Google's vision for Search in an AI era is starting to come together.
It's rolling out its AI Mode to users in the US, giving them a full AI search experience.
At its annual I/O event, the company also announced several new search tricks it has planned.
We're finally getting a good glimpse of what Google looks like when it's fully transformed by AI.
Last year's I/O brought an onslaught of AI announcements and a sense that the search giant was trying to prove it still had the juice to lead this race, with a lot of disparate products and demos. This year, a clearer picture is emerging of how Google sees the future of its core products, including what CEO Sundar Pichai called a "total reimagining" of Search during a roundtable with the press ahead of the event. This includes a more conversational-type search called AI Mode and eventually an artificial intelligence assistant that understands the world around you.
Google has faced a major dilemma: Search advertising generates the lion's share of the company's multibillion-dollar business, though it knows it can't just stay still and let rivals eat its lunch. It's trying to build AI into its main product before someone else does it better. But not so fast that it risks hurting the company's profit engine.
The company has been inching forward with AI Overviews, and this week it's rolling out its AI Mode to everyone. While AI Overviews gives a response summary at the top of the normal search page, AI Mode allows users to click a new tab, which opens a conversational-type experience that surfaces a more diverse variety of sources, all still based on Google's search index. Users can also ask follow-up questions.
"AI Mode is not just this AI-powered experience end to end, but it also is a glimpse of what's to come in Search overall," said Google Search's head, Liz Reid.
AI Mode uses what Google calls a "query fan-out" technique, which means it runs multiple queries simultaneously and returns the results all at once. Google says it will make searching better and allow users to ask more complex questions.
The feature today is just the start of how Google sees search evolving. Google is announcing a bag of new tricks that it's keeping in Labs for now, so they'll only be available to early testers. Still, they show what Google sees as the future of search.
One example is Deep Search, which lets users punch in a superlong and complicated question and returns a fully cited report, much like Google's Deep Research feature in Gemini. There's also a version that returns real-time data and visualizations (think charts on sports teams' statistics).
Google is also set to let users give AI Mode access to other Google apps and their search history so it can return more tailored answers and recommendations.
Reid said Google would feed some of the features from AI Mode into its standard search engine and AI Overviews, the idea being that Google's standard search experience benefits from the leaps it's making in the underpinning AI models.
"You put all of this together. This really is building the future of search," Reid said. "Searching starts to feel effortless."
Does Google envision AI Mode being the default one day? That's the implication here, though the company will closely watch over the next few months to see just how many people click the "AI Mode" tab.
The everything assistant
Google also has a vision for an AI assistant that's with you all the time.
If you've seen Google's Project Astra, an AI agent that uses vision to see the world around it, you already have a good idea of what Google is thinking here. It wants to build an assistant that's with you anywhere — be it in your phone or in a pair of augmented reality glasses — and can see the world, answer questions, and relay information to you in a matter of seconds. Or maybe it's just helping you code.
At I/O, Google is announcing it's extending its frontier Gemini 2.5 Pro model to be a "world model," which really just means it's going to be able to understand what it's seeing and, Google says, make plans. In AI speak, it's becoming more agentic.
Google DeepMind CEO Demis Hassabissaid these updates were "critical steps" toward building a "universal AI assistant" that can better understand the user and take actions on their behalf.
"This is our ultimate goal for the Gemini app: an AI that's personal, proactive, and powerful," Hassabis added.
Google is set to make its camera-enabled and screen-sharing Gemini Live available to everyone with the Gemini App and launch Veo 3, a video generation model that includes support to combine sound effects.
It needs to build fast here. While generative AI is not yet a critical business in the way Search is, the company said its Gemini app had more than 400 million monthly active users. Google's own internal analysis found that Gemini still trailed OpenAI's and Meta's apps as of earlier this year, according to documents shown in court.
A lawsuit over the death of a pilot on one of Sergey Brin's planes has ended in a settlement.
Two pilots died after a plane belonging to the Google cofounder crashed off the California coast.
Another lawsuit related to the crash was voluntarily dismissed in January.
A lawsuit over the fatal crash involving one of Google cofounder Sergey Brin's planes has ended in a settlement.
Two pilots, Dean Rushfeldt and Lance Maclean, died in May 2023 after a plane belonging to Brin crashed off the coast of California. A lawsuit filed last July by the family of Rushfeldt claimed the crash was the result of a failure in the plane's fuel equipment.
The settlement, filed in the Superior Court of Santa Clara County on May 9, is "pending the exchange of closing documents and disbursement of the settlement fund," per a filing with the court. The parties anticipate that the settlement will be completed in about 60 days, the filing said.
The family of Rushfeldt filed suit against Brin, his private family office Bayshore Global, and several related entities, alleging multiple counts of negligence that led to the fatal crash. The suit offered a rare glimpse into the highly secretive inner workings of Brin's personal life and family office.
Google, a defendant in the original complaint, was also named in the settlement. The plaintiff filed a request in February to dismiss the company from the suit.
Jessica McBryant, an attorney for Rushfeldt's family, declined to comment via email, citing a "confidentiality agreement." Google and a spokesperson for Brin's family office did respond to requests for comment.
The pilots who died were enlisted to fly the plane from Santa Rosa, California via Honolulu to Brin's private island in Fiji, the complaint read. However, the plane ran out of fuel around 30 miles off the coast of California after a fuel bladder malfunctioned.
The plane crashed into the Pacific Ocean, and Rushfeldt and Maclean were found dead inside. Their bodies could not be recovered before the plane sank, the complaint said.
In the complaint, Rushfeldt's family alleged that Bayshore Global and a related corporate entity, Seafly, which maintained the plane, made several errors in the aircraft's maintenance, including installing the fuel bladder "from memory" instead of following a checklist and not properly logging the alterations.
The widow of Maclean, Maria Magdalena Olarte, filed a separate lawsuit against Brin and Bayshore in February 2024. The plaintiff filed to dismiss the case with prejudice in early January. In a filing in November 2024, the parties said they were "attempting to resolve the matter" via negotiations. It's unclear if a formal settlement was reached, and an attorney for Olarte did not respond to a request for comment.
Google stock tumbled after Apple senior vice president Eddy Cue said Safari searches had dropped.
A filing reveals another reason Google watchers worry about search: slowed growth in paid clicks.
Some analysts are split over whether Google's Search empire is under threat.
Apple senior vice president of services Eddy Cue set alarm bells ringing on Wednesday after dropping a bombshell at Google's antitrust trial: Google searches through Safari dropped in April for the first time ever.
While his comments triggered a frenzied sell-off in Google stock, it might not be the only reason the company's watchers should be concerned about Google's ability to keep full control over the search market.
A little-noticed number in Google's latest financial disclosure may be the realest sign yet that investors have reason to worry.
After reporting blockbuster Q1 earnings last month, Google revealed in a 10-Q filing with the SEC that paid clicks for the quarter grew 2%, down from 5% growth in the same quarter a year ago. That's the slowest growth rate since the company began reporting the metric.
Paid clicks are exactly what they sound like: people click on ads across Google Search and other services such as Google Play and Gmail. Each click translates to money in Google's pocket.
Why those paid clicks are down, exactly, Google hasn't said.
"It's possible macro played a role, or searches with AI overviews delivered better results, requiring fewer 'paid clicks' to get to conversion," Bernstein analysts wrote in a note published Wednesday. "But mostly, it's a worrying KPI."
The analysts said they believe the timing of the dip, combined with Cue's comments and surging numbers of ChatGPT and Meta AI users, suggests that Google's control of the search market may be lower than previously believed.
"Combined, we estimate Google's search share is closer to 65-70% vs. the 90% we often hear," they wrote.
Google declined to comment.
Google's slice of pie
Google insists that it's seeing more searches than ever.
Since the 2000s, the company has managed to harvest vast amounts of searches by paying Apple a fee to make its search engine the default on Apple's Safari web browser. As recently as 2022, Google had paid Apple at least $20 billion — a massive fee that signals how much value Google sees in having Apple users turn to its search engine for all their queries.
It maintains that this partnership continues to drive growth in searches. Cue's comments were provocative enough to prompt the search giant to issue a public statement stating that it continues to see "overall query growth" in Search, including an increase in total queries coming from Apple.
There's little doubt among industry watchers that the overall search pie is growing — though figures from research firm Statcounter suggest Google's control of the global search has fallen slightly. The big question is whether Google's slice of that pie is shrinking relative to rivals.
According to Statcounter, Google's share of global search traffic fell to 89.71% in March 2025, down from about 91% in March 2024 and about 93% in March 2023.
Meanwhile, competing search products are growing. In April, OpenAI said that around 10% of the world uses ChatGPT, which would be at least 800 million users. Meta also said that about 1 billion people use AI across its various products.
The search market expands with AI as chatbots and generative tools expand the definition of search. Google could reap the rewards here, though this also creates an opening for competitors charging as fast as possible to stay ahead of the search giant.
Bernstein analysts estimated that generative AI queries that run through chatbots such as ChatGPT are reaching volumes close to 15% of the queries processed by Google and other traditional search engines.
Analysts are split
Other analysts are divided on just how much of a threat Google's search business faces.
For instance, longtime Apple analyst Ming-Chi Kuo took to X on Wednesday to explain why he felt it was a mistake to think generative AI would not affect Google's advertising business.
He said that despite the "continued growth of Google's advertising business," the company hasn't had much competition yet.
"GenAI service providers have not launched advertising businesses, so Google Ads remains the best choice for online advertisers," Kuo wrote.
The following statement by Apple’s senior vice president of services, Eddy Cue, implies that Google search and advertising business are facing potential threats from generative AI (GenAI): Cue noted that searches on Safari dipped for the first time last month, which he attributed…
Kuo likened Google's situation to the one Yahoo faced during the 2000s. The company's advertising business, launched in 1995, only started declining in 2008, despite newfound competition from Google's AdWords business arriving back in 2000.
Analysts at investment bank Jefferies have a different view. In a research note on Wednesday, the analysts had a particular word to describe the roughly $155 billion sell-off in Google's stock following Cue's comments: "overblown."
While they acknowledged that Google's AI-powered "Overviews" feature may act as a headwind right now as it is resulting in "fewer searches," they said Google will "be able to ramp monetization" of its AI summary feature over the long run.
They also don't see a scenario where Apple shifts away from Google and causes as much harm as investors might think.
"While Safari is significant, it does not represent the entirety of search activity; iOS accounts for 18% of operating systems, and Safari holds 17% of the browser market share compared to Chrome's 66%," the analysts wrote.
Tech firms like Amazon, Google, and Microsoft are doubling down on performance-based rewards.
Top performers are better incentivized, while underachievers face increased scrutiny.
Tech companies are also reducing new-hire offers and lowering pay raises.
Big Tech is rewriting the rules this year: Top talent gets big rewards, while underperformers face the heat.
The latest to follow suit is Amazon, which recently overhauled its compensation model to reward long-running top performers with larger payouts. Four years at the highest performance rank now pays off big with 110% of the pay band, which exceeds the previous 100% cap. But employees who achieve the top rank for the first time take a hit, with total comp cut to 70% of the pay range, down from last year's 80%.
The move is part of a broader tech trend of using performance reviews and compensation to better incentivize high achievers and eliminate underperformers and their drag on the balance book. Big Tech firms, including Google, Microsoft, and Meta, have made similar changes, all in pursuit of leaner, top-tier teams.
These companies handed out lavish perks, pampering, and sky-high salaries to stay competitive. A pandemic-era hiring spree exacerbated a tech tradition of hoarding top talent and only added to the ranks. Now, a shift makes one thing clear: It's no longer about head count — it's about talent.
The changes go hand in hand with sweeping pivots across tech to replace perks and pampering with performance and efficiency. Big Tech still rewards workers handsomely, but the era of resting and vesting at the giants may be well and truly over.
"The industry overall has made big efforts to do more with less," Brian Nguyen, a product manager at the salary data platform Levels.fyi, told Business Insider. "This has allowed them to scale back costs while pushing for higher-quality work with fewer employees and rewards for only the top-tier performers."
Google declined to comment, and Microsoft did not respond to a request for comment.
A Meta spokesperson reiterated what the company previously shared with BI about its low-performer cuts: "Employees at Meta have always been held accountable to a goal-based culture of high performance."
The new approach "creates a steadier compensation progression for employees as they develop within their roles," an Amazon spokesperson said, adding the company offers "multiple channels" for employees to raise concerns about pay.
'Appropriate distinction'
Amazon managers this year received a list of potential pay-related questions and suggested answers, according to internal talking points obtained by BI. If asked about pay cuts, managers are encouraged to focus on employees getting raises for improved performance.
"Employees moving to a higher OV level will generally receive an increase," one talking point said, referring to Amazon's performance rating called overall value. "This ensures an appropriate distinction between newer and consistently high-performing employees at each OV level."
At Google, a larger share of employees can now earn top performance ratings, which lead to bigger bonuses and equity. Managers have also been given the authority to hand out more discretionary rewards to employees who are excelling.
But with a fixed budget, those gains come at the cost of lower-ranked employees, reinforcing the company's focus on high performance and raising the stakes for those falling behind.
"High performance is more important than ever to achieve the goals we've set, and so we're making some changes to further reward top contributors," John Casey, Google's vice president of global compensation and benefits, told staff in a memo last week.
Microsoft, meanwhile, has rolled out policies aimed at enhancing performance management and addressing what the companycalls "low performers." Managers also received "more transparency and clarity" in the rewards process, an internal note seen by BI said. The change came after the company laid off 2,000 employees deemed to be underperforming earlier this year.
Meta has ramped up what it calls performance-based cuts, aiming to shed about 5% of its lowest-rated staff. An internal memo revealed plans to make these layoffs annual under a "non-regrettable attrition" policy. Adding to the sting, some ex-employees — even high performers — have landed on internal block lists, barring them from being rehired.
Lower pay for new hires
While these companies are weighing an employee's performance rating history more heavily, some are also reducing new-hire offers.
In 2022, Amazon regularly offered over $300,000 a year to newly hired midlevel software engineers, according to data from Levels.fyi. Nowadays, more offers land around $270,000 a year, with recruiters being especially strict about needing "extra approvals" for anything higher, said Levels.fyi's Nguyen. Other companies, like Meta and Apple, are taking a similar approach, he added.
Switching jobs no longer helps either.
The Atlanta Fed's Wage Growth Tracker shows that median pay increases for job switchers dropped to 4.2% in February, down sharply from 7.3% in early 2023.
Tech professionals, in particular, are feeling the heat with lower pay offers and significant pay cuts, BI previously reported, as reduced competition has weakened employees' bargaining power.
Not everyone is sold that the shift to crack down on underperformance is here to stay.Peter Capelli, a management professor at the University of Pennsylvania, told BI that many companies follow these trends simply because "others are doing them," cycling through similar changes every decade or so.
"These are just fads," Capelli said.
Still, the changes haven't gone unnoticed. At one Amazon office, some employees quietly voiced their frustration on an elevator whiteboard, a photo of which was shared with BI.
"No raise, no RSUs — thanks!" the whiteboard read, referring to the lack of employee stock awards. "Do more with less."
Elevance Health's acquisition of Verily's insurance business has closed.
Verily's CEO told staff in a memo that the sale will strengthen the company's financial position.
Verily has been on a mission to shed projects and pivot to AI as it eyes life beyond Google's walls.
Alphabet life sciences unit Verily has officially sold its insurance business to Elevance Health, its CEO told employees last week, as the company continues to shed projects and focus on AI.
In an April 30 memo to staff, which was seen by Business Insider, Verily CEO Stephen Gillett said Elevance's acquisition of its stop-loss insurance subsidiary, Granular, had officially closed.
"This acquisition represents a strategic win for Elevance Health, Granular, and Verily," Gillett wrote in a note to staff.
"On our end, this sale strengthens Verily's financial position and allows us to reinvest the proceeds towards Verily's strategic priorities," he later added.
BI first reported in February that Verily had entered an agreement to sell Granular to insurance provider Elevance Health. A Verily spokesperson referred BI to a previous statement confirming the company's agreement to sell Granular.
Companies sometimes take out top-loss insurance to pay their employees' medical bills and limit their financial exposure, potentially protecting them from sharp increases in spending. Granular launched in 2020 and used "proprietary technology" for its services, the company said.
With its insurance business sold, Verily has shed another project as it aims to further streamline itself and refocus its strategy around AI.
Verily sits among the "other bets" owned by Alphabet but live outside Google.
It started out in Google's moonshot lab in 2015 and focused on an array of projects, including wearables and surgical robots. It has been criticized for entertaining too many bets and lacking a clear focus. The unit completely pivoted to COVID-19 screening and testing during the pandemic. Last year, Verily announced a new AI-powered chronic care product named Lightpath, designed to help patients living with conditions such as diabetes and obesity.
Verily is also eyeing a future beyond Alphabet's walls. In January, the life sciences group separated some of its internal systems from Google's, such as certain employee benefit systems. Last year, Verily issued employees new laptops, office badges, and email addresses that are no longer aliases of the Google email domain, multiple people familiar with the matter told BI.
Verily is also looking to raise another round of capital in the next few months, BI previously reported.
Google CEO Sundar Pichai speaks at a Google I/O event in Mountain View, California.
Jeff Chiu / AP Photo
Google has a message for employees: higher performers will be rewarded.
The company is changing how employees are ranked in their yearly performance reviews.
"High performance is more important than ever," Google's head of compensation told staff.
Google is making changes to its compensation structure in a bid to incentivize higher performance from staff.
The company announced on Tuesday that it will change how employee performance ratings work. The change gives more staff the opportunity to achieve one of the highest scores in their annual review and, therefore, higher compensation, while lower performers receive smaller bonuses and equity.
In an email to staff with the subject line "Strengthening our performance culture," John Casey, Google's vice president of global compensation and benefits, said that managers will be able to allocate the "Outstanding Impact" score to more employees than previously.
"This means more Googlers will have the opportunity to achieve that rating during annual reviews, and their bonus and equity award will be modeled using the O's individual multiplier in 2026," he wrote in the email, which was seen by Business Insider.
Googlers' performances are rated once a year using an internal system known as Googler Reviews and Development (GRAD). An employee's impact is scored on a scale, with "not enough impact" being the lowest and "Transformative Impact" being the highest. The ranking a Googler is given usually determines their bonus and equity.
Most Googlers fall into the second-highest bracket of "Significant Impact." The "Outstanding Impact" bucket captures a smaller number of higher performers, while the "Transformative" bucket is reserved for a very small number of excelling Googlers.
Casey said Google will also increase the discretionary budget it gives to managers so they can dish out more rewards to high performers who fall within that "Significant Impact" bracket.
He added that the changes would be "budget-neutral," meaning employees with lower ratings could receive smaller bonuses and equity in their compensation packages as more of the budget is allocated to the higher brackets.
"We want to be upfront that to fund this we'll be slightly reducing the bonus and equity individual multipliers for Significant Impact and Moderate Impact ratings," Casey told staff. "It's important to note that Significant Impact will remain a strong rating — achieving it will still get you more than your target bonus."
Google spokesperson Courtenay Mencini told Business Insider in a statement, "We're making these changes to further reward top performers and continue our momentum across the company."
The changes at Google come amid a broader shift across Big Tech to run their businesses more efficiently and push employees to perform better. Microsoft recently rolled out new policies aimed at dialing up performance pressure on employees, while in January, Meta announced it would cut 5% of its workforce, focusing on low performers.
While Google hasn't gone to such lengths itself, the changes to its performance ratings are designed to push staff to work harder and aim higher.
"High performance is more important than ever to achieve the goals we've set," wrote Casey in the email to staff, adding that the changes were being made to "further reward top contributors" at the company.
Read the full email sent to Google employees:
Hi Googlers,As recent moments like the Gemini 2.5 Pro launch and Cloud Next have shown, there's incredible momentum across the company right now — it's so exciting to see Googlers pull together to deliver on our ambitious product roadmaps. High performance is more important than ever to achieve the goals we've set, and so we're making some changes to further reward top contributors, in all teams across the company.First, we'll increase the ratings distribution guidance we give to leaders for Outstanding Impact (O). This means more Googlers will have the opportunity to achieve that rating during annual reviews, and their bonus and equity award will be modeled using the O's individual multiplier in 2026.Second, we're increasing the discretionary budget we give to managers so they can further reward high performers within the Significant Impact rating.We want to be upfront that to fund this we'll be slightly reducing the bonus and equity individual multipliers for Significant Impact and Moderate Impact ratings. It's important to note that Significant Impact will remain a strong rating — achieving it will still get you more than your target bonus.The above changes are budget-neutral, and overall we're continuing to invest in comprehensive and highly competitive compensation and benefits.These changes apply to the end-of-year reviews and 2026 compensation planning.John Casey
Jerry Dischler, president of cloud applications at Google.
Google
Google Cloud executive Jerry Dischler told staff on Monday he plans to leave the company.
Dischler was leading cloud applications at Google.
Dischler previously ran Google's powerful ads business and spent nearly 20 years at the company.
Jerry Dischler, a Google veteran who spent several years steering the company's crucial advertising business, plans to depart Google.
Dischler, who has been at Google for almost 20 years, announced his departure in a memo to staff on Monday, which was seen by Business Insider. A Google spokesperson confirmed the departure.
Dischler was most recently president of cloud applications, overseeing Google's Workspace office software product and integrating AI tools into customers' businesses. He joined Google in 2005 and worked on technology that eventually became Google Pay. He later ran Google's entire ads operations.
"The most difficult aspect of my decision was stepping away from the incredible opportunity we have before us," he wrote in the email, adding that he did so with "immense confidence" in the teams.
Dischler's exit marks another notable departure for the teams working on Google Workspace, a critical product for Google in generative AI that competes with Microsoft's productivity tools.
Aparna Pappu, former vice president and general manager of Google Workspace, announced late last year she would step down and hand over the reins to Dischler.
In his departing note, Dischler wrote that senior managers on Workspace will report to Google Cloud CEO Thomas Kurian, effective May 9, until a new leader is named.
Dischler wrote that it was time for him "to explore something new," although he did not say if he was leaving for a new opportunity.
Chorus has built tech to help companies track their goods in the supply chain.
Alphabet
Chorus, an Alphabet moonshot for getting a real-time view into the supply chain, is spinning out.
Series X Capital led its funding round, with Alphabet also taking a stake.
X said that spinning out Chorus will help it make its services more generally available faster.
Chorus, a project focused on making the global supply chain more efficient, is spinning out of Alphabet's moonshot incubator and into an independent company.
Chorus was formed several years ago inside Alphabet's famed X moonshot lab, which tries to make technological breakthroughs by pursuing hugely ambitious and often zany ideas.
X — formerly Google X — has spun up projects such as Alphabet's self-driving company Waymo and its drone delivery business Wing.
Chorus sells technology to businesses that give them a real-time view of what's happening with their goods. It has spent several years maturing as a business inside Alphabet's X incubator, but flying solo should help Chorus move faster, the companies believe.
"The next big milestones are making our visibility, condition monitoring and remote inventory management services generally available," Suresh Vishnubhatla, Chorus CEO, told Business Insider.
As part of the spin-out process, Series X Capital led a funding round for Chorus for an undisclosed amount, which included a stake from Alphabet.
Series X Capital, which has a limited public profile, is run by Gideon Yu, the former YouTube and Facebook CFO and San Francisco 49ers co-owner, according to corporate filings reviewed by Business Insider and a person with direct knowledge of the matter. Yu and representatives for Series X Capital did not respond to a request for comment.
Chorus develops tools that track how goods move through the supply chain and sells sensors that businesses attach to their products. A machine learning system that runs in the cloud then pumps all that data to a dashboard. The goal is to give businesses better and more up-to-the-minute information about their items, whether it's the condition and temperature of their product or data on exactly when a shipment will reach its destination.
Chorus has its roots in an earlier X initiative known as Project Delta. This initiative focused on tackling food waste by building a system that connected grocers, food banks, and logistic companies in an effort to match food supply with demand. While Google later absorbed some of that technology, what remained inside X became Chorus.
The company aims to tackle the billions of dollars of cargo that go missing each year and prevent pharmaceuticals from spoiling due to temperature issues. During the COVID-19 pandemic, Chorus worked with New Zealand's Healthcare Logistics to deliver 11 million COVID-19 vaccinations to the country, Alphabet said.
Chorus is the latest in a series of spin-outs from X, which, as Business Insider previously reported,has put more focus on finding outside investment for its projects. Last month, Alphabet announced that Taara, a project using lasers to beam internet, would also spin out and be backed by both Alphabet and Series X Capital.
In an email interview, X chief Astro Teller told Business Insider that spinning out some of the incubator's best projects can help them scale up faster with partners that bring the right industry expertise.
"This next stage of work — bringing products to market and scaling a business — can best be done outside of the prototyping-focused environment of X," said Teller.
While Alphabet has previously chosen to graduate some X projects into businesses that stay within Alphabet's walls, turning to outside investment also offers more paths to incubated projects at X, many of which have historically never made it out of the lab.
"This provides a clearer growth path for our projects and enables us to focus on what we do best — early-stage invention," said Teller.
Google DeepMind uses noncompetes that last as long as a year to stop employees from going to rivals.
A former DeepMind director commented publicly about the issue last week.
Several factors, including an employee's seniority and work focus, affect the noncompete length.
The battle for AI talent is so hot that Google would rather give some employees a paid one-year vacation than let them work for a competitor.
Some Google DeepMind staff in the UK are subject to noncompete agreements that prevent them from working for a competitor for up to 12 months after they finish work at Google, according to four former employees with direct knowledge of the matter who asked to remain anonymous because they were not permitted to share these details with the press.
Aggressive noncompetes are one tool tech companies wield to retain a competitive edge in the AI wars, which show no sign of slowing down as companies launch new bleeding-edge models and products at a rapid clip. When an employee signs one, they agree not to work for a competing company for a certain period of time.
Google DeepMind has put some employees with a noncompete on extended garden leave. These employees are still paid by DeepMind but no longer work for it for the duration of the noncompete agreement.
Several factors, including a DeepMind employee's seniority and how critical their work is to the company, determine the length of noncompete clauses, those people said. Two of the former staffers said six-month noncompetes are common among DeepMind employees, including for individual contributors working on Google's Gemini AI models. There have been cases where more senior researchers have received yearlong stipulations, they said.
"Our employment contracts are in line with market standards," a Google spokesperson told Business Insider in a statement. "Given the sensitive nature of our work, we use noncompetes selectively to protect our legitimate interests."
As the AI field has blossomed in the past two years with new startups and opportunities at heavyweight tech companies, some DeepMind employees feel their lengthy noncompetes have restricted their movement.
"Who wants to sign you for starting in a year?" said one former DeepMind employee. "That's forever in AI."
Noncompete laws in the US vary by state, though noncompete clauses are unenforceable in California, where Google and several other tech giants are headquartered. New legislation introduced in 2023 expanded California law to ban the enforcement of noncompetes entered into outside the state.
In the UK, where DeepMind's headquarters is located, noncompetes are enforceable if they are deemed reasonable to protect the employer's legitimate business interests.
It can pose a serious challenge to some talent at one of the world's leading AI labs seeking career opportunities elsewhere during a boom period for the industry, particularly as some of DeepMind's competitors, such as OpenAI and Microsoft, grow their UK offices and try to poach staff.
One former DeepMind employee said they were aware of colleagues who considered leaving London for jobs in California just to get out of the noncompete.
'It's abuse of power'
The issue of noncompetes spilled into the public domain last week when Nando de Freitas, Microsoft AI vice president and former DeepMind director, posted a message to DeepMind employees on X.
"Every week one of you reaches out to me in despair to ask me how to escape your notice periods and noncompetes," he wrote. He added that employees unhappy about the terms should reach out to DeepMind leaders, including CTO Koray Kavukcuoglu and senior research director Douglas Eck, whom de Freitas said are "against it."
"Above all don't sign these contracts," de Freitas wrote. "No American corporation should have that much power, especially in Europe. It's abuse of power, which does not justify any end."
Dear @GoogDeepMind ers, First, congrats on the new impressive models.
Every week one of you reaches out to me in despair to ask me how to escape your notice periods and noncompetes. Also asking me for a job because your manager has explained this is the way to get promoted, but…
In the past, the idea of receiving full compensation without working for several months didn't seem so bad to some staff, though in the red-hot fight for talent now happening in the AI field, it risks becoming a problem.
"It's becoming less popular now because there's loads of cool startups that would not be willing to wait six months-plus, so people end up missing out on some good opportunities," said one former DeepMind employee.
A former Google employee said the noncompetes prevalent in the generative AI boom are a stark contrast to those seen in the tech industry in the previous decade, when "people working on some of the highest value systems in the world" could more readily take a job offer elsewhere without being bound by such agreements.
The former employee also drew parallels between noncompetes in AI and those seen in hedge funds, where clauses have proven notoriously aggressive.
"AI is interesting. It seems to be the first time in my career that you have this insane race, like a space race," the former employee said. "People really feel like to be six months ahead, a year ahead, could make all the difference."
Some Google employees said they saw smaller pay bumps in compensation packages for 2025.
It was raised in an all-hands meeting this week.
Google's benefits chief said staff in some locations saw smaller equity refreshers.
Google employees recently got their new compensation packages, and some say they've once again seen smaller pay bumps.
The issue was raised in a company all-hands on Tuesday, according to a transcript of the meeting shared with Business Insider.
At Google's monthly TGIF (Thank God It's Friday) meetings, employees can submit questions and vote for the ones they want to ask the most. One of the top-voted questions this week asked why some employees had seen a decrease in their refreshed stock grants and overall compensation despite Google's recent strong financial performance.
John Casey, Google's vice president for global compensation and benefits, told employees that over 80% of employees saw a year-on-year increase in their 2025 compensation. However, he added that some Googlers in less technical roles in some locations received smaller equity packages this year to calibrate pay to local markets.
Google's compensation packages typically consist of a base salary, equity awards, and, in some cases, bonuses. The actual amount a Googler makes varies based on performance.
Casey said that Google wanted to pay employees who make an impact and that the compensation structure was set up to reward high performers.
A Google spokesperson told Business Insider that its 2025 compensation cycle was similar to last year's and that Google continues to align its pay to local markets overall.
In emails sent to staff Tuesday, Google's top bosses explained why the search giant plans to spend $32 billion to acquire the cybersecurity startup, which marked the year's biggest deal so far — and the biggest deal of all time for Alphabet.
"We don't make deals like this every day, so I wanted to share a little more about why, and why now," Google CEO Sundar Pichai wrote in an email, before summarizing some of what was included in the company's public blog post about the deal on Tuesday.
Pichai also forwarded an email from Google Cloud CEO Thomas Kurian. Kurian wrote that the announcement comes at a "critical time" for Google and its partners as customers are "accelerating" their move of data to the cloud.
"Most deployments are now multicloud or hybrid — introducing complex management challenges," he wrote in the memo, which was seen by Business Insider.
Multicloud, where organizations use services from more than one cloud provider, has become increasingly important to Google's strategy in the artificial intelligence boom.
Kurian wrote to staff that "customers are broadening their use of software and need to integrate their cybersecurity tools with their software development tools to protect and remediate vulnerabilities."
He also added that "the growing adoption of AI has accelerated the threats both to and from AI models."
"Together, we believe Google Cloud and Wiz will vastly improve how security is designed, operated and automated in the future," Kurian wrote.
Matt Renner, Google's president of global revenue, also emailed staff acknowledging the "rising number" of cybersecurity breaches.
"We believe Wiz and Google Cloud together will accelerate organizations' ability to improve security, lower the cost to do so, and ultimately spur the adoption of multicloud and cloud computing for customers, partners, and the industry as a whole," he wrote.
This week's announcement followed a second run-up at the deal, after Google and Wiz entered talks last year for a $23 billion acquisition that ultimately fizzled out. Wiz, which builds technology that scans data companies put onto the cloud and identifies security risks, would continue to be available through other cloud providers such as Amazon Web Services and Microsoft Azure, the companies announced on Tuesday.
In a note this week, Bernstein analysts wrote that acquiring Wiz "plugs a need for an Enterprise-credible cloud security platform," which Google "currently lacks" compared to rivals Amazon and Microsoft.
In a briefing held Tuesday, Google CFO Anat Ashkenazi said the company expects the Wiz acquisition to close in 2026, pending regulatory scrutiny.
"Until the acquisition closes, Google Cloud and Wiz remain independent companies in all respects and it is crucial that our teams are aware of and follow this guidance," Renner wrote in his memo.
A trifecta of economics, pandemic-era realities, and political pressure is changing tech culture.
From Big Tech to Silicon Valley startups, companies are pushing to "do more with less."
Leaders aren't shy about wielding their power and aligning around a hard-driving strategy.
For years, Shopify CEO Tobi Lütke enjoyed a reputation for growing the $125 billion e-commerce company without working the grueling hours expected of startup founders.
"My job is incredible, but it's also just a job. Family and personal health rank higher in my priority list," he wrote in a now-deleted post on X, then Twitter, shortly before the pandemic, as reported by Business Insider at the time. "The only times I worked more than 40 hours in a week was when I had the burning desire to do so."
This year, even Lütke appeared to change his tune.
"I'm at home for dinner but I work at least 10 or so hours a day and a lot of the weekend," Lütke wrote on X. He was responding to a user who called him a "counter-example" to a meme suggesting you can't have work-life balance and a breakthrough startup. "I don't want people to get misguided by this meme."
Across tech, the tables have turned for employees as performance pressure and proclamations of "efficiency" and "intensity" replace perks and pampering. Sweeping layoffs have become the norm in an industry that, in recent memory, enjoyed job security. The pressure to dominate in AI has created intense competition, as companies use the technology to do more with fewer workers. Already hard-driving workplaces have become even harder.
While the situation for tech employees has been changing since the pandemic boom ended in 2022, more recent developments include a decidedly different tone from executives. Now, companies aren't just making these changes; they want to be seen making them.
Meta earlier this year said it was cutting 4,000 employees deemed low performers as CEO Mark Zuckerberg said the "culturally neutered" corporate world had gotten away from "masculine energy." Amazon insisted that employees return to the workplace every weekday, a policy some employees say is stricter than before the pandemic.
Other companies have cracked down, too. Microsoft, which was once referred to as a "country club" for its relatively lax culture, cut 2,000 employees as it overhauled its review process to eliminate underperformers more quickly.
Google, which practically invented tech perks like free lunch, started an "efficiency drive." Its cofounder Sergey Brin, who had stepped away from leading Google but now often shows up to work on the company's Gemini AI models, recently recommended that employees working on its Gemini tools should work 60 hours a week and go into the office "at least every weekday." Wall Street has rewarded this rigor, as stock prices of Meta, Amazon, Microsoft, and Google's parent, Alphabet, have surged since 2022.
Startups also see a trickle-down effect from Big Tech companies' pressures. Krish Ramadurai, a partner at AIX Ventures, said he had noticed a "pronounced shift" toward leaner teams and rigorous performance standards at startups.
Between performance-based cuts, return-to-office mandates, and the stripping of workplace perks, it's clear not only that the tech industry is done coddling employees, but that companies want to send the message those days are over. BI interviewed employees from tech giants, including Microsoft, Google, Amazon, and Meta, as well as various tech startups, about the changes. Some spoke on the condition of anonymity since they're not authorized to talk to the press, though their identities are known to BI.
Meta, Microsoft, Google, and Amazon did not comment. Shopify did not respond to a request for comment.
From comfy to collapsed
For years, fierce competition for tech workers meant companies spoiled employees with astonishing salaries and swanky perks, such as in-office massages and free food cooked by fancy chefs.
By 2022, tech companies seemingly couldn't throw enough money at workers. Early that year, Amazon more than doubled its maximum base salary, and Microsoft gave across-the-board raises to employees up to a certain level of seniority to dissuade them from leaving for competitors.
As the pandemic boom ended, tech stocks plummeted, and interest rates increased through 2022. This prompted an efficiency drive by many companies as investors demanded profitability over growth at all costs.
Also that year, companies watched the billionaire Elon Musk's handling of the Twitter acquisition, in which he cut thousands of employees, plus perks like free lunches, and demanded a commitment to a new "extremely hardcore" vision and "long hours at a high intensity." At one point, Twitter workers were begging on Slack for toilet paper and clean bathrooms amid Musk's drastic cost-cutting.
As of late last year, Fidelity valued X at only about 20% of the $44 billion that Musk bought it for in 2022. Still, his approach may have expanded what the tech industry thought possible in terms of workforce and cost cutting.
"People paid attention because the prevailing wisdom was you couldn't take out that much of an engineering organization and put that much instability on it and not have it fall over," Brad Porter, the founder and CEO of Cobot, told BI. "It did come close to falling. He pushed right to the edge of it actually falling over, but it didn't fall over."
'Do more with less'
By the end of 2022 and in early 2023, tech giants had conducted unprecedented rounds of layoffs. Meta, Amazon, Google, and Microsoft collectively laid off more than 60,000 employees during that time.
Layoffs have remained at a steady drip across the industry since. Such cuts have become so frequent at Google, for example, that employees have taken to crowdsourcing information on layoffs in an internal Google Doc.
Employees told BI about the pressure across the industry to "do more with less." "There's lots of uncertainty," one longtime Amazon employee said, "and lots of pressure to perform the jobs of multiple people at the mercy of ruthless middle management."
Tech companies are also culling middle management layers. Amazon in September announced a plan to increase the ratio of individual contributors to managers by 15% by the end of this month. In December, CEO Sundar Pichai told his staff that Google had cut vice president and manager roles by 10% as part of its efficiency drive. Microsoft also monitors what it calls "span of control," tracking the number of reports per manager.
Performance pressures
Amid the cuts, employees across the industry say companies are dialing up the performance demands.
Meta told its staff in January that it would eliminate roughly 5% of its workforce, or about 4,000 employees, to "raise the bar on performance management," as Zuckerberg wrote in an internal memo.
Google also increased pressure on employees. Perhaps most telling was Pichai's December comments attempting to clarify what "Googleyness" means for a modern Google. Once a squishy and vague philosophy for the search giant's corporate culture, Pichai said he believed it now meant, among other things, being "mission first."
"There is more pressure for individuals to be better in their roles, and there is much more aggressive performance management happening these days," a longtime Google manager said.
"We're being asked to do more for less," said another current longtime Google employee.
That same Google employee said that Silicon Valley had been moving toward more ruthless, efficient workplaces for a while — and that the current political climate "gives them the green light to do it openly." Google has been working to become more efficient since its chief investment officer Ruth Porat joined the company as CFO from Morgan Stanley in 2015, "but now the masks are off," the person said.
Microsoft was once referred to as the tech industry's "country club," meaning a place employees would go after they were done working hard in their careers and wanted to coast before retirement. A change this year shows how far Microsoft has shifted when it fired 2,000 employees deemed low performers without severance and ended their health benefits the same day. This kind of performance-based mass cut showed a shift for the tech giant.
One longtime Microsoft senior-level employee said they felt that the "culture shifts toward firmer performance expectations" at peer tech companies like Google, Meta, and Amazon made it more acceptable for Microsoft to do the same.
At TikTok, the pressure to perform jumped last year after the company directed managers to deliver more low scores in performance reviews, leading to PIPs and eventual exits. At the same time, six current and former employees told BI their goals had become much harder to hit. One staffer called the goals "unattainable."
The company has also recently heightened RTO requirements for some teams. In February, it told its US e-commerce workers that in addition to being in the office five days a week, they would physically need to be in the building for eight hours a day. Ten current and former workers told BI that burnout had become common, leading to some going on mental health leave to get a break. TikTok did not respond to a request for comment.
"You feel like if you're not hitting a target, even if it's a moving target, you're in trouble," a former staffer who went on leave for mental health reasons told BI. "For me, it was just feeling like a failure, like I couldn't do anything right."
It's gotten hardcore in the 'valley of death'
The increasing pivot to performance has even made it to already hard-charging startups.
Startups have a time-honored tradition of an always-on, work-first lifestyle. Early employees are expected to put in grueling hours of coding and customer support during this critical phase, known as the "valley of death," when startups are flush with initial funding but not yet profitable.
The free-money era tested this tradition of hustle and thriftiness. Investors heaped money into small startups when interest rates bottomed out, and the blitz scaling that followed set off an arms race of perks to help startups attract top talent. Employees could work from home and set their own schedules. They pocketed wellness stipends and trotted the globe on extravagant off-sites. The tech startup Bolt gave many employees Fridays off.
"I think many individuals — founders included — lost sight of the true goal of a company. It is to make money," Mang-Git Ng, the founder of Anvil, a paperwork automation company, told BI.
Now, the executives who had lavished high salaries and fancy perks on their employees are resetting expectations, winding down remote work, and cutting head count.
"Everyone who comes into our office at Decagon has opted into working with a team that's here because we want to do big things and see bigger and better results," said Jesse Zhang, the founder of Decagon, who now badges into the office six days a week. "There's no such thing as a rocketship that doesn't have a certain level of intensity to fuel its trajectory."
Call it the Big Tech trickle-down effect.
"Founders aren't sugarcoating it," said Natan Fisher, who runs a recruiting firm, SingleSprout, that specializes in hiring technical talent. "I've had a few cofounders tell employees they aren't working hard enough, and, 'If you're not all in, no hard feelings, we can give severance, but we can't slow down.' Late nights, weekends, even people crashing at the office, it's real."
Google is buying the cybersecurity firm Wiz for $32 billion — its biggest acquisition ever.
It's also the biggest deal of the year so far and a major test for Donald Trump's antitrust policy.
A deal of this size could also supercharge more M&A following a slow few years.
Just as 2025 was off to a slow start for M&A, Google announced its largest-ever acquisition.
Google's parent company, Alphabet, and Wiz confirmed Tuesday that they had reached an agreement on the $32 billion all-cash deal.
The two companies were in talks last year for a $23 billion deal, but they fizzled out, and Wiz later said it would instead pursue an initial public offering.
Google CEO Sundar Pichai said during a Tuesday briefing that the Wiz acquisition would boost its cloud security offering at a time when AI is bringing "new risks" and "multi-cloud and hybrid are becoming the norm."
"Against this backdrop, organizations are looking for cybersecurity solutions that improve cloud security and span multiple clouds," he added.
Wiz, an Israeli-founded startup headquartered in New York, specializes in technology that scans everything companies put onto the cloud and identifies security risks. Under the agreement, Wiz's services would continue to be available through other cloud providers such as Amazon Web Services and Microsoft Azure, Thomas Kurian, Google Cloud's CEO, said in the Tuesday briefing.
Anat Ashkenazi, Google's CFO, said the company expects the deal to close in 2026, subject to regulatory approvals.
For Google, the deal would give it the chance to show customers its cloud offerings — which trail behind those of Microsoft and Amazon — are as secure as can be. "With Wiz, we believe we will vastly improve how security is designed, operated, and automated," Kurian added.
Here's why the acquisition is a big deal for startup exits and a big test for the Trump administration.
Return of startup dealmaking?
A deal of this size could catalyze more startup M&A activity following a sluggish few years.
Last year, there were 2,066 VC-backed startup M&A exits, worth $83.6 billion, according to PitchBook data. The first quarter of 2025 was off to a slower start, with 382 M&A deals worth $13.6 billion.
As the biggest deal of the year so far, Google's $32 billion purchase of Wiz has tipped the scales significantly, potentially pointing to more appetite for acquisitions as an exit route.
"Acquisitions allow startups to avoid the inherent volatility of public markets altogether and just focus on long-term growth and their daily operations," Mariam Pettit, the managing partner of Graph Theory Capital, told Business Insider, adding that an acquisition could deliver faster returns than an IPO.
Startups have been gravitating toward secondary share sales and acquisitions in lieu of public listings for multiple reasons. The IPO process can be tough, requiring thorough auditing of a company's financials, potential restructurings, and regulatory compliance — as well as more intense quarterly scrutiny, Pettit said.
The US IPO market remained relatively dormant in 2024 despite high multiples and low volatility. PitchBook data showed that most VC-backed listings underperformed last year, with Instacart, Klaviyo, and Ibotta trading significantly lower relative to the Morningstar Growth Index.
Wiz's deal with Google would offer the company an avenue to bolster its growth, a person familiar with the process told BI, adding that the company surpassed $700 million in annual recurring revenue in the last quarter.
A test for Trump's antitrust policy
The $32 billion acquisition could serve as a test of the regulatory environment under the Trump administration.
While President Donald Trump was expected to usher in a more favorable M&A environment, Vice President JD Vance had previously voiced support for stricter dealmaking rules. Last year, he backed legislation to eliminate tax breaks for corporate mergers, signaling a stance closer to former FTC chair Lina Khan, who, under Biden, launched investigations into Microsoft and Amazon's businesses.
Andrew Ferguson, Khan's replacement, said in February that the Trump administration would continue to use strict corporate merger guidelines adopted under Biden.
Against this backdrop, Google is facing two antitrust lawsuits, including one against its search business that was brought during Trump's first term.
Two recent Google mergers have come under close scrutiny by regulators but ultimately passed: its $2.1 billion acquisition of Fitbit in 2021 and its purchase of the cybersecurity firm Mandiant for $5.4 billion in 2022.
Evelyn Mitchell-Wolf, an EMARKETER senior analyst, described the merger as a "bold move" in the current regulatory landscape.
"Given its price tag — and the fact that Google's already been found guilty of anticompetitive conduct in the search market — the acquisition will attract too much scrutiny to go through completely unimpeded," she said.
Google Cloud restructured its strategy team to "respond faster" to the market.
One of the biggest changes has been made to boost sales productivity, per an internal memo.
Cloud has become a crucial part of Google's business in the AI race.
Google has shaken up its Cloud group yet again, this time consolidating teams in its strategy unit.
Staff in Google Cloud's strategy and operations team were told in February that the reorganization would help the company "respond faster to market demands," according to a memo sent by Google Cloud's Go-To-Market COO Greta Krupetsky and seen by Business Insider. She described the new structure as a "network model" in the memo, which will better support Cloud's various sales teams.
Google has been streamlining many of its internal teams across the company over the past two years in an effort to move faster. Krupetsky said the latest reorg included consolidating its core Business Services into one team led by Google Cloud managing director Abhi Sharma to "help increase sales productivity." A new central "Deal Management" team has also been created under Google Cloud COO Francis deSouza, who joined the company earlier this year.
The unit will include a new "Business Functions - Customer Experience Organization" led by Erez Wohl, managing director at Google Cloud. Wohl's previous strategy and operations functions will now report to Krupetsky.
"We continue to evolve our business to meet our customers' needs and the significant opportunity ahead. We will continue to invest in areas that are critical to our business and ensure our long-term success," a spokesperson told BI in a statement.
The memo mentions a "small number" of roles being cut as part of the reorganization. Those cuts, which BI reported on last month, included some roles being moved to other regions including India and Mexico City, according to an internal employee-crowdsourced document. BI also reviewed several internal job postings for roles within Google Cloud's Go-To-Market Strategy and Operations team based in Mexico City.
Google's Cloud group has become a crucial part of the company's efforts to compete in artificial intelligence. Google Cloud sales hit $11.96 billion in the fourth quarter, up 30% from the previous year, though falling short of analyst expectations. Investors are closely watching Google's Cloud business, which is still smaller than those of rivals Amazon and Microsoft.
Google previously said it plans to spend $75 billion in capital expenditure this year.
Anderson had for years been fascinated by the potential for additive manufacturing — also known as 3D printing — to transform the world. Page, for his part, had pushed Kittyhawk engineers to explore the use of 3D printing to build parts of its flying vehicles, believing it could dramatically cut the cost of production.
Their interests have collided in a new startup they formed called Dynatomics, which explores ways to use AI and additive manufacturing to build aircraft, according to business filings reviewed by Business Insider and multiple people familiar with the startup, who asked to remain anonymous because they were not permitted to speak publicly about the project.
Dynatomics is funded by Page and run by Anderson out of Palo Alto, California, those people said. State corporation filings show the business was registered in California in June 2023, and a website domain name for Dynatomics was secured around the same time. The site is almost entirely empty except for a logo and a teaser: "Working on something new."
Two former Kittyhawk employees who work at Dynatomics are Stanford graduates who studied aerial robotics. On his personal website, Anderson describes working on a new stealth project "at the intersection of AI and advanced manufacturing." For a while, he had the name Dynatomics listed on his LinkedIn but later removed it.
Anderson, as well as representatives for Page, did not respond to requests for comment from BI.
Additive manufacturing involves building an object one layer of material at a time. The aviation industry already uses this process to reduce the cost and increase the speed of some components, and it becomes more difficult when producing larger parts or combining different materials. According to people familiar with the project, Page and Anderson are exploring how to use AI to improve the manufacturing process. The Information had previously reported on Page forming Dynatomics.
At Kittyhawk, Page asked Anderson and then-CEO Sebastian Thrun to have employees design flying vehicles using 3D-printed parts. Page believed it might be possible to reduce the cost of building flying cars by several magnitudes by using new forms of manufacturing, according to people who worked with him at the time.
In 2023, Dynatomics turned to a research group at the Tallinn University of Technology in Estonia, which had just won a national research prize for its work in additive manufacturing of machines. Dynatomics paid the university 60,000 EUR for consultation services "related to additively manufactured electrical machines," according to filings in an online database that posts grant information and a person familiar with the partnership. Dynatomics' name was scrubbed from the grant after BI reached out to the university to inquire about it.
A 2023 research paper by one Tallinn University researcher who consulted with Dynatomics described how using AI with additive manufacturing can "mitigate print failure, reducing waste, cost, and production time."
Page pushes for 3D printing
Page was compelled by the idea that 3D printing and cheap off-the-shelf components could offer a new path forward for electric vertical take-off and landing (eVTOL) aircraft. When he mandated Kittyhawk employees explore a series of projects using 3D printing technology in 2022, Anderson butted heads with some employees who felt Page was asking for things that defy the laws of physics.
"He's a lot more knowledgeable about aviation than I had any reason to believe before coming here," Anderson told staff during a 2022 meeting with Kittyhawk, a recording of which was obtained by BI. "This is his thing." To illustrate Page's obsession with aviation, Anderson told staff that Page's son had given his dad a T-shirt for Christmas that read, "I'm plane crazy."
The frustration sometimes went both ways. Anderson made notes in an online calendar that was accessible to other staff. In one case, he wrote how he wished he could fire everyone at the company and start over, according to three former Kittyhawk employees who saw the note.
Dynatomics may be his opportunity. When Anderson began forming his Dynatomics project, he plucked at least two people who had worked in a highly secretive skunkworks group within Kittyhawk known as "Feather," which busied away on more ambitious and experimental ideas for flying vehicles — and had a rare direct line to Page.
Page and Anderson's close ties
Page's bet on Dynatomics is a bet on Anderson.
An English-American writer and entrepreneur, Anderson's career has taken him from journalist to tech founder. He spent his early years studying physics and playing bass in a post-punk band named REM (culminating in a battle-of-the-bands showdown with the Michael Stipe-fronted namesake for the title, which Anderson's band lost).
He later worked as a journalist and editor at the Economist and Wired. In 2009, he co-founded the consumer drone company 3D Robotics and later sold it to Kittyhawk, the flying taxi startup run by driverless car luminary and ex-Googler Thrun and funded by Page. Anderson later became Kittyhawk's chief technology officer.
Larry Page and Chris Anderson (right) at the Wired NextFest in 2007
Amy Tierney/WireImage
While Anderson and Page knew each other long before Kittyhawk, former employees of the flying car company said the two men became closer after Anderson joined. Anderson would often journey over to Page's nearby home to discuss ideas, sometimes in Page's home garage, which was filled with 3D printers that he would tinker with, according to one of the people who heard it from Anderson.
"He's a good engineer, which is what Larry likes," said a person familiar with both men.
Page's decision to run Kittyhawk outside Alphabet — which houses a collection of moonshot projects — allowed him to fund the project wholesale without answering to stockholders, who had often scrutinized Alphabet's spend on blue-sky ideas like driverless cars and internet balloons.
At Kittyhawk, Page eventually wanted to design a flying car that could be manufactured for $50,000, according to former employees, who said the company never got near that figure. Dynatomics could allow Anderson and Page to make good on what was once Kittyhawk's promise to "Democratize electric aviation."
"One of the reasons the sky is empty is because it's quite expensive to get up there, and if airplanes were as cheap as cars and were made in the same volume of cars, it would change the nature of transportation and perhaps our world," Anderson previously said on the Vertical Space podcast.
He later remarked: "The empty sky is market failure"
Broadcom CEO Hock E. Tan speaks at an event held by President Donald Trump.
AP
Broadcom has cut VMware's workforce by roughly half since acquiring the company in 2023.
VMware had more than 38,000 employees in early 2023. That's down to about 16,000, two sources said.
Wall Street loves CEO Hock Tan's M&A playbook.
Broadcom's aggressive cost-cutting strategy has at least halved VMware's workforce while pleasing Wall Street analysts.
The chip giant closed one of the biggest tech deals ever when it acquired cloud software provider VMware in late 2023. Since closing the acquisition, Broadcom CEO Hock Tan has focused on making the subsidiary more profitable.
Broadcom has been cutting jobs in the past year, including cuts to VMware's salesforce in October and to its professional services team last week, according to current and former employees and LinkedIn posts. In addition, Broadcom has reduced the workforce in other business units or offices over the past few months.
VMware stated that it had more than 38,000 employees as of February 2023, according to a regulatory filing, though many employees and executives departed before the deal closed. Attrition has also contributed to the shrinking workforce.
By this January, VMware had around 16,000 employees, according to two sources. Employees in marketing, partnerships, and VMware Cloud Foundation have been let go over the past few months, according to former staff and LinkedIn posts.
Broadcom has implemented other workplace changes at VMware, such as requiring employees to return to the office. In cases where not enough employees badged into certain offices, Broadcom has shut down those locations and cut staff there, one former employee said. Broadcom did not respond to a request for comment last week.
Tan has concentrated on VMware's largest customers and switched to a subscription-based business model from a perpetual license approach. He said on Thursday's earnings call that 60% of customers had so far been converted.
The company also raised prices. VMware customers have said they've faced massive hikes due to product bundling, which means multiple products are packaged together and customers have to pay for them all.
"The VMware kicker"
Broadcom's stock has climbed about 40% in the past year. In December, Broadcom's market cap hit $1 trillion.
Analysts have largely been pleased by Broadcom's handling of the merger integration.
"The VMware kicker continues to execute, which shouldn't be a surprise, given Hock Tan's expertise on his wash, rinse, and repeat M&A playbook," said Dave Wagner, portfolio manager at Aptus Capital Advisors, ahead of Thursday's blockbuster earnings release.
William Blair analysts called VMware the "star in software," saying it was an "opportunity for Broadcom to drive sustained software growth and potentially reduce the impact of heightened customer churn going into 2027."
Charles O'Rear/Getty, thawornnurak/Getty, valiantsin suprunovich/Getty, twomeows/Getty, aluxum/Getty, Tyler Le/BI
Tech giants Amazon, Google, IBM, and Microsoft are racing to develop a functional quantum computer.
Each has released a prototype quantum chip with different approaches and potential applications.
The field is rapidly evolving, but major hurdles remain before it becomes commercially useful.
The quantum race is heating up.
Tech titans Amazon, Google, IBM, and Microsoft each recently announced advancements in their prototype chips, tightening the race to develop a commercially useful quantum computer that could solve some of the universe's stickiest problems faster than a classical computer ever could.
Quantum computing is a rapidly evolving — though still largely theoretical and deeply technical — field. Butcracking it open could help discover new drugs, develop new chemical compounds, or break encryption methods, among other outcomes, researchers say.
Naturally, each of the major players in Big Tech wants to be the one to take quantum computing mainstream.
"You're hearing a lot about it because this is a real tipping point," Oskar Painter, the director of quantum hardware at Amazon Web Services, told Business Insider in late February, following the company's announcement of its Ocelot chip.
Stick with us — here's where it gets complicated.
Where classical computing uses binary digits — 0s and 1s, called bits — to represent information, quantum computing relies on a foundation built from the quantum equivalent of bits, called qubits. When they behave predictably at a large enough scale, qubits allow quantum computers to quickly calculate equations with multiple solutions and perform advanced computations that would be impossible for classical computers.
However, qubits are unstable, and their behavior is unpredictable. They require specific conditions, such as low light and extremely cold environments, to reduce errors. When the number of qubits is increased, the error rate goes up — making advancement in the field slowgoing.
Small-scale quantum computers already exist, but the race is on to scale them up and make them useful to a wider audience rather than just scientists.
Recently, Amazon, Google, and Microsoft have announced new prototype chips, and IBM has made strides in its existing quantum road map. Each company is using unique approaches to solve the error reduction and scalability problems that have long plagued the field and make useful quantum computing a reality.
Here's how each approach stacks up.
Microsoft
Microsoft's Majorana 1 chip is the first quantum computing chip powered by topological qubits.
Microsoft
Approach to quantum: Topological qubits
Most powerful machine: Majorana 1
In February, Microsoft unveiled its new quantum chip, Majorana 1. The aim is for the chip to speed up the development of large-scale quantum computers from decades to years.
Microsoft said the chip uses a new state of matter to produce "topological" qubits that are less prone to errors and more stable. Essentially, this is a qubit based on a topological state of matter, which isn't a liquid, gas, or solid. As a result, these quantum particles could retain a "memory" of their position over time and move around each other. Information, therefore, could be stored across the whole qubit, so if any parts fail, the topological qubit could still hold key pieces of information and become more fault-resistant.
"Microsoft's progress is the hardest to get an idea about because it's very niche," said Tom Darras, founder of quantum computing startup Welinq. "Even experts in the industry find it difficult to assess the quality of these results."
Quantum experts agree that Microsoft still has many roadblocks to overcome, and its peer-reviewed Nature paper only demonstrates aspects of what its researchers have claimed to achieve — but some in the quantum ecosystem see it as a promising outcome.
Google
Google researchers are aiming to reverse a long-standing qubit problem.
Google
Approach to quantum: Superconducting qubits
Most powerful machine: Willow
In December, Google announced Willow, its newest quantum chip, which the company claims takes just five minutes to solve a problem that would take the world's fastest supercomputer 10 septillion years.
Perhaps more impressive was Google's breakthrough in how quantum computers scale. Historically, the more qubits that are added, and the more powerful the computer becomes, the more prone it is to errors. With Willow, Google's researchers said that adding more physical qubits to a quantum processor actually made it less error-prone, reversing the typical phenomenon.
Known as "below threshold," the accomplishment marks a significant milestone by cracking a problem that has been around since the 1990s. In a study published in Nature, Google's researchers posit this breakthrough could finally offer a way to build a useful large-scale quantum computer. However, much of this is still theoretical, and now Google will need to prove it in practice.
Amazon
A superconducting-qubit quantum chip being wire-bonded to a circuit board at the AWS Center for Quantum Computing in Pasadena, Calif.
Amazon Web Services
Approach to quantum: Superconducting qubits
Most powerful machine: Ocelot
In late February, Amazon Web Services announced its Ocelot chip, a prototype designed to advance the company's focus on cloud-based quantum computing.
An Amazon spokesperson told Business Insider the Ocelot prototype demonstrated the potential to increase efficiency in quantum error correction by up to 90% compared to conventional approaches. The chip leverages a unique architecture that integrates cat qubit technology — named for the famous Schrödinger's cat thought experiment — and additional quantum error correction components that can be manufactured using processes borrowed from the electronics industry.
Troy Nelson, a computer scientist and the chief technology officer at Lastwall, a cybersecurity provider of quantum resilient technology, told Business Insider that Amazon's Ocelot chip is another building block that the industry will use to build a functioning quantum computer. However, its error rate needs to be substantially lowered, and its chips would require more qubit density before they're useful.
"There's lots of challenges ahead. What Amazon gained in error correction was a trade-off for the complexity and the sophistication of the control systems and the readouts from the chip," Nelson said. "We're still in prototype days, and we still have multiple years to go, but they've made a great leap forward."
IBM
CES patrons take a look as IBM unveils this quantum computer, Q System One.
Ross D. Franklin/ASSOCIATED PRESS
Approach to quantum: Superconducting qubits
Most powerful machine: Condor
IBM has been a quantum frontrunner for some time, with several different prototype chips and its development of Q System One, the first circuit-based commercial quantum computer, unveiled in January 2019.
IBM's Condor chip is the company's most powerful in terms of its number of qubits. However, since its development, IBM has focused its approach on the quality of its gate operations and making its newer quantum chips modular so multiple smaller, less error-prone chips can be combined to make more powerful quantum computing machines.
Condor, the second-largest quantum processor ever made, was unveiled at the IBM Quantum Summit 2023 on December 4, 2023. At the same time, IBM debuted its Heron chip, a 133-qubit processor with a lower error rate.
Rob Schoelkopf, cofounder and chief scientist of Quantum Circuits, told Business Insider that IBM has prioritized "error mitigation" over traditional error correction approaches. While IBM has so far been successful in what Schoelkopf calls "brute force scaling" with this approach, he said the methodology will need to be modified in the long run for efficiency.
Who leads the race?
Sankar Das Sarma, a theoretical condensed matter physicist at the University of Maryland, told Business Insider that the Amazon Web Services Ocelot chip, Google's Willow, and IBM's Condor use a "more conventional" superconducting approach to quantum development compared to other competitors.
By contrast, Microsoft's approach is based on topological Majorana zero modes, which also have a superconductor, but in "a radically different manner," he said. If the Majorana 1 chip works correctly, Das Sarma added, it is protected topologically with minimal need for error correction, compared to claims from other tech companies that they have improved conventional error correction methods.
Still, each company's approach is "very different," Das Sarma said. "It is premature to comment on who is ahead since the whole subject is basically in the initial development phase."
Big Tech companies should be cautious about "raising expectations when promoting results," said Georges-Olivier Reymond, CEO of quantum computing startup Pasqal. "Otherwise, you could create disillusionment."
Reymond's sentiment was echoed by IBM's VP of quantum adoption and business development, Scott Crowder, who told Business Insider he is concerned "over-hype" could lead people to discount quantum technology before its promise can be realized.
"We think we are on the cusp of demonstrating quantum advantage," said Crowder, referring to when a quantum computer outperforms classical machines. "But the industry is still a few years from a fully fault-tolerant quantum computer."
Andrew Caballero-Reynolds/Getty Images; Jenny Chang-Rodriguez/BI
The price of building AI is falling to new lows.
New, cheaper AI development techniques have developers rejoicing — but it's not all upshot.
As costs hit rock bottom, Big Tech foundation model builders must justify expensive offerings.
How much does it cost to start an AI company?
The answer is less and less each day as large language models are being created for smaller and smaller sums.
The cost of AI computing is falling. Plus, a technique called distillation to make decent LLMs at discount prices is spreading. This has sent a spark through parts of the AI ecosystem and a chill through others.
Distillation is an old concept gaining new significance. For most, that's good news. For a select few, it's complicated. And for the future of AI, it's important.
Distillation defined
AI developers and experts say distillation is, at its core, using one model to improve another. A larger "teacher" model is prompted to generate responses and paths of reasoning and a smaller "student" model mimics its behavior.
Chinese firm DeepSeek caused a stir with OpenAI-competitive models reported to have trained for around $5 million. It sent the stock market into a panic, punishing Nvidia with a loss of $600 billion in market capitalization for the potential downshift in chip demand. (Such a decline has yet to materialize.)
A UC Berkeley team of researchers, flying further under the radar, trained two new models for under $1000 in computing costs each, according to research released in January.
In early February, researchers from Stanford University, The University of Washington, and the Allen Institute for AI were able to train a serviceable reasoning model for a fraction of that.
Distillation was an unlock for all of these developments.
It is a tool in developers' toolboxes, alongside fine-tuning, to improve models in the training phase, but at a much lower cost than other methods. Both techniques are used by developers to give models specific expertise or skills.
This could mean taking a generic foundation model like Meta's Llama and using another model to distill it into an expert on US tax law, for example.
It could also look like using DeepSeek's R1 reasoning model to distill Llama to have more reasoning capabilities — meaning when AI takes a longer time to generate an answer in order to question its own logic and lay out the process of reaching an answer step-by-step.
"Perhaps the most interesting part of the R1 paper was being able to turn non-reasoning smaller models into reasoning ones via fine-tuning them with outputs from a reasoning model," wrote the analysts at Semianalysis in January.
In addition to the bargain price tag — at least for AI — DeepSeek released distilled versions of other open-source models using the R1 reasoning model as the teacher. DeepSeek's full-sized models, along with the largest versions of Llama are so large that only certain hardware can run them. Distillation helps with that too.
"That distilled model has a smaller footprint, fewer parameters, less memory," explained Samir Kumar, a general partner at Touring Capital. "You can run it on your phone. You could run it on edge devices," he said.
DeepSeek's breakthrough was that the distilled models didn't get worse as they got smaller as was expected. In fact, they got better.
Distillation isn't new but it has changed
The distillation technique first surfaced in a 2015 paper authored by prominent Google AI chiefs Jeff Dean and Geoffrey Hinton, and current Google DeepMind research VP Oriol Vinyals.
Vinyals recently said the paper was rejected from the prestigious NeurIPS conference because it was not deemed to have much impact on the field. A decade later, distillation is suddenly at the forefront of AI discussion.
What makes distillation so powerful now as opposed to back then, is the number and quality of open-source models to use as teachers.
"I think by releasing a very capable model — the most capable model to date — in the open source with a permissible MIT license, DeepSeek is essentially eroding that competitive moat that all the big model providers have had to date, keeping their biggest models behind closed doors," Kate Soule, director of technical management for IBM's LLM Granite, said on the company's Mixture of Experts podcast in January.
How far distillation can go
Soule said Hugging Face, the internet repository for LLMs, is full of distilled versions of Meta's Llama and Alibaba's Qwen, both open-source traditional models.
Indeed, of the 1.5 million models available on Hugging Face, 30,000 of them contain the word "distill" in the name, which conventionally indicates a distilled model. But none of the distilled models have made the site's leaderboard.
Just like shopping at the dollar store in the physical world, distillation presents some of the lowest cost-to-performance ratios on the market, but the selection is somewhat limited and there are drawbacks.
Making a model particularly good at one type of task through distillation can erode its performance in other areas.
Apple researchers attempted to create a "distillation scaling law" that can predict the performance of a distilled AI model based on factors including the size of the model being built, the size of the "teacher" model, and the amount of computing power used.
They concluded that distillation can work better than traditional supervised learning in some cases, but only when a high-quality "teacher" model is used. The teacher also needs to be. larger than the model being trained, but not beyond a certain threshold. Improvement stops as teacher models grow too big.
Still, the technique can, for instance, close the distance between idea and prototype for founders and generally lower the barrier to entry for building AI.
Finding a shortcut to smarter, smaller models doesn't necessarily negate the need for big, expensive foundation models, said multiple AI experts. But it does call into question the financial prospects of the companies that build those big models.
Are foundation models doomed?
"Just about every AI developer in the world today," is using DeepSeek's R-1 to distill new models, Nvidia CEO Jensen Huang said on CNBC following the company's latest quarterly earnings.
Distillation has brought opportunity, but it is poised to meet opposition due to the threat it poses to massive, expensive, proprietary models like those made by OpenAI and Anthropic.
"I think the foundation models will become more and more commoditized. There's a limit that pre-trained models can achieve, and we're getting closer and closer to that wall," Jasper Zhang, cofounder of cloud platform Hyperbolic said.
Zhang said the answer for the big names of LLMs is to create beloved products, rather than beloved models — perhaps lending credence to Meta's decision to make its Llama models somewhat open.
There are also more aggressive tactics foundational model companies can take, according to a Google Deepmind researcher who asked to remain anonymous to discuss other companies.
Companies with reasoning models could remove or reduce the reasoning steps or "traces" shown to the user so that they can't be used for distillation. OpenAI hides the full reasoning path in its large o1 reasoning model but has since released a smaller version, o3-mini, which does show this information.
"One of the things you're going to see over the next few months is our leading AI companies trying to prevent distillation," David Sacks, President Donald Trump's adviser for cryptocurrency and artificial intelligence policy told Fox News in January.
Still, it may be difficult to put the genie back in the bottle, by tamping down distillation in the Wild West of open-source AI.
"Anyone can go to Hugging Face and find tons of data sets that were generated from GPT models, that are formatted and designed for training and likely taken without the rights to do so. This is like a secret that's not a secret that's been going on forever," Soule said on the same podcast.
Anthropic and OpenAI did not respond to requests for comment.
Google's AI Mode gives a glimpse into the future of search.
Justin Sullivan/Getty Images
Google said it would begin testing a new "AI Mode" built into its search page.
It says AI Mode aims to give users "a wider and more diverse" set of results than AI Overviews.
It brings AI even closer to the core Google search experience.
Google said it planned to test a new "AI Mode" for search that would provide an entire page of artificial-intelligence-generated results in response to user queries.
The new feature, which is set to become available to early testers, would place an AI Mode tab at the top of the Google search page. Google says that when a user clicks it, they'll be taken to a new page that answers the query with "a wider and more diverse" set of results powered by AI.
Google already offers AI Overviews, which attempt to answer some queries with a direct answer at the top of the page. AI Mode takes this a step further by generating an entire page of results powered by a custom version of the Gemini 2.0 model that uses reasoning and multimodal AI.
In one example Google provided, a user searching for information about sleep trackers and finds that AI Mode creates a table comparing options.
Google said it would start inviting Google One AI Premium subscribers in the US to test the new AI Mode via Search Labs, adding that it had already become available to a few "trusted testers." It did not provide a timeline for a wider rollout.
A glimpse into the future of search
Google is infusing AI into its major products while trying not to disrupt its primary cash cow, which is search. ChatGPT and other AI chatbots have been seen as an existential threat to Google's core search business, though data suggests they're not making a dent.
The new AI Mode, while still in development, offers a glimpse into a new approach to how Google's search engine may eventually work.
"With this new mode, people can ask nuanced questions that might have previously taken multiple searches — like exploring a new concept to comparing options and beyond — and get a comprehensive AI-powered response," Google said Wednesday in an accompanying explainer.
Google's new AI Mode is being rolled out to early testers.
Google
AI Mode also feels like a bridge between classic search and its Gemini chatbot, which can be accessed only via its own website or through an app. AI Mode would attempt to answer queries directly where it can but also prominently show links to the sources of information and, Google said, tap into shopping data for millions of products.
Like the Gemini chatbot, AI Mode could let users ask follow-up questions. It's also multimodal, letting users ask queries using text, voice, or images.
Google said AI Mode would work only when it has high confidence in the results. In cases where it doesn't, it would simply spit out a list of search results. Some features, such as the comparison tables, would also not be available in AI Mode from the start, but a spokesperson said the company planned to roll them out over time.
Google also said on Wednesday that it had launched its Gemini 2.0 model for AI Overviews, which it said would improve results for more complicated searches, including coding, advanced math, and multimodal queries.
Have a tip? Contact this reporter via email at [email protected] or Signal at 628-228-1836.
New data from Google suggests Search queries are growing at a healthy clip.
The new data bucks the narrative that OpenAI's ChatGPT is eating at Google's business.
The arrival of AI agents could shake things up.
OpenAI's ChatGPT may pose an existential threat to Google's business, though new numbers suggest it isn't making a dent in Google's search growth just yet.
Google said Monday that it now sees more than 5 trillion searches a year. Barclays analysts took that figure — along with previous company disclosures — and did some back-of-the-napkin math to conclude Google queries may have grown over 20% over the two years since ChatGPT's launch in late 2022.
ChatGPT's active users have been rising for two years, reaching 100 million weekly active users (WAUs) in November 2023 and 400 million WAUs in February 2025. That had some Google investors worried.
By estimating that Google had 3.5 trillion searches a year in early 2023 based on Microsoft's disclosure that it had 10 billion searches a day around that time, the analysts found that OpenAI isn't hurting the search giant just yet. Google first said it was handling "trillions" of searches a year in 2016.
"While this could still play out in the future, the opposite appears to be true over the past two years," the Barclays analysts wrote, referring to ChatGPT's perceived threat to Google's business.
It's unclear how this increase is spread across Google's various Search products. The company has built new AI Search tools such as Lens, which lets users point their phone at something and have Google scan it for relevant information; and Circle to Search, a newer tool that lets users search for information on their phone without leaving their current app.
Then there's AI Overviews, the biggest disruptor to Google's classic page of 10 blue links. These provide direct answers to queries right at the top of the search page. Their prevalence has ebbed and flowed, in part thanks to some unfortunate snafus (glue pizza, anyone?) that led to Google scaling back AI Overviews last year.
Jim Yu, CEO of the SEO firm BrightEdge, told BI last week that he has seen more user-generated content appear in Search AI Overviews in recent weeks. This suggests that Google is becoming more confident again in pulling from sources such as Reddit.
