• Microsoft hired ex-Facebook global head of engineering Jay Parikh to lead a new AI unit called CoreAI.
• Internal memos Parikh has sent to employees reveal the unit's early ambitions and accomplishments.
• Parikh's initiatives focus on cultural shifts, operational improvements, and customer focus.

Microsoft envisions an "age of AI agents," and CEO Satya Nadella recently tapped one of Mark Zuckerberg's former top lieutenants to bring it to reality.

In January, Nadella put Jay Parikh in charge of a new AI unit called CoreAI, central to Microsoft's ambition to help developers build digital personal assistants capable of taking over tasks from human workers.

Amid Parikh's first Microsoft Build developer conference in this new role, internal memos reveal his goals for the unit, its early accomplishments, and his advice to address what he sees as problems within the company. Microsoft declined to comment.

A fresh perspective for the 'next phase of Microsoft'

Behind the scenes at Microsoft, Nadella prides himself on hiring outside talent from other big technology companies to add fresh perspective and giving them wide latitude to change how things are done, several people close to the CEO told BI.

Those reports include Charlie Bell, who helped build Amazon's cloud from its earliest days before defecting to Microsoft to become its security boss, and AI CEO Mustafa Suleyman, an ex-Google executive who joined the company from AI startup Inflection.

Parikh joined their ranks in October after running cloud security company Laceworks, acquired in 2024. He previously was vice president and global head of engineering for Meta. Zuckerberg has publicly credited Parikh for many technological achievements during his 11-year tenure at the company.

When Nadella announced Parikh's hiring in an email to employees, he wrote that the "next phase of Microsoft" would require "adding exceptional talent" from outside the company.

In January, when Microsoft reorganized to create a new organization under Parikh. The group, called CoreAI, combined teams from Parikh's new direct reports like Eric Boyd, a corporate vice president of AI platform; Jason Taylor, a deputy CTO for AI infrastructure; Julia Liuson, president of the developer division; and Tim Bozarth, a corporate vice president of developer infrastructure.

Nadella said at the time that Parikh would also work closely with the cloud-and-AI chief Scott Guthrie; the experiences-and-devices leader Rajesh Jha; Bell, the security boss; Suleyman, Microsoft's AI CEO; and Kevin Scott, the company's CTO.

A copy of Parikh's latest org chart viewed by Business Insider shows he has nearly 10,000 reports, most of whom (about 7,000) are in the developer division under Liuson.

Parikh's 'agent factory' vision

Four months in, Parikh has started to make his mark on Microsoft with a vision to create an AI "agent factory." In the early days of Microsoft, cofounders Bill Gates and Paul Allen had ambitions to create the world's first "software factory," a company full of programmers who would build everything from applications to operating systems.

Parikh said he met with Gates and discussed his own concept, a production line for AI agents and applications.

"Building our vision demands this type of culture — one where Al is embedded in how we think, design, and deliver," Parikh wrote in an April 14 email to his team. "The Agent Factory reflects this shift — not just in what we build, but in how we build it together. If we want every developer (and everyone) to shape the future, we have to get there first."

The memos reveal some of the developments at CoreAI since Parikh's arrival.

Since January, Foundry — Microsoft's AI platform for developers — has "delivered $337 million of favorable COGS (cost of goods sold) impact year-to-date, with a projected $606 million on an annualized basis," according to one of Parikh's memos.

Microsoft won new customers for its AI programming tool GitHub Copilot, deploying "5,000+ Copilot Business seats" for Fidelity with 5,000 more expected, another memo stated. Copilot Business sells for $19 per user per month, which would make the deal worth as much as $2.28 million annually at full price, though customers often get discounts for large deals. Fidelity declined to comment.

Startup Harvey AI, meanwhile, has agreed to a two-year $150 million commitment to consume Azure cloud services, according to one of Parikh's memos. Harvey AI declined to comment.

Making Microsoft think macro

The memos viewed by BI show how Parikh appears to be taking seriously his mandate to introduce a new perspective to the company and fix procedural problems that Microsoft may not be able to see that it has.

In a May 10 email to his team, Parikh said shifting the company's culture is "essential" to its future, and outlined progress toward priorities like accelerating the pace at which employees work, breaking down siloes to work better as one team, and making products more reliable and secure.

"One of my early observations coming into Microsoft is that we sometimes treat symptoms rather than systems," Parikh wrote in a May 5 email. "We often focus too much on the micro, which results in band-aids and bolt-ons vs taking a broader system view (which may mean thinking beyond what one team directly owns). This often leads to more complexity and operational burden. We'll help each other get better at this."

Parikh's plan to get Microsoft to focus on the macro is to create a "learning loop" with a debrief after every product launch, incident, customer meeting, internal meeting, or decision. He's started new processes to make this happen, according to the memos.

Parikh has an "Ops Review" series, going team by team to make specific improvements but also to "find common patterns of engineering pain that need broader improvements," he wrote. The reviews, he explained, focus on longer-term operational metrics to help with strategy. "We are zooming out and taking a more end-to-end view of a team's operational setup, creating space for an open discussion around what's working and what's not." The reviews began in April with the App Services team.

Also among Parikh's mandates: more customer focus. His organization is required to conduct reviews of major incidents, like outages, that could impact customers, and chart how quickly the teams identified the problem and deployed a fix.

He also started "get well plans" for unhappy customers after he "encountered a couple of fairly unhappy customers" in recent meetings, according to an April 26 email. His solution? Weekly reviews to "understand where we went off track, identify solutions, and execute the recovery plan," tracking progress until the accounts "get well again."

What Parikh thinks Microsoft should change so far

In the May 5 email, Parikh shared "several recurring themes and insights" within Microsoft that he believes the company should seek to change or simplify.

First, he encouraged his organization to engage engineers from outside their direct team because "different perspectives help."

In his view, Microsoft also takes too long and the process is too hard to deprecate, or discourage use of, old versions of software. "Supporting too many versions is unattainable," Parikh wrote. "We are following up with C+Al (the Cloud + AI organization, under Scott Guthrie) to brainstorm how we can modernize and streamline this."

Incident reviews are overloaded with metrics that don't have enough value, Parikh wrote, and Microsoft sends out too many alerts, which creates noise. "It's important to periodically zoom out and audit how your monitoring is running and to simplify if you are overloaded on alerts and metrics. Use Al to help triage complex alerting situations," he urged.

Parikh encouraged his teams to "see the forest for the trees on scalability," and to organize brainstorming sessions when faced with a traffic load they can't support to see what it would take to support five or 10 times as much traffic. "You may be stuck in a local maxima with incremental improvements, and it might be time to brainstorm how you can get a step function more scale," he wrote.

He also recommended employees seek to address classes of problems, not just one-offs. "Quick fixes lead to complexity," Parikh wrote. "Instead of band-aids, we should aim for broader system improvements that solve whole categories of issues and boost long-term efficiency."

"We're building muscle in spotting patterns, not just patching symptoms," Parikh wrote. "And that's a big deal."

Got a tip? Contact the reporter Ashley Stewart via the encrypted messaging app Signal (+1-425-344-8242) or email ([email protected]). Use a nonwork device.

SilverStone's first '80s throwback PC case started life as an April Fools' joke, but the success of the FLP01 was apparently serious enough to merit a follow-up. The company brought another beige case to the Computex trade show this week, the vertically oriented FLP02 (via Tom's Hardware).

If the original horizontally oriented FLP01 case called to mind a 386-era Compaq Deskpro, the FLP02 is a dead ringer for the kind of case you might have gotten for a generic 486 or early Pentium-era PC. That extends to having a Turbo button built into the front—on vintage PCs, this button could actually determine how fast the processor was allowed to run, though here, it's actually a fan speed control instead. A lock on the front also locks the power switch in place to keep it from being flipped off accidentally, something else real vintage PCs actually did.

Despite its retro facade, the FLP02 is capable of fitting in even higher-end modern PC parts than the original FLP01. Front USB-A and USB-C ports are hidden behind a magnetic door on the front of the case, and its faux-5.25-inch floppy drives are just covers for drive bays that you could use for an optical drive or extra front I/O.

Read full article

Comments

Microsoft is updating Windows 11 with a set of new encryption algorithms that can withstand future attacks from quantum computers in a move aimed at jump-starting what’s likely to be the most formidable and important technology transition in modern history.

Computers that are based on the physics of quantum mechanics don’t yet exist outside of sophisticated labs, but it’s well-established science that they eventually will. Instead of processing data in the binary state of zeros and ones, quantum computers run on qubits, which encompass myriad states all at once. This new capability promises to bring about new discoveries of unprecedented scale in a host of fields, including metallurgy, chemistry, drug discovery, and financial modeling.

Averting the cryptopocalypse

One of the most disruptive changes quantum computing will bring is the breaking of some of the most common forms of encryption, specifically, the RSA cryptosystem and those based on elliptic curves. These systems are the workhorses that banks, governments, and online services around the world have relied on for more than four decades to keep their most sensitive data confidential. RSA and elliptic curve encryption keys securing web connections would require millions of years to be cracked using today’s computers. A quantum computer could crack the same keys in a matter of hours or minutes.

Read full article

Comments

• Microsoft is working on a new "Tenant Copilot" offering, according to an internal memo.
• The company is also developing news ways for customers to manage AI agents alongside human staff.
• Microsoft at the time was planning to announce the developments at next week's Build.

Microsoft is working on a new Copilot and could unveil it at the company's Build conference next week, according to an internal memo viewed by Business Insider.

The software giant also has grand "Agent Factory" ambitions, and is developing new ways for corporate customers to manage AI agents alongside human employees, the memo shows.

The Tenant Copilot project is run by the organization behind the Microsoft 365 business. This new Copilot is designed to "rapidly channel an organization's knowledge into a Copilot that can 'talk,' 'think,' and 'work' like the tenant itself," according to an April 14 email sent by Microsoft executive Jay Parikh.

A "tenant" is the term used to describe corporate users of the Microsoft 365 suite of business applications. A Copilot that has access to these tenants would essentially be able to access customer information stored within their Microsoft 365 accounts.

Parikh explained in the email that Microsoft is using different AI techniques to power the Tenant Copilot feature. Supervised fine-tuning helps "to capture a tenant's voice." The tool will also tap into OpenAI's o3 reasoning model "to shape its thought process." Lastly, "agentic" fine-tuning will "empower real-world tasks," he wrote.

Microsoft at the time planned to offer a public preview of Tenant Copilot at Build, according to the memo. The company sometimes changes what it plans to announce at the conference.

Meanwhile, the CoreAI Applied Engineering team is also "working to launch a collaborative go-to-market plan for top-tier customers to drive successful adoption of our Al cloud," Parikh added in the memo.

Microsoft declined to comment.

Parikh's 'Agent Factory' concept

Parikh is the former head of engineering at Facebook. Microsoft CEO Satya Nadella hired Parikh in October and tapped him in January to run a new group called CoreAI Platform and Tools focused on building AI tools. The group combined Microsoft's developer division and AI platform team and is responsible for GitHub Copilot, Microsoft's AI-powered coding assistant.

This year's Build event will be Parikh's first at the helm of this new organization. In the email to the nearly 10,000 employees in the organization, Parikh discussed a new "Agent Factory" concept. That's likely a nod to cofounder Bill Gates, who talked about Microsoft being a "software factory."

"Building our vision demands this type of culture — one where Al is embedded in how we think, design, and deliver," Parikh wrote. "The Agent Factory reflects this shift — not just in what we build, but in how we build it together. If we want every developer (and everyone) to shape the future, we have to get there first."

Parikh has been trying to work across organizations to collaborate on AI agents, through a "new type of cross-product review" combining teams such as security services like Entra and Intune with "high-ambition agent efforts" within LinkedIn, Dynamics, and Microsoft 365.

Meet your new AI agent co-worker

Part of this effort focuses on how to manage AI agents alongside human employees.

Microsoft, for example, has been working on how to handle identity management for AI agents, according to the memo. This technology usually controls security access for human users. Now, the company is trying to spin up a similar system for AI agents.

"Our hypothesis is that all agent identities will reside in Entra," Parikh wrote, although "not every agent will require an identity (some simpler agents in M365 or Studio, for instance, don't need one)."

Microsoft is taking a similar approach to M365 Admin Center, which is used by IT administrators to manage employee access to applications, data, devices, and users. Future versions of this system will accommodate AI agents as "digital teammates" of human workers, according to Parikh's memo.

Microsoft's Copilot Analytics service is also expanding into broader workforce analytics to give corporate customers a view of how work gets done both by humans and AI agents.

And Parikh aims to make Azure AI Foundry, its generative AI development hub, "the single platform for the agentic applications that you build," he wrote. "At Build, we will have the early versions of this, and we'll iterate quickly to tackle a variety of customer use cases."

Have a tip? Contact this reporter via email at [email protected] or Signal at +1-425-344-8242. Use a personal email address and a nonwork device; here's our guide to sharing information securely.

The world of computers is dominated by binary. Silicon transistors are either conducting or they're not, and so we've developed a whole world of math and logical operations around those binary capabilities. And, for the most part, quantum computing has been developing along similar lines, using qubits that, when measured, will be found in one of two states.

In some cases, the use of binary values is a feature of the object being used to hold the qubit. For example, a technology called dual-rail qubits takes its value from which of two linked resonators holds a photon. But there are many other quantum objects that have access to far more than two states—think of something like all the possible energy states an electron could occupy when orbiting an atom. We can use things like this as qubits by only relying on the lowest two energy levels. But there's nothing stopping us from using more than two.

In Wednesday's issue of Nature, researchers describe creating qudits, the generic term for systems that hold quantum information—it's short for quantum digits. Using a system that can be in three or four possible states (qutrits and ququarts, respectively), they demonstrate the first error correction of higher-order quantum memory.

Read full article

Comments

• Microsoft plans to cut about 6,000 jobs, less than 3% of its global workforce.
• The cuts aim to reduce middle managers and increase coders versus noncoders, as BI earlier reported.
• Tech industry trends show a shift toward fewer managers, as seen with Amazon and Google.

Microsoft plans to cut less than 3% of its global workforce, or about 6,000 employees, with notifications beginning Tuesday, the company confirmed.

A person familiar with the cuts said some affected employees would stay on the payroll for 60 days and still be eligible for rewards and bonuses. Microsoft's spokesperson did not comment on or confirm these terms.

As Business Insider reported last month, these cuts are intended to reduce the number of middle managers and increase the ratio of coders versus noncoders on projects. Microsoft organizations want to increase their "span of control," or the number of employees who report to each manager. A spokesperson said these latest cuts were not performance-driven.

Across the tech industry, a culling of middle managers is already underway. Amazon has been trying to increase the ratio of individual contributors to managers. And in December, Google CEO Sundar Pichai told staff that the company cut vice president and manager roles by 10% as part of an efficiency drive.

Microsoft is also trying to decrease the "PM ratio" on some teams, which is the ratio of product managers or program managers to engineers.

Have a tip? Contact this reporter via email at [email protected] or Signal at +1-425-344-8242. Use a personal email address and a nonwork device; here's our guide to sharing information securely.

Israeli quantum computing startup Classiq has landed $110 million in Series C funding, the largest ever for a quantum software company. The latest round brings its total haul to $173 million since launch five years ago. The round was led […]

The post Israeli startup Classiq raises $110M in funding to become the ‘Microsoft of quantum computing’ first appeared on Tech Startups.

• Amazon is developing a new AI coding tool, internally codenamed Kiro.
• Kiro aims to enhance software development with AI agent technology and a multi-modal interface.
• AI coding assistants are rapidly growing, with major tech firms investing heavily in the space.

AI coding assistants are exploding in popularity. Amazon wants a piece of it.

According to an internal document obtained by Business Insider, Amazon Web Services is building a new AI coding tool, codenamed Kiro.

The software development application taps into AI agents to analyze user prompts and existing data, generating code in "near real-time," the document said.

Kiro is a web and desktop app that can be customized to work with first-party and third-party AI agents, Amazon explained in the document. It also taps into knowledge bases, extensions, and themes further enhancing developer productivity.

Kiro will feature a multi-modal interface, allowing developers to input not just text but also visual diagrams and other contextual information, the document stated.

AWS offers an existing AI coding assistant called Amazon Q. The document obtained by BI suggests that Kiro may be a broader application that taps into multiple AI agents to automate or speed up many aspects of software development.

The tool is expected to be able to auto-generate technical design documents, flag potential issues, and offer code optimizations, Amazon explained in its internal document.

"There is an opportunity to reimagine how AI is used to build software at an exponentially faster rate of innovation and higher product quality," the company wrote.

Amazon disses other AI coding tools

The internal document critiques existing AI coding tools as being locked into "code-centric" interfaces that slow developers down. Kiro aims to "democratize" software creation, minimizing time-to-code and maximizing productivity, it said.

AWS had considered launching Kiro in late June, according to the document, though it remains uncertain whether that timeline is still in effect.

AWS's spokesperson declined to comment on Kiro specifically, but told BI that the company is working on AI agent features for its existing products, like the Q developer tool.

"AI agents are quickly transforming the developer experience, and we are rapidly creating innovative new approaches to software development that take full advantage of these powerful agentic capabilities," the spokesperson said. "We're only getting started."

'Explosion of coding agents'

AI coding assistants have seen a sharp surge in growth lately.

CEO Andy Jassy called out the growth of AI tools such as Cursor and Vercel during Amazon's earnings call last week, highlighting an "explosion of coding agents" among AWS customers.

Google and Microsoft both said that around 30% of their code is now written by AI. David Sacks, the White House's AI and crypto czar, recently called coding assistants the "first big break-out application of AI," noting explosive growth in tools like Cursor and Replit.

"The ramifications of moving from a world of code scarcity to code abundance are profound," Sacks wrote on X last week.

Startups in the space are attracting significant attention. Anysphere, which built Cursor, raised a huge funding round recently, and OpenAI agreed to buy AI coding startup Windsurf for $3 billion.

AI may change the role of human coders

By 2028, 9 out of 10 enterprise software engineers will use AI coding assistants, up from less than 14% in early 2024, according to Gartner estimates. It's unclear how this will reshape the role of human coders.

Last year, AWS CEO Matt Garman said it's possible that most software developers will not be coding in the future because of new AI tools, and that Amazon has to help employees "upskill and learn about new technologies" to boost their productivity, BI previously reported.

Amazon faced early hurdles with its Q coding assistant, sparking internal concerns over high costs and lackluster performance compared to rivals like Microsoft's Copilot, BI previously reported. The company's spokesperson said user experience with generative AI is "constantly evolving" and pointed to customers, including Deloitte and ADP, that saw productivity gains with Amazon Q.

Amazon believes tools like Kiro will simplify common tasks, such as integrating Stripe payments, while empowering developers to do more with less, according to the document.

"With Kiro, developers read less but comprehend more, code less but build more, and review less but release more," it said.

Have a tip? Contact this reporter via email at [email protected] or Signal, Telegram, or WhatsApp at 650-942-3061. Use a personal email address and a nonwork device; here's our guide to sharing information securely.

• Microsoft has a new, two-year rehire ban on employees who are ousted due to performance issues.
• It's also embracing an infamous Amazon approach that treats some employee attrition as "good."
• The tech industry in general is moving toward stricter performance management and reduced perks.

Microsoft is embracing two controversial management approaches that suggests the software giant is getting tougher on employees.

When the company ousts employees due to performance issues, it will now put them on a two-year block list that bars them from being rehired, according to an internal document viewed by Business Insider.

In addition, Microsoft will count these job cuts as "good attrition," according to the document. This means the company is shedding staff that it's happy to see leave.

These two tools are new for Microsoft and they're part of a broader effort by the company to change its performance-management process to shed low performers faster and keep them out.

There are no goals for this "good attrition" metric at present — or at least BI has not uncovered any yet. However, this is already being reviewed at the executive level and appears to be becoming more of a focus as the company dials up performance expectations, two Microsoft managers told BI. They asked not to be identified discussing internal matters. A Microsoft spokesperson declined to comment.

Good attrition is similar to an infamous "unregretted attrition" metric at Amazon, which gives a goal to organizations for the percentage of employees they want to lose every year.

Microsoft's new two-year block list is also similar to an approach used by other tech companies. Meta, for example, maintains internal lists barring some former employees from rejoining the company through systems that track rehire ineligibility, including a "non-regrettable attrition" designation and a "do not rehire" flag, as BI previously reported.

Overall, the industry overall is shifting toward more rigorous performance expectations and less coddling. Performance-based cuts are becoming more of a regular occurrence as tech companies get tougher on employees.

Earlier this year, Microsoft fired 2,000 employees deemed underperformers without severance and started a new performance improvement plan. A recent internal email sent to Microsoft managers, viewed by BI, disclosed that this new plan is "globally consistent" with "clear expectations and a timeline for improvement."

The new process gives employees an option to enter the PIP or quit and accept a "Global Voluntary Separation Agreement (GVSA)," according to another email that BI viewed. Another document, also viewed by BI, shows the agreement includes a payout equal to 16 weeks pay.

Have a tip? Contact this reporter via email at [email protected] or Signal at +1-425-344-8242. Use a personal email address and a nonwork device; here's our guide to sharing information securely.

• Amazon Web Services recently issued guidance to staff on tariffs and data sovereignty concerns.
• AWS aims to reassure customers amid uncertainty over tariffs and other geopolitical issues.
• Amazon previously hinted that third-party sellers might raise prices on its site due to tariffs.

As tariffs spark growing uncertainty across Amazon's retail operations, the company's cloud division is quietly moving to head off similar concerns from business customers.

According to an internal document obtained by Business Insider, Amazon Web Services has issued new guidance to frontline sales and technical staff, instructing them on how to respond to customer questions about tariffs, data sovereignty, and potential restrictions tied to US government policy.

Among the talking points: If an AWS customer asks about possible price increases due to tariffs, employees are told to avoid direct answers and instead reaffirm pricing terms for those covered under existing Private Pricing Agreements (PPAs).

"In the event that AWS does increase prices, these increases will not change any agreed upon discounts, credits or service-specific rates in your PPA," the internal document stated.

While AWS may be less directly impacted by tariffs than Amazon's e-commerce business, the document reveals the company is concerned enough to prep staff with answers to potential tough customer questions.

The document covers questions ranging from potential price hikes and data-privacy concerns. It even broaches the possibility that US President Donald Trump might ban foreign companies from using AWS.

In a recent CNBC interview, Amazon CEO Andy Jassy acknowledged the situation remains fluid and emphasized efforts by the company's e-commerce business to keep consumer prices low. Still, he hinted that some third-party sellers might raise prices in response to tariffs. He also noted that, despite the uncertainty, Amazon continues its data center expansion.

Amazon, whose stock has dropped about 15% this year, is set to report first-quarter earnings on Thursday.

A spokesperson for the company referred BI to a statement from the internal document:

"We're closely monitoring the situation, and we are working to assess the impact on our business. As we navigate the evolving trade policy landscape, our focus remains on delivering value to our customers and innovating on their behalf."

Do not 'speculate'

Tariff-driven price hikes have already become a flashpoint in Amazon's retail division.

As BI previously reported, internal teams have struggled with forecasting, and vendors say Amazon has offered cost relief in exchange for strict margin guarantees. Meanwhile, third-party sellers say they're being forced to raise prices due to rising import costs.

What this means for AWS pricing remains unclear. Internal guidance tells employees not to "speculate," citing the rapidly evolving nature of trade policy.

Some cloud industry experts suggest tariffs could squeeze AWS more than the company lets on.

AWS relies heavily on high-end computing gear, much of it manufactured in China or Taiwan. While some semiconductor components were recently exempted from tariffs, other critical data center parts may still be affected. Trump has paused most new tariffs for 90 days, but a 145% tariff on Chinese goods remains in effect.

"AWS and other hyperscalers could choose to absorb the cost or pass it on to customers," said Travis Rehl, CTO of cloud consultancy Innovative Solutions. "I'm unsure which direction they'd take."

Ben Schaechter, CEO of cloud cost optimization firm Vantage, said tariffs could force AWS to tighten future discounts or slow infrastructure growth due to higher hardware costs.

The bigger threat, some say, is reduced cloud spending.

Randall Hunt, CTO of cloud advisory firm Caylent, told BI that customers are already cutting back in broad spending in anticipation of slower growth and rising costs.

Data sovereignty and Trump-era fears

The growing uncertainty over Trump's actions has pushed Amazon to prepare for even more extreme scenarios, including potential US government demands for cloud customer data or a move to block non-US users from accessing AWS.

Those concerns over privacy and data access have grown recently as Trump's tariff-driven trade war increased tensions between the US and European countries.

If asked about potential US government data requests, Amazon instructed employees to emphasize that AWS does not disclose customer information unless legally required and that all requests are thoroughly reviewed.

The guidance also clarifies AWS's position on the CLOUD Act, or Clarifying Lawful Overseas Use of Data Act. The CLOUD Act, passed in 2018, gives US law enforcement agencies the authority to access data held by US-based companies, even if stored abroad.

AWS has not provided enterprise or government customer data stored outside the US since at least 2020 and it will challenge any "over-broad" or unlawful requests, the document stated.

"The CLOUD Act does not provide the U.S. government with unfettered access to data held by cloud providers," the document added.

On the question of whether Trump could block foreign access to AWS, the document stops short of addressing whether the president has the authority, but notes there's no indication such action is imminent.

In fact, it argues that doing so would contradict the administration's stated goal of supporting US tech companies abroad.

"AWS is closely plugged into US policy and this Administration's efforts, and can confirm we have heard nothing about restricting cloud services to non-US customers in response to addressing trade imbalances or unfair trade barriers, and expect their focus to continue to be on tariffs as the 'rebalancing' mechanism," the document said.

Sanctions and 'Buy Canada'

AWS also addresses fears that US sanctions could restrict access to its services in certain countries. The guidance notes that full country-wide sanctions are rare and that in the past, companies have been given time to wind down operations when sanctions do occur.

"US country-wide sanctions or services restrictions are exceedingly rare," the document said. "But in the theoretical case that such sanctions ever came to pass, AWS would do everything practically possible to provide continuity of service."

Finally, AWS is preparing for patriotic backlash in some markets, such as a potential "Buy Canada" movement. Employees are told to clarify that AWS's Canada office is a registered Canadian corporation headquartered in Toronto, and that customers can choose to store their data locally and encrypt it.

Still, the guidance urges caution. Employees should be careful framing AWS as a "Canadian business," given the complexity of the term.

"Whether AWS is a 'Canadian business' will depend on how that is defined in particular circumstances," the document concludes.

Have a tip? Contact this reporter via email at [email protected] or Signal, Telegram, or WhatsApp at 650-942-3061. Use a personal email address and a nonwork device; here's our guide to sharing information securely.

• The AI field faces a significant talent shortage, with too few skilled workers to fuel the industry.
• Quantum computing startups have noticed and are trying to avoid the same problem.
• Some quantum companies are funding certificate programs and university courses to train new talent.

Quantum computing companies are learning from missteps made during the artificial intelligence boom and are investing heavily in training programs to ensure the fledgling industry maintains its momentum.

The goal is to avoid a talent shortage like the one AI companies are trying to navigate. While hiring for AI-related roles has surged in recent years, the pool of potential workers with the educational background and technical know-how to train large language models, write AI algorithms, and engineer new AI applications hasn't kept up.

Erik Garcell, director of quantum enterprise development at quantum computing software company Classiq Technologies, told Business Insider that quantum computing companies have taken notice of the hiring challenges facing the AI industry and are trying to circumvent them before they become a problem.

"At Classiq, we actually have a whole academic program we're kicking off here, working with universities to deploy quantum curriculum, because not every school has one, and those that do, we're trying to advocate for more of the hands-on practicum side of things," Garcell said.

Research published by the international consulting firm Randstad in November found that 75% of companies adopt AI and hire for AI-related roles. Still, just 35% of talent have received AI training in the last year. That labor pool comes with a significant age gap, with just one in five baby boomers being offered AI skill development opportunities, compared to almost half of Gen Z workers, and a 42 percentage point gender gap favoring men.

That gap persists despite a surge in the need for new talent. PwC found in its 2024 Job Barometer that the share of jobs that require specialized AI skills has grown 700% since 2016 — three times the rate of other job types — and having AI skills comes with the potential for a 25% wage premium.

The talent pool in the quantum industry, which is just a fraction of the size of the AI field, is even smaller, but demand is growing. The trade publication The Quantum Insider reported that quantum computing is expected to create an estimated 250,000 new jobs by 2030, jumping to 840,000 by 2035.

Preparing for quantum's 'ChatGPT moment'

Quantum computing is a rapidly evolving field attracting major interest from tech giants like IBM, Microsoft, Nvidia, and Google. Since it combines computer science, math, and quantum mechanics, it remains deeply technical and expensive to advance.

But while the industry is still in its infancy, Garcell said the major players in the quantum field believe its potential benefits — including advancements in medicine, materials science, and cybersecurity — are worth early investment to ensure a talent shortage doesn't derail its progress.

"IBM really did put out quite a lot of really good educational content for quantum computing; they nurtured the industry very early on with a lot of that," Garcell said. "I'm very happy to say that they have a lot of really great learning available, same with companies like Pennylane, just great educational content for free that you can start diving into."

IBM has partnered with quantum startups like Q-CTRL to build its quantum learning program, which includes free online courses outlining fundamental topics like building quantum algorithms and error-correcting codes. Other companies, like Google, Pennylane, and Microsoft, offer similar independent courses with certificates of completion to identify those with in-demand quantum skills that recruiters are searching for.

The Massachusetts Institute of Technology, the University of Chicago, and the University of California, Berkeley, offer some of the most comprehensive quantum computing courses available at academic institutions.

Garcell said he recently taught a three-day introductory course at MIT to introduce the university's student body to quantum technology. The school also partners with Classiq to offer an official certification course online.

"Right now is just a great time for finding content out there for those who just want to get their hands dirty with quantum computing," Garcell said. "In the industry, there's been a lot of talk about when we're going to have the 'ChatGPT moment,' where everyone turns their head and goes, 'Oh, this technology exists!' where they might not have known it existed before. People are going to start turning their heads, and it's going to make a lot of noise — we just want to be ready."

• Microsoft is trying to simplify AI sales, according to slides from an internal presentation.
• The current approach slowed sales, confused customers, and affected cost and quality, insiders say.
• Microsoft plans to slash the number of "solution areas."

Microsoft is trying to simplify its many AI offerings by streamlining how the products are pitched to customers, according to internal slides from a recent presentation.

The software giant has a bunch of different AI tools called Copilot. There's Copilot for its Teams chat app, Copilot for its PowerPoint presentation tool, Copilot for its Outlook email service — just to name a few.

These products are often split into different "solution areas," as Microsoft calls them. Having Copilot tools in many different buckets can slow down sales, confuse customers, and affect cost and quality of the tools, people in the organization told Business Insider. They asked not to be identified discussing private matters.

Microsoft has sales teams focused on each solution area, which will now be consolidated.

Microsoft Chief Commercial Officer Judson Althoff this week unveiled plans for addressing these issues in the company's upcoming fiscal year, which begins in July. BI obtained copies of slides from his presentation.

According to one of the slides, three major changes include:

• Consolidate Microsoft's solution areas.
• Accelerate regional skills at scale.
• Align teams working with small, medium, and corporate customers with those working with outside channel partners who market and sell Microsoft products.

The organization currently has six solutions areas: Modern work, Business Applications, Digital & App Innovation, Data & AI, Azure Infrastructure, and Security.

Beginning in July, these areas will be combined into three: AI Business Solutions, Cloud & AI Platforms, and Security.

AI Business Solutions will include tools such as Copilot for Microsoft 365, Copilot for Teams, Copilot for Outlook, plus a data visualization product called Power BI, according to a person who attended a Thursday all-hands for Althoff's organization. This person asked not to be identified discussing private matters.

"We are evolving the commercial solution areas within our sales organization to better reflect the era of AI and support the growth of our customers and partners," a Microsoft spokesperson said in a statement. "This evolution reflects the shift in how customers and partners are buying and will better serve their needs."

The other changes include expanding training for salespeople and a reorganization to Small, Medium Enterprise & Channel (SME&C) team, which was announced internally earlier this year.

The changes come as Microsoft is trying to figured out how to make money from its significant AI investments. It has mulled changes including new software bundles with Copilot. The company earlier this year said it plans to spend $80 billion on expanding its network of AI data centers.

Have a tip? Contact this reporter via email at [email protected] or Signal at +1-425-344-8242. Use a personal email address and a nonwork device; here's our guide to sharing information securely.

• Ahead of ServiceNow's earnings, Wall Street worried about DOGE hitting government software spending.
• The company won six new US government customers in the first quarter.
• ServiceNow CEO Bill McDermott tells Business Insider the company remains "un-DOGE-ed."

When I got on the phone with ServiceNow CEO Bill McDermott on Wednesday, one of the first things I asked was, "Have you been DOGE-ed?"

The White House DOGE Office has made an ambitious effort to slash federal spending. The US government buys a lot of software, and since this efficiency drive kicked off in January, Wall Street has worried about which tech companies might lose valuable contracts.

ServiceNow was among those in the potential firing line, helping to push its stock down by more than 20% this year. Ahead of ServiceNow's first-quarter results, TD Cowen analysts wrote about "ongoing DOGE concerns." The company gets roughly 10% of its revenue from the US federal government, so "risks are more acute," the analysts wrote in a preview.

On Wednesday afternoon, ServiceNow reported Q1 numbers, and these concerns seem to have been unfounded, at least for now. The company beat Wall Street expectations and raised guidance for subscription revenue. 

More importantly, ServiceNow said its US public sector business grew more than 30% year-over-year, and it added six new government customers in the first quarter. The stock jumped 11% in after-hours trading.

"Un-DOGE-ed," McDermott said.

Avoiding DOGE carnage

I asked him why ServiceNow has managed to avoid the DOGE carnage. 

The CEO said the company helps organizations save money by providing cloud software that automates many humdrum, but important tasks. ServiceNow's software can also make it easier to consolidate multiple different IT tools and services under one roof, another way to save.

With DOGE on the prowl and tariff risks denting confidence, if organizations can use software to cut costs, be more efficient, and reduce duplicative services, it might be less painful to keep paying ServiceNow.  

That last point may be particularly pertinent to government agencies, which often have many older, less efficient, legacy software systems.

"We're working with agencies to replace costly legacy systems," McDermott said. "They realize this is the moment where the software industrial complex has to be collapsed onto ServiceNow. It's grown in cost and complexity over time due to maverick buying. We're here to help reduce and simplify that."

Savings in Raleigh

McDermott cited the city of Raleigh, North Carolina, which uses ServiceNow to auto-populate personnel forms so different teams, such as HR, IT, Facilities, and Payroll, don't have to enter the same information more than once. That saves city employees more than 1,302 hours annually, according to a ServiceNow case study.

ServiceNow software also helped Raleigh replace six legacy service-management solutions and reduce the number of employees in the city's IT call center from eight to two. Those six staffers now work in other areas with the city. Raleigh estimates that it still saved $315,000 a year. 

That seems small, though such savings add up over time and across multiple customers. And when Elon Musk goes around demanding agencies cut spending, every little bit helps. 

"I like to say that everyone wins in this business, and I still believe that. However, the customer wants some losers now and is looking to consolidate software systems and services," McDermott said. "In uncertain times, we help organizations consolidate their legacy technology spending."

• Microsoft's new performance improvement plan includes an offer for employees to quit.
• The new system is similar to Amazon's Pivot program, which has been criticized by some employees.
• The tech industry is shifting to stricter performance expectations.

Microsoft is offering a new choice to employees who are deemed low performers: Accept a payout to quit or risk termination on a performance improvement plan.

An internal email sent to Microsoft managers on Friday disclosed the company's new "globally consistent" performance improvement plan with "clear expectations and a timeline for improvement."

Affected employees can enter the PIP or quit and accept a "Global Voluntary Separation Agreement (GVSA)," the email, viewed by Business Insider, said.

"This performance improvement process is available year-round so you can act quickly to transparently address performance issues, while offering employees choice," said the email, which was from Amy Coleman, Microsoft's new chief people officer.

Another internal Microsoft document viewed by BI said that the payout to employees would be equal to 16 weeks of pay. Affected staff have five days to decide which option to take. If they choose to start the PIP, the payout is off the table, this internal document said.

The new PIP system may work differently outside the US, as other countries have different laws. Microsoft declined to comment.

Shades of Pivot

Microsoft's approach is similar to Amazon's Pivot program, which offers a PIP option or a payout if the employee leaves.

BI has reported extensively on Pivot, a central cog in a broader Amazon performance review system that some employees have said is designed to meet firing quotas rather than revive careers.

Amazon has defended its system in the past. In 2021, the company told BI that it provided managers "with tools to help employees improve their performance and grow in their careers." It added: "This includes resources for employees who are not meeting expectations and may require additional coaching."

More PIP, less coddling

The latest changes come as Microsoft, and the tech industry overall, shifts toward more rigorous performance expectations and less coddling.

Earlier this year, Microsoft fired 2,000 employees deemed underperformers without severance.

Leaders in at least some parts of the company are considering additional performance-based cuts as soon as May. They're looking at cutting middle managers and how to increase the ratio of coders versus noncoders on projects.

Have a tip? Contact this reporter via email at [email protected] or Signal at +1-425-344-8242. Use a personal email address and a nonwork device; here's our guide to sharing information securely.

• Wall Street analysts say Amazon has paused some data center deals.
• The data center market may be slightly cooling, after a frenzied couple of years.
• Microsoft has also taken its foot off the AI accelerator a bit recently.

First, it was Microsoft. Now Amazon is raising eyebrows on Wall Street as fresh signs suggest the cloud giant may be easing off the accelerator in the race to build AI data centers.

Some analysts are concerned that Amazon Web Services, the dominant cloud provider, may be entering a digestion phase that could slow momentum in the data center market.

The speculation gained traction Monday when famed short-seller Jim Chanos posted on X with a simple and ominous remark, alongside an analyst note suggesting caution around AWS's data center plans.

That note, published by Wells Fargo analysts, cited industry sources who reported this weekend that AWS paused discussions for certain new colocation data center deals, particularly international ones. The analysts stressed that the scale of the pause remains unclear, though they're worried.

"It does appear like the hyperscalers are being more discerning with leasing large clusters of power, and tightening up pre-lease windows for capacity that will be delivered before the end of 2026," the analysts wrote.

Oh no, colo

The same day, TD Cowen analysts published similar findings from their own data center research.

"Our most recent checks point to a pullback in US colocation deals from Amazon," they wrote in a note to investors. Colo deals, as they're known in the industry, involve different companies sharing space in the same data center.

"We are aware of select colocation deals that it walked away from, as well as expansion options that it chose not to exercise," the Cowen analysts added.

They also said that their recent industry checks point to a slowdown in Amazon's AI ambitions in Europe.

"This is a dynamic we will continue to monitor," the analysts wrote.

Three signs of moderation

More broadly, Cowen's analysts have spotted a cooling in the data center market — relative to the frenzied activity of recent years.

"We observed a moderation in the exuberance around the outlook for hyperscale demand which characterized the market this time last year," they wrote, laying out three specific signs of calmer times:

• Data center demand has moderated a bit, particularly in Europe.
• There has been a broader moderation in the urgency and speed with which cloud companies seek to secure data center capacity.
• The number of large deals in the market appears to have moderated.

Some context is important here. The AI data center market has gone gangbusters ever since OpenAI's ChatGPT burst onto the scene in late 2022 and showed the potential of generative AI technology.

These signs of moderation are pretty small in relation to this huge trend. However, trillions of dollars in current and planned investments are riding on the generative AI boom. With so much money on the line, any inkling that this rocket ship is not ascending at light speed is unnerving.

Microsoft made similar moves

These signals from Amazon echo similar moves by Microsoft, which recently halted some data center projects.

"Like Microsoft, AWS seems to be digesting recent aggressive leasing activity," the Wells Fargo analysts wrote.

They clarified that this doesn't mean signed deals are being canceled, but rather that AWS is pulling back from early-stage agreements like Letters of Intent or Statements of Qualifications—common ways that cloud providers work with partners to prepare for data center projects.

Amazon says it still sees strong AI demand

In response to these growing concerns, Kevin Miller, vice president of Global Data Centers at AWS, posted on LinkedIn on Monday to offer some clarity.

"We continue to see strong demand for both Generative AI and foundational workloads on AWS," he wrote.

He explained that AWS has thousands of cloud customers around the world and must weigh multiple solutions to get them the right capacity at the right time.

"Some options might end up costing too much, while others might not deliver when we need the capacity," Miller wrote. "Other times, we find that we need more capacity in one location and less in another. This is routine capacity management, and there haven't been any recent fundamental changes in our expansion plans."

Amazon did not respond to a request for comment from Business Insider.

Digestion or indigestion?

Miller's comments aim to position the pause not as a red flag, but as part of the normal ebb and flow of data center growth.

Historically, these digestion periods, marked by slowing new leases or deferred builds, can last 6 to 12 months before a rebound, the Wells Fargo analysts wrote. Google, for instance, pulled back from leasing in the second half of 2024, only to return aggressively in early 2025, they noted.

The Cowen analysts said Amazon's recent cautious moves to pull back on colocation deals may be related to efforts to increase efficiency across its data center operations. Also, AWS typically doesn't do a lot of colocation deals anyway, preferring instead to build its own data centers, the analysts wrote.

They also wrote that other tech giants, such as Meta and Google, are still aggressively pursuing new capacity.

The bottom line? While AWS appears to be taking a breath, the AI cloud race is far from over. Analysts and investors will watch closely to see whether this pause marks a brief recalibration or a more significant shift in AI strategy.

• GPU shortages delayed projects in Amazon's retail division last year.
• The company created a more efficient approval and monitoring process for internal GPU requests.
• Amazon says it now has "ample" GPU capacity across the company.

Last year, Amazon's huge retail business had a big problem: It couldn't get enough AI chips to get crucial work done.

With projects getting delayed, the Western world's largest e-commerce operation launched a radical revamp of internal processes and technology to tackle the issue, according to a trove of Amazon documents obtained by Business Insider.

The initiative offers a rare inside look at how a tech giant balances internal demand for these GPU components with supply from Nvidia and other industry sources.

Early in 2024, the generative AI boom was in full swing, with thousands of companies vying for access to the infrastructure needed to apply this powerful new technology.

Inside Amazon, some employees went months without securing GPUs, leading to delays that disrupted timely project launches across the company's retail division, a sector that spans its e-commerce platform and expansive logistics operations, according to the internal documents.

In July, Amazon launched Project Greenland, a "centralized GPU capacity pool" to better manage and allocate its limited GPU supply. The company also tightened approval protocols for internal GPU use, the documents show.

"GPUs are too valuable to be given out on a first-come, first-served basis," one of the Amazon guidelines stated. "Instead, distribution should be determined based on ROI layered with common sense considerations, and provide for the long-term growth of the Company's free cash flow."

Two years into a global shortage, GPUs remain a scarce commodity —even for some of the largest AI companies. OpenAI CEO Sam Altman, for example, said in February that the ChatGPT-maker was "out of GPUs," following a new model launch. Nvidia, the dominant GPU provider, has said it will be supply-constrained this year.

However, Amazon's efforts to tackle this problem may be paying off. By December, internal forecasts suggested the crunch would ease this year, with chip availability expected to improve, the documents showed.

In an email to BI, an Amazon spokesperson said the company's retail arm, which sources GPUs through Amazon Web Services, now has full access to the AI processors.

"Amazon has ample GPU capacity to continue innovating for our retail business and other customers across the company," the spokesperson said. "AWS recognized early on that generative AI innovations are fueling rapid adoption of cloud computing services for all our customers, including Amazon, and we quickly evaluated our customers' growing GPU needs and took steps to deliver the capacity they need to drive innovation."

"Shovel-ready"

Amazon now demands hard data and return-on-investment proof for every internal GPU request, according to the documents obtained by BI.

Initiatives are "prioritized and ranked" for GPU allocation based on several factors, including the completeness of data provided and the financial benefit per GPU. Projects must be "shovel-ready," or approved for development, and prove they are in a competitive "race to market." They also have to provide a timeline for when actual benefits will be realized.

One internal document from late 2024 stated that Amazon's retail unit planned to distribute GPUs to the "next highest priority initiatives" as more supply became available in the first quarter of 2025.

The broader priority for Amazon's retail business is to ensure its cloud infrastructure spending generates the "highest return on investment through revenue growth or cost-to-serve reduction," one of the documents added.

Amazon's new GPU "tenets"

Amazon's retail team codified its approach into official "tenets" — internal guidelines that individual teams or projects create for faster decision-making. The tenets emphasize strong ROI, selective approvals, and a push for speed and efficiency.

And if a greenlit project underdelivers, its GPUs can be pulled back.

Here are the 8 tenets for GPU allocation, according to one of the Amazon documents:

1. ROl + High Judgment thinking is required for GPU usage prioritization. GPUs are too valuable to be given out on a first-come, first-served basis. Instead, distribution should be determined based on ROl layered with common sense considerations, and provide for the long-term growth of the Company's free cash flow. Distribution can happen in bespoke infrastructure or in hours of a sharing/pooling tool.
2. Continuously learn, assess, and improve: We solicit new ideas based on continuous review and are willing to improve our approach as we learn more.
3. Avoid silo decisions: Avoid making decisions in isolation; instead, centralize the tracking of GPUs and GPU related initiatives in one place.
4. Time is critical: Scalable tooling is a key to moving fast when making distribution decisions which, in turn, allows more time for innovation and learning from our experiences.
5. Efficiency feeds innovation: Efficiency paves the way for innovation by encouraging optimal resource utilization, fostering collaboration and resource sharing.
6. Embrace risk in the pursuit of innovation: Acceptable level of risk tolerance will allow to embrace the idea of 'failing fast' and maintain an environment conducive to Research and Development.
7. Transparency and confidentiality: We encourage transparency around the GPU allocation methodology through education and updates on the wiki's while applying confidentiality around sensitive information on R&D and ROI sharable with only limited stakeholders. We celebrate wins and share lessons learned broadly.
8. GPUs previously given to fleets may be recalled if other initiatives show more value. Having a GPU doesn't mean you'll get to keep it.

Project Greenland

To address the complexity of managing GPU supply and demand, Amazon launched a new project called Greenland last year.

Greenland is described as a "centralized GPU orchestration platform to share GPU capacity across teams and maximize utilization," one of the documents said.

It can track GPU usage per initiative, share idle servers, and implement "clawbacks" to reallocate chips to more urgent projects, the documents explained. The system also offers a simplified networking setup and security updates, while alerting employees and leaders to projects with low GPU usage.

This year, Amazon employees are "mandated" to go through Greenland to obtain GPU capacity for "all future demands," and the company expects this to increase efficiency by "reducing idle capacity and optimizing cluster utilization," it added.

$1 billion investment in AI-related projects

Amazon's retail business is wasting no time putting its GPUs to work. One document listed more than 160 AI-powered initiatives, including the Rufus shopping assistant and Theia product image generator.

Other AI projects in the works include, per the document:

• A vision-assisted package retrieval (VAPR) service that uses computer-vision technology to help drivers quickly identify and pick the correct packages from vans at delivery stops.
• A service that automatically pulls in data from external websites to create consistent product information.
• A new AI model that optimizes driver routing and package handling to reduce delivery times and improve efficiency.
• An improved customer service agent that uses natural language to address customer return inquiries.
• A service that automates seller fraud investigations and verifies document compliance.

Last year, Amazon estimated that AI investments by its retail business indirectly contributed $2.5 billion in operating profits, the documents showed. Those investments also resulted in approximately $670 million in variable cost savings.

It's unclear what the 2025 estimates are for those metrics. But Amazon plans to continue spending heavily on AI.

As of early this year, Amazon's retail arm anticipated about $1 billion in investments for GPU-powered AI projects. Overall, the retail division expects to spend around $5.7 billion on AWS cloud infrastructure in 2025, up from $4.5 billion in 2024, the internal documents show.

Improving capacity

Last year, Amazon's heavy slate of AI projects put pressure on its GPU supply.

Throughout the second half of 2024, Amazon's retail unit suffered a supply shortage of more than 1,000 P5 instances, AWS's cloud server that contains up to 8 Nvidia H100 GPUs, said one of the documents from December. The P5 shortage was expected to slightly improve by early this year, and turn to a surplus later in 2025, according to those December estimates.

Amazon's spokesperson told BI those estimates are now "outdated," and there's currently no GPU shortage.

AWS's in-house AI chip Trainium was also projected to satisfy the retail division's demand by the end of 2025, but "not sooner," one of the documents said.

Amazon's improving capacity aligns with Andy Jassy's remarks from February, when he said the GPU and server constraints would "relax" by the second half of this year.

But even with these efforts, there are signs that Amazon still worries about GPU supply.

A recent job listing from the Greenland team acknowledged that explosive growth in GPU demand has become this generation's defining challenge: "How do we get more GPU capacity?"

Do you work at Amazon? Got a tip? Contact this reporter via email at [email protected] or Signal, Telegram, or WhatsApp at 650-942-3061. Use a personal email address and a nonwork device; here's our guide to sharing information securely.

• Microsoft is adding policies meant to enhance performance management and address low performers.
• The tech industry is shifting toward more rigorous performance expectations and less coddling.
• Microsoft's new policies include a two-year rehire ban for employees found to be underperforming.

Microsoft has created new policies and tools for managers as part of an effort to dial up performance pressure on employees, according to an internal email viewed by Business Insider.

Amy Coleman, Microsoft's new chief people officer, on Friday emailed managers about "new and enhanced tools to help you accelerate high performance and swiftly address low performance."

There's a new option for exiting underperformers, and a policy that bars these people from transferring within Microsoft or getting rehired by the company for two years, according to the email.

Many tech companies have been getting tougher on employees in the past year or so. Efficiency has replaced perks and pampering, as performance-based job cuts become more of a regular occurrence.

For instance, Mark Zuckerberg targeted low performers when Meta eliminated thousands of jobs earlier this year. And, similar to Microsoft's new policy, Meta puts ousted employees on "block lists" meant to stop them from being rehired by the company. 

Earlier this year, Microsoft fired 2,000 employees deemed underperformers without severance. Managers spent months evaluating employees all the way up to the executive level as part of changes to the company's performance review and management process, insiders told BI.

Coleman's email, in full below, details some of those changes and suggests others on the horizon. Microsoft declined to comment.

"Employees with zero to 60% rewards" refers to employees with low scores in Microsoft performance reviews, which use a scale from 0 to 200 that influences how much they receive in stock awards and cash bonuses.

Read the full email to Microsoft managers:

Managers,

Thank you for your patience yesterday with the unexpected technical product issues. As customer zero, this was an opportunity to learn and quickly make improvements.

As Satya shared at the recent Employee Town Hall, our success as a company depends on our relevance in year 51 and beyond - in terms of our innovation, the products we deliver, and the impact we have for our customers and partners. With that, our focus remains on enabling high performance to achieve our priorities spanning security, quality, and leading Al. This focus and our growth mindset encourage excellence, motivates us to push ourselves through challenges, and enables us to deliver results.

Today, we're rolling out new and enhanced tools to help you accelerate high performance and swiftly address low performance. Our goal is to create a globally consistent and transparent experience for employees and managers (subject to local laws and consultation). These tools will also help foster a culture of accountability and growth by enabling you to address performance challenges with clarity and empathy.

Manager Readiness: FY25 Performance, Connect and Rewards: Register for a 60-minute virtual, facilitated session to dive deeper into the performance landscape at Microsoft. Each session will explore what's new for FY25 rewards, provide guidance on differentiating rewards outcomes, and define what it means to deliver "significant impact." Next week, you'll also see a mail from Performance and Development to all employees with more details on Connects.

Clarity and Transparency in Rewards: This year, we'll ensure more transparency and clarity in the Rewards process for managers including additional guidance for each rewards outcome and showing payout percentages to help you make decisions that align with our high-performance expectations. More Rewards details will be shared by the end of the month.

Performance Improvement Process: If an employee is not meeting expectations, you can use the Performance Improvement Plan (PIP), a new globally consistent approach to set clear expectations and a timeline for improvement. The employee can accept the improvement plan or choose to transition out of the company with the offer of a Global Voluntary Separation Agreement (GVSA). This performance improvement process is available year-round so you can act quickly to transparently address performance issues, while offering employees choice.

Updated Internal Movement/External Rehire Policy: Employees with zero and 60% Rewards outcomes and/or on an active PIP will not be eligible for internal transfers. Former employees who left with zero or 60% Rewards or during/after a PIP will not be eligible for rehire until two years after their termination date.

Manager Excellence Initiatives: Additionally, in the coming months, we'll launch several initiatives to strengthen how we measures manage, and motivate teams to deliver for our customers. You'll have access to scenario-based, Al-supported tools designed to help you prepare for constructive or challenging conversations by practicing in an interactive environment.

Thank you for your leadership and commitment to driving high performance and accountability across your team. This isn't just about Microsoft's success. This is about your success, your team's success, our customers' success, and together, fostering a culture where high-performing, winning teams can thrive.

Have a tip? Contact this reporter via email at [email protected] or Signal at +1-425-344-8242. Use a personal email address and a nonwork device; here's our guide to sharing information securely.

• AMD sees the AI inference shift as a chance to grab market share from Nvidia.
• AI inference will move from data centers to edge devices, like smartphones, AMD's CTO says.
• Mark Papermaster expects an AI 'killer app' in the next three to six years.

The lion's share of artificial intelligence workloads moving from training to inference is great news for AMD, its CTO said.

AI training workloads — the ones that make up the gargantuan task of building large language models, imbuing them with a familiar writing or speaking style, and knowledge — used to be most of what AI computing was used for. Inference is the computing process that happens when AI generates outputs like answering questions or creating images.

It's hard to pin down exactly when the switch happened — probably some time last year. But inference is now and will likely stay the largest segment of accelerated computing going forward. Since then, AMD executives have been hyping up a window of opportunity to wrest market share from Nvidia.

"People like the work that we've done in inference," CEO Lisa Su said on the company's February earnings call.

AI at scale is all about inference.

If you ask Mark Papermaster, AMD's Chief Technology Officer, where it all goes from there, he'll tell you that as inference grows, it's headed for the edge.

"Edge devices" are the industry term for computers that live outside the data center. Our phones and laptops all qualify, but so could smart traffic lights or sensors in factories. Papermaster's job is to make sure AMD is headed in the right direction to meet the demand for AI computing across devices as it grows.

AMD has had to play catch-up in the data center since Nvidia's 10-year head start. But at the edge? The field is more open.

Business Insider asked Papermaster what he thinks the future of handheld AI looks like.

This Q&A has been edited for clarity and length.

What's the most prominent use for AI computing in edge devices like laptops and phones?

The use case you're starting to see is local, immediate, low-latency content creation.

Why do we use PCs? We use them to communicate, and we use them to create content. As you and I are talking — this is a Microsoft Teams event — AI is running underneath this. I could have a correction on it such that if I look side to side, you just see me centered. That's an option. I can hit automatic translation — you could be in Saudi Arabia and not speak any English, and we could have simultaneous translation once these things become truly embedded and operational, which is imminent.

It's truly amazing what's coming because just locally on your PC, you'll be able to verbally describe: 'Hey, I'm building a PowerPoint. I need this. I need these features. I'm running Adobe. This is what I want.'

Today, I've got to go back to the cloud. I've got to run the big, heavy compute. It's more expensive and it takes more time.

That's the immediate example that's front and center, and this is why we've invested heavily in AI PCs. That's imminent from Microsoft and others in the next six months.

The other application that we're already seeing is autonomous anything. It starts with cars, but it's way beyond cars. It's the autonomous factory floor.

OK, say it's 2030 — how much inference is done at the edge?

Over time, it'll be a majority. I can't say when the switch over is because it's driven by the applications — the development of the killer apps that can run on edge devices. We're just seeing the tip of the spear now, but I think this moves rapidly.

You might consider phones as an analogy. Those phones were just a nice assist until the App Store came out and made it really easy to create a ton of applications on your phone.

Now, things that used to always be done with more performant computing could be done more locally. Things that were done in the cloud could be done locally. As we start to get killer applications, we're going to start to see that shift go very rapidly. So it's in the next three to six years, no doubt.

I keep running into examples that suggest the way models are getting better is to just keep piling on more inference compute.

How do you know that three years from now, there's not going to be some breakthrough that makes all these devices being designed now completely out of date?

Everything you're describing is to gain even more capability and accuracy. It doesn't mean that what we have is not useful. It's just going to be constantly improving, and the improvement goes into two vectors.

One vector is becoming more accurate. It can do more things, and typically drives more compute. There's an equal vector that runs in parallel, saying, 'How could I be more optimized?'

I call it the DeepSeek moment. It sort of shook the world. Now you have everybody — Microsoft, Meta, Google — making their models more efficient. So you have both examples where it's taking more and more compute and examples where there's innovation driving more and more efficiency. That's not going to change.