• Tech's heaviest hitters are racing to build new AI tech for healthcare.
• They're landing partnerships with health systems and Big Pharma and selling consumer health AI tech.
• Here's how tech powerhouses from Microsoft to Alphabet to Nvidia are investing in healthcare AI.

Artificial intelligence is gaining ground in healthcare, and tech's biggest players are racing to stake their claims.

Many tech giants have been building healthcare businesses long before the AI boom. Amazon, for example, invested heavily in the industry when it acquired primary care provider One Medical for $3.9 billion in 2022.

Now, those tech powerhouses are sharpening their healthcare strategies with AI at the center. Chipmaker Nvidia has rapidly expanded its healthcare ambitions in the past two years, aiming to tackle everything from surgical robotics to drug discovery.

Their healthcare ambitions reflect a broader shift: as AI advances, Big Tech sees fresh opportunities for transformation — and revenue — in the complex industry.

From AI-powered wearables to robotic surgery, here's how tech's most powerful companies are investing in healthcare AI.

Microsoft: Doubling down on its healthcare cloud

Microsoft first entered healthcare nearly two decades ago. Now, it's integrating AI into its cloud solutions to automate hospital operations.

Microsoft bought the ambient intelligence company Nuance in 2022 for nearly $20 billion. Nuance dominates the market for AI-powered medical scribing, though it now faces tough competition from startups like $2.75 billion Abridge. Its latest release is Dragon Copilot, a solution Microsoft announced in March that integrates the company's voice dictation tech with Nuance's ambient listening, in a move the company said would help doctors save even more time documenting patient visits.

In an October 2024 KLAS report, most healthcare organizations said they considered Nuance when purchasing a clinical documentation solution, in part due to prior contracts with Microsoft and the company's extensive healthcare software suite. The tech giant is integrating AI into those other healthcare cloud offerings to organize and analyze medical records and automate tasks like patient scheduling and paperwork.

Microsoft also partners with Nvidia to combine Nvidia's AI tech with Microsoft's cloud solutions to power advanced healthcare and biopharma research and better medical imaging.

Apple: Banking on AI-powered consumer health tracking

Apple's AI efforts have been understated compared to its peers, as have its healthcare investments.

Much of Apple's healthcare attention to date has focused on its Apple Watch, which includes several AI-powered health features for consumers, from AI-powered fall detection and irregular heartbeat tracking to sleep analysis.

After Apple released its augmented reality headset Apple Vision Pro last year, the company said multiple healthcare organizations were using the new tech for various applications, including reviewing patient surgical plans or training clinicians to use new medical devices.

Apple's next big health AI push could be on the horizon. Bloomberg News reported at the end of March that Apple was developing an AI-powered health coach to provide personalized lifestyle recommendations based on its consumers' health data, as tracked through devices like the Apple Watch and iPhone.

TK APPLE COMMENT

Nvidia: Bringing "physical AI" to the hospital

Chip giant Nvidia is digging into a wide range of healthcare specialties, from radiology to drug discovery, primarily by partnering with other healthcare companies.

Nvidia VP of healthcare Kimberly Powell told Business Insider in April that medical imaging was one of Nvidia's entry points into healthcare. The company has notched a number of medical imaging partnerships powered by its AI platforms, most recently with GE Healthcare in March. GE Healthcare plans to use Nvidia's tech to simulate autonomous medical imaging, including autonomous X-rays and ultrasounds, to test their application in physical medical devices.

It's a similar collaboration to Nvidia's existing deal with IT solutions provider Mark III to work with healthcare systems to create simulations of their hospital environments for AI development. "Physical AI," or AI that can analyze and interact with the physical world, is a key part of Nvidia's vision for healthcare. 

"This physical AI thing is coming where your whole hospital is going to turn into an AI," Powell told BI in November. "You're going to have eyes operating on your behalf, robots doing what is otherwise automatable work, and smart digital devices. So we're super excited about that, and we're doing a lot of investments."

Nvidia is also an investor in several healthcare startups, most notably clinical documentation startup Abridge, which raised $250 million in February from investors including Nvidia's venture arm NVentures at a $2.75 billion valuation, and Hippocratic AI, which raised $141 million in Series B funding from NVentures and other firms at a $1.64 billion valuation.

It's also partnered with many of its portfolio companies, such as Moon Surgical, a robotics company that uses Nvidia's medical device AI platform Holoscan for its surgical assistant robot.

Amazon: Adding more tech to consumer care

Amazon is bringing AI to doctors, patients, and pharma companies across its healthcare businesses.

In March, Amazon announced it was testing a new chatbot feature called Health AI, which the company says can give users advice and suggest products for common medical needs. The health AI assistant can direct users to Amazon's online pharmacy or to talk to a doctor at its primary care chain, One Medical, for further care. An Amazon spokesperson told BI that the chatbot is in beta, and the retailer is collecting feedback from customers and working on new features for its AI assistant.

The retail giant also offers its own medical transcription tool, called HealthScribe, which analyzes doctor-patient conversations to create clinical notes for providers. It's one of several AI capabilities that One Medical's clinicians use, in addition to features that assist with patient messaging and care coordination.

Amazon provides a number of generative AI tools through Amazon Web Services for life sciences companies, too. Genentech and AstraZeneca, for example, have used AWS's AI tools for drug discovery research and clinical trial tasks.

Amazon has also suffered more false starts in healthcare than its peers. The retailer shut down its telehealth service Amazon Care in 2022, three years after its launch, and discontinued its wearable fitness tracker Amazon Halo in 2023. Last year, One Medical came under fire after media reports and a malpractice lawsuit raised concerns about the primary care chain's patient safety practices. An Amazon spokesperson said that the company is prohibited by law from discussing patient records, but that One Medical has extensive quality and safe measures in place for patient care.

Alphabet: Powering better health research with foundation models and search tools

Alphabet has created several healthcare AI tools that build upon Google's core search capabilities, with a focus on healthcare-specific foundation models.

Google launched MedLM, a set of healthcare foundation models, in 2023 to enable tech to summarize patient-doctor conversations, conduct clinical research, and automate insurance claims processing. In October, it announced Vertex AI Search for Healthcare, a specialized search engine that gives clinicians answers to their medical questions about a patient's health records or medical documents. It's also built a research AI system designed to assist with patient diagnosis, which can analyze data such as medical images and simulate patient-provider interactions.

On the consumer front, Google has released a number of tools with health applications, such as Google Lens, which allows people to take a picture of their own skin and search for visually similar skin conditions. It's also working on personal health AI models that can interpret sleep and fitness data to offer personalized wellness suggestions.

Isomorphic Labs, Alphabet's AI research arm spun out of Google DeepMind, has partnered with pharma giants like Novartis and Eli Lilly on more efficient drug development, building on DeepMind's AlphaFold protein structures.

Dr. Karen DeSalvo has led Google's health initiatives since 2019. At the beginning of May, she announced her retirement as chief health officer; Google Health chief clinical officer Dr. Michael Howell will replace her when she steps down in August.

Oracle: Leveling up health records with AI

Oracle's long-awaited plans to revolutionize electronic medical records with AI will come to fruition this year, according to the tech company.

In October, Oracle unveiled its "next-generation" electronic health record system, which the company said would incorporate several of Oracle's cloud and AI capabilities, including clinical AI agents, search capabilities, and patient data analytics. Its goal is to seamlessly integrate AI into providers' workflows and automate healthcare administrative tasks. The AI-powered EHR is slated for release to early adopters this year, the company said.

Oracle made its biggest investment toward this goal in 2022, when it bought EHR company Cerner for $28.3 billion and rebranded it as Oracle Health. Its efforts to transform Cerner have hit a number of snags, however. BI reported in May 2024 that flaws in Cerner's EHR rollout at the Department of Veterans Affairs led to thousands of medical orders vanishing, resulting in delayed or missed treatments for many veterans, and leaving Oracle scrambling to fix Cerner's tech after the acquisition.

Oracle is also financially backing Stargate, a joint venture with OpenAI, SoftBank, and investment firm MGX, to invest up to $500 billion in US-based AI infrastructure. Oracle cofounder and executive chairman Larry Ellison said in January that the team is working on tools for advancing disease detection, including cancer detection, with AI.

Salesforce: Selling ready-made health AI agents

Salesforce is jumping onto the AI agent trend with a range of pre-built AI assistants for healthcare organizations.

In February, the cloud-based software company announced Agentforce for Health, a library of AI agents that healthcare companies can use to automate patient tasks like appointment booking, provider tasks like summarizing a patient's medical history, and life sciences tasks like clinical trial matching. Salesforce also landed a partnership with electronic health record company Athenahealth to integrate Agentforce for Health capabilities into Athenahealth's EHR.

The agentic AI release builds on Salesforce's previous healthcare assistant tool, Einstein Copilot, which launched in March 2024 to allow providers to query patient data as consolidated in Salesforce's Health Cloud.

Salesforce is also powering new AI platforms created on its Health Cloud, such as Blue Shield of California's AI-powered prior authorization tech, which the two companies said would enter testing in early 2025.

Palantir: Partnering with health systems on AI transformation

Palantir is primarily known for its billions of dollars in defense tech contracts with the US government. But it's also spent the past four years building out a healthcare business, working with top healthcare systems to improve their operations with AI.

Palantir works with health systems including Cleveland Clinic, Tampa General, and Nebraska Medicine to automate key hospital functions, including revenue cycle management, staffing and scheduling, and patient flow, co-heads of healthcare Jeremy David and Drew Goldstein told BI in April.

Palantir said in May it would work with the Joint Commission, a nonprofit that evaluates healthcare organizations for accreditation, to streamline the Joint Commission's data collection process and help hospitals manage their quality standards using Palantir's AI and analytics.

Palantir is also partnering with R1 RCM, the AI-powered revenue cycle management company acquired by TowerBrook and CD&R in an $8.9 billion take-private deal in August. Beyond its partnerships, Palantir aims to equip more healthcare startups with its AI tools through its software platform HealthStart.

• Tech giants are known for their challenging interviews.
• Google, Meta, and Nvidia top the list of rigorous interviews with multiple rounds and assessments.
• But tough questions show up across industries, according to employee reports on Glassdoor.

It's tough to break into high-paying companies.

Google is notorious for having a demanding interview process. Aside from putting job candidates through assessments, preliminary phone calls, and asking them to complete projects, the company also screens candidates through multiple rounds of interviews.

Typical interview questions range from open-ended behavioral ones like "tell me about a time that you went against the status quo" or "what does being 'Googley' mean to you?" to more technical ones.

At Nvidia, the chipmaking darling of the AI boom, candidates must also pass through rigorous rounds of assessments and interviews. "How would you describe __ technology to a non-technical person?" was a question a candidate interviewing for a job as a senior solutions architect shared on the career site Glassdoor last month. The candidate noted that they didn't receive an offer.

Tech giants top Glassdoor's list of the hardest companies to interview with. But tough questions show up across industries — from luxury carmakers like Rolls-Royce, where a candidate said they were asked to define "a single crystal," to Bacardi, where a market manager who cited a difficult interview, and no offer, recalled being asked, "If you were a cocktail what would you be and why?"

The digital PR agency Reboot Online analyzed Glassdoor data to determine which companies have the most challenging job interviews. They focused on "reputable companies" listed in the top 100 of Forbes' World's Best Employers list and examined 313,000 employee reviews on Glassdoor. For each company, they looked at the average interview difficulty rating as reported on Glassdoor.

Here's a list of the top 90 companies that put candidates through the ringer for a job, according to self-reported reviews on Glassdoor.

• The Abilene City Council approved a tax abatement for Crusoe's data center campus in Texas.
• Business Insider obtained a copy of the agreement via a public records request.
• Crusoe, the data center developer, could get an 85% property tax break.

The developer of the first Stargate data center in Abilene, Texas, could get an 85% tax break on billions of dollars worth of property, according to a new tax abatement agreement obtained by Business Insider through a public records request.

To qualify for the tax break, AI startup Crusoe will need to spend a minimum of $2.4 billion of a $3.5 billion "targeted investment."

President Donald Trump announced Stargate, a joint venture between OpenAI, Oracle, and SoftBank, at a White House press conference in January. That month, Business Insider reported that Oracle is the tenant of two data center buildings under construction on Crusoe's data center campus in Abilene, which is widely assumed to be Stargate's first site. The estimated cost for those buildings, $1.1 billion, will count toward the targeted investment.

Since then, Crusoe has registered two more data center buildings with a state agency on the site. No tenant is listed for one of those buildings; Oracle is listed as the tenant for the other.

Crusoe has agreed to build six new data centers, each with a minimum square footage of 100,000, on Lancium's land in Abilene. These will join the two it is already constructing for Oracle.

Data centers create jobs—although some say it's not enough

A local news outlet first reported the abatement after the Abilene City Council approved it in February, naming the site's landowner Lancium as the tax break recipient.

The abatement requires Crusoe and Lancium to create 357 new full-time jobs with minimum salaries of $57,600. The does not specify what kind of jobs need to be created, though the hiring timeframe indicates that construction jobs will likely count toward the quota.

Critics of tax abatements for data centers have said that they don't create enough jobs to justify the tax breaks. Proponents like to point out that the sites can employ thousands of skilled laborers during the construction period.

Property tax abatements are common financial incentives used by local governments to encourage companies to operate in their cities and towns. Specific property tax breaks for data centers vary widely depending on locality. Loudoun County, Virginia, home to the world's largest concentration of data centers, does not abate property taxes for data centers. In New Albany, Ohio, a fast-growing data center, data centers have been granted 100% property tax abatements.

Stargate's expansion plans

Oracle executive chairman and CTO Larry Ellison has said the first Stargate data center is in Abilene, though scant details have emerged about the massive AI infrastructure initiative since Trump's announcement.

In February, OpenAI CFO Sarah Friar wrote in a LinkedIn post that the company was evaluating additional locations in Texas, Pennsylvania, Oregon, and Wisconsin for Stargate.

Crusoe announced it is developing a data center campus in Abilene, though it has not publicly confirmed that it is connected to Stargate. Public filings in Texas show that Oracle is Crusoe's tenant, Business Insider reported in January.

To finance the Abilene development, Crusoe entered into a $3.4 billion joint venture with private credit company Blue Owl. JP Morgan has lent $2.3 billion in construction financing to the project.

Lancium declined to comment. Crusoe, Oracle, and the Development Corporation of Abilene did not respond to requests for comment from BI.

Contact Ellen Thomas at [email protected] or on Signal at 929-524-6964.

Oracle isn’t commenting on recent reports that it has experienced two separate data breaches that have exposed sensitive personal information belonging to thousands of its customers.

The most recent data breach report, published Friday by Bleeping Computer, said that Oracle Health—a health care software-as-a-service business the company acquired in 2022—had learned in February that a threat actor accessed one of its servers and made off with patient data from US hospitals. Bleeping Computer said Oracle Health customers have received breach notifications that were printed on plain paper rather than official Oracle letterhead and were signed by Seema Verma, the executive vice president & GM of Oracle Health.

The other report of a data breach occurred eight days ago, when an anonymous person using the handle rose87168 published a sampling of what they said were 6 million records of authentication data belonging to Oracle Cloud customers. Rose87168 told Bleeping Computer that they had acquired the data a little more than a month earlier after exploiting a vulnerability that gave access to an Oracle Cloud server.

Read full article

Comments

TikTok has until April to finalize a sale to a U.S.-based buyer, yet uncertainties about the future of the short-form video app remain. However, a new report has emerged, suggesting that Oracle is the leading candidate to serve as the cloud technology partner for managing TikTok in the U.S., according to The Information.  Sources including […]

© 2024 TechCrunch. All rights reserved. For personal use only.

• Tech companies seek to hire thousands of skilled foreign workers through H-1B visas each year.
• President Donald Trump's immigration crackdown is raising questions about the future of such visas.
• See which tech companies file for the most H-1B visas, according to publicly available data.

Tech industry giants are hiring thousands of foreign workers through H-1B visas each year, even as the program faces renewed scrutiny under President Donald Trump's second term and growing skepticism from Silicon Valley leaders who once championed it.

The H-1B program allows US companies to hire up to 85,000 foreign workers with specialized skills annually. Workers are chosen through an annual lottery, which kicked off last week and will run through March 24. While Trump expressed support in December, calling it "a great program" that he has "used many times," key figures in his political base have voiced opposition.

Things escalated late last year when Trump appointed Sriram Krishnan, a first-generation Indian American who immigrated to the US from India in 2007, to serve as a senior White House advisor for AI. The appointment drew backlash from some MAGA supporters, including former Trump aide Steve Bannon, who called the program "a total and complete scam to destroy the American worker."

In late January, Republican Senators John Kennedy and Rick Scott introduced a joint resolution under the Congressional Review Act to reverse a Biden-era rule that extended the automatic renewal period for employment authorization documents from 180 days to 540 days. Kennedy said the extension "hampers the Trump administration's efforts to enforce our immigration laws," signaling there would be additional scrutiny of work permits for foreign nationals.

Even tech leaders have softened their stance. Elon Musk and Marc Andreessen, once unequivocal supporters, have recently acknowledged the need for improvement.

Andreessen said on Lex Fridman's podcast last month that the US has been conducting "a 60-year social engineering experiment to exclude native-born people from the educational slots and jobs that high-skill immigration has been funneling foreigners into."

Musk has called for raising the minimum salary requirements for people on H-1B visas and adding a "yearly cost" to make it more expensive for companies to hire from overseas. "I've been very clear that the program is broken and needs major reform," he posted on X.

While the program's future remains uncertain, any significant changes or restrictions to H-1B visas would profoundly impact America's largest technology companies, which have built their workforces around access to global talent.

Business Insider used publicly available data from the Department of Labor and US Citizenship and Immigration Services to analyze which tech companies filed the most H-1B requests during the 2024 government fiscal year. The data comes from applications submitted by businesses seeking to sponsor skilled workers' visas.

Our analysis shows that tech giants collectively file for thousands of these visas annually, using them to fill critical roles that they claim cannot be adequately staffed domestically.

Notably, not every visa filing results in an actual hire, and occasionally multiple filings might be associated with a single position. Companies sometimes submit new applications to accommodate amendments or extend existing visas. Nevertheless, the data available to the public offers a reliable glimpse into the H-1B visa requirements of major corporations.

We have excluded IT consulting firms from this analysis to focus specifically on tech product companies, despite consulting giants like Infosys and Tata Consultancy Services traditionally being among the program's largest users.

The analysis reveals that tech giants like Amazon, Microsoft, Google, Meta, and Apple are among the program's heaviest users, with thousands of filings each.

While most positions are for software engineers and other technical roles, companies also use the program to fill specialized positions in research, product management, and data science. The employee head count for each firm comes from the latest publicly available data such as the company's latest annual report, their corporate website, or according to sources BI spoke with.

The firms listed either did not respond to a request for comment or declined to comment on the record.

Here are the top 40 tech companies sponsoring H-1B visas, ranked by their number of filings:

1. Amazon

Total certified H-1B filings: 14,783 (including 23 for Whole Foods).

Total employees worldwide: 1,556,000 as of the end of 2024.

2. Microsoft

Total certified H-1B filings: 5,695 flings (including 970 from LinkedIn).

Total employees worldwide: 228,000 as of the second quarter of 2024.

3. Alphabet

Total certified H-1B filings: 5,537 (including 115 from Waymo and Verily).

Total employees worldwide: 183,323 as of the end of 2024.

4. Meta

Total certified H-1B filings: 4,844.

Total employees worldwide: 74,067 as of the end of 2024.

5. Apple

Total certified H-1B filings: 3,880.

Total employees worldwide: 164,000 as of the third quarter of 2024.

6. IBM

Total certified H-1B filings: 2,907.

Total employees worldwide: More than 293,400 as of the end of 2024.

7. Intel

Total certified H-1B filings: 2,558.

Total employees worldwide: 108,900 as of the end of 2024.

8. Oracle

Total certified H-1B filings: 2,141.

Total employees worldwide: 159,000 as of the end of May 2024.

9. Tesla

Total certified H-1B filings: 1,677.

Total employees worldwide: 125,665 as of the end of 2024.

10. Bytedance

Total certified H-1B filings: 1,611.

Total employees worldwide: More than 150,000, according to the company's website.

11. Salesforce

Total certified H-1B filings: 1,525 (A Salesforce spokesperson said that the company filed 1,808 H-1B petitions in fiscal year 2024 including new hires, amendments, and extensions).

Total employees worldwide: 76,453 as of the end of January 2025.

12. Nvidia

Total certified H-1B filings: 1,519.

Total employees worldwide: 36,000 as of the end of fiscal year 2025.

13. Cisco

Total certified H-1B filings: 1,330.

Total employees worldwide: 90,400 as of the end of fiscal year 2024.

14. Qualcomm

Total certified H-1B filings: 1,291.

Total employees worldwide: 49,000 employees as of the end of the third quarter of 2024.

15. Adobe

Total certified H-1B filings: 787.

Total employees worldwide: More than 30,708 of as November 2024.

16. Intuit

Total certified H-1B filings: 770.

Total employees worldwide: 18,200 at the end of fiscal year 2024.

17. Uber

Total certified H-1B filings: 703.

Total employees worldwide: 31,100 as of the end of 2024.

18. Paypal

Total certified H-1B filings: 623.

Total employees worldwide: 24,400 as of the end of 2024.

19. eBay

Total certified H-1B filings: 548 (An eBay spokesperson said eBay filed 494 H-1B visas in fiscal year 2024, noting that the publicly available information doesn't disclose the exact number of roles hired for.)

Total employees worldwide: 11,500 as of the end of 2024.

20. Rivian

Total certified H-1B filings: 584.

Total employees worldwide: 14,861 as of the end of 2024.

21. ServiceNow

Total certified H-1B filings: 578.

Total employees worldwide: 26,293 as of the end of 2024.

22. HP

Total certified H-1B filings: 533.

Total employees worldwide: 58,000 as of the end of 2024.

23. Dell

Total certified H-1B filings: 489.

Total employees worldwide: 120,000 as of February 2, 2024.

24. Lucid Motors

Total certified H-1B filings: 488.

Total employees worldwide: 6,800 as of the end of 2024.

25. DoorDash

Total certified H-1B filings: 427.

Total employees worldwide: 23,700 as of the end of 2024.

26. Fiserv

Total certified H-1B filings: 403.

Total employees worldwide: 38,000 as of the end of 2024.

27. Micron Technology

Total certified H-1B filings: 369.

Total employees worldwide: 48,000 as of August 29, 2024.

28. VMWare

Total certified H-1B filings: 359.

Total employees worldwide: 16,000 according to Business Insider's sources.

29. ADP

Total certified H-1B filings: 350.

Total employees worldwide: 64,000 as of June 2024.

30. Workday

Total certified H-1B filings: 347.

Total employees worldwide: 20,400 as of January 31, 2025.

31. Expedia

Total certified H-1B filings: 331.

Total employees worldwide: 16,500 as of the end of 2024.

32. MathWorks

Total certified H-1B filings: 295.

Total employees worldwide: 6,500, according to the corporate website.

33. Snowflake

Total certified H-1B filings: 285.

Total employees worldwide: 7,004 as of January 31, 2024.

34. Databricks

Total certified H-1B filings: 283.

Total employees worldwide: More than 7,000, according to the company's website.

35. Synopsys

Total certified H-1B filings: 267.

Total employees worldwide: 20,000 as of November 2024.

36. Stripe

Total certified H-1B filings: 265.

Total employees worldwide: Approximately 8,200 according to BI's reporting.

37. Snap

Total certified H-1B filings: 258

Total employees worldwide: 4,911 as of December 2024.

38. Netflix

Total certified H-1B filings: 256.

Total employees worldwide: 14,000 as of the end of 2024.

39. Block

Total certified H-1B filings: 231.

Total employees worldwide: 11,372 as of the end of 2024.

40. Pinterest

Total certified H-1B filings: 225.

Total employees worldwide: 4,666 as of the end of 2024.

• Major companies have been relocating their headquarters to Texas.
• Since 2020, 200 companies, including Tesla and Chevron, have moved HQs to Texas, state data say.
• Gov. Greg Abbott has cited a good regulatory environment, drawing firms from states like California.

Texas has become a hot spot for the corporate operations of major companies across the US.

Since 2020, a growing number of major businesses have moved their headquarters or reincorporated in Texas, flocking from pricier states like California. Many cite the lower cost of living and benefits for corporations as reasons they chose the Lone Star State.

KFC's parent company, Yum! Brands, announced in February that it would move its US headquarters from Louisville, Kentucky, to dual HQs in Plano, Texas, and Irvine, California. Meanwhile, Elon Musk, CEO of SpaceX and Tesla and owner of X, has announced that all three companies will be headquartered in Texas.

HQ relocations were in an "acceleration period" in 2020 and 2021, the office of Gov. Greg Abbott said, with a total of 121 companies moving to Texas during that time. The number of those that moved from California made up more than half of the relocations.

Since then, the rate has leveled out to be consistent with historical data. A total of 200 companies have moved to Texas since 2020, according to the data from Gov. Abbott. In 2024, 24 companies, including Chevron and SpaceX, announced they would establish an HQ there.

The moves are fueled by the "reasonable regulatory environment," "exceptional quality of life," and the lower cost of operating a business in Texas, Abbott's office said in its report.

Here's a list of companies that have shifted their business operations to Texas.

KFC

Yum! Brands announced in February that it would establish two HQ locations in Texas and California to "foster greater collaboration among brands and employees."

About 100 KFC corporate workers will have to relocate from Louisville, Kentucky, to Texas over six months. The company said it will also ask 90 US-based remote workers to return to the office and relocate to "the campus where their work happens."

Yum! Brands is the parent company of KFC, Taco Bell, Pizza Hut, and Habit Burger & Grill. The KFC Foundation and Yum! Brands are expected to maintain corporate offices in Kentucky.

The New York Stock Exchange Chicago

On February 12, the New York Stock Exchange announced it would move its Chicago branch to Dallas. NYSE Chicago will be reincorporated to NYSE Texas, "offering companies the opportunity to list their securities" there.

"As the state with the largest number of NYSE listings, representing over $3.7 trillion in market value for our community, Texas is a market leader in fostering a pro-business atmosphere," Lynn Martin, NYSE Group president, said in a press release.

Chevron

Chevron announced in August 2024 that its headquarters would move from San Ramon, California, to Houston before the end of the year.

The energy giant said the relocation would "enable better collaboration and engagement with executives, employees, and business partners."

The oil company had been sued by California, which accused Chevron and other energy giants of downplaying the risks of fossil fuels.

But Texas Gov. Greg Abbott tweeted, "WELCOME HOME Chevron! Texas is your true home."

Before the move, Chevron had about 7,000 employees in the Houston area and 2,000 in San Ramon. It said it expects all corporation functions to move to Texas by 2029.

X

Musk announced X's relocation from California to Texas at the same time as SpaceX in July 2024, citing living costs, safety, and political reasons.

He said California laws are "attacking both families and companies" and expressed concerns over the safety of San Francisco. Court filings from September 2024 showed that Musk requested to change X's HQ address from San Francisco to Bastrop, Texas, Forbes reported.

Tesla

Tesla was the first of Musk's companies to move from Silicon Valley to Texas. Musk officially moved Tesla's headquarters from Palo Alto, California, to Austin in 2021, citing the lack of affordable housing in the Bay Area.

"There's a limit to how big you can scale in the Bay Area," Musk said at the time.

Oracle

Oracle moved its corporate HQ to Austin in 2020, ending its four-decade tenure in Silicon Valley.

The move offered employees "more flexibility about where and how they work," a spokesperson told Business Insider at the time.

Though it's been years since the move, Oracle's California offices employ nearly triple the number of workers than its Texas HQ, Bloomberg reported in 2024.

CBRE

Brokerage giant CBRE moved its HQ from Los Angeles to Dallas in 2020.

It was established in San Francisco over a century ago, though CBRE said it had large operations in North Texas before the official move to Dallas.

According to data from CBRE, Texas led the pack in net gains of Fortune 500 companies relocating between 2018 and 2023.

AECOM

AECOM, a Fortune 500 construction firm, said it would relocate its headquarters from Los Angeles to Dallas in 2021.

The company called Texas a "talent magnet" for consulting and engineering, and CEO Troy Rudd participated in the move to AECOM's existing Texas offices from California.

SpaceX

Musk announced his plans to relocate SpaceX from Hawthorne, California, to Starbase, Texas, on X in July 2024.

The move is Musk's response to Gov. Gavin Newsom signing AB 1955, prohibiting schools from enforcing policies that would require parents to be notified about students who may identify as transgender.

"Because of this law and the many others that preceded it, attacking both families and companies, SpaceX will now move its HQ from Hawthorne, California, to Starbase, Texas," Musk said in an X post.

Charles Schwab

Financial services company Charles Schwab moved its HQ to Westlake, Texas, in 2021, citing California's high taxes. It was formerly based in San Francisco.

"The costs of doing business here are so much higher than some other place," Chairman and founder Charles Schwab told Forbes.

McKesson

McKesson announced that it'd move its HQ from San Francisco to Las Colinas, Texas, in 2018, with plans to move most jobs from Silicon Valley to Texas and other hub locations by 2021.

Four years after the move, CEO Brian Tyler said the city "was absolutely the right community for McKesson to call home."

"Since making the move to Irving, McKesson has quickly benefited from the deep, diverse talent pool in the Dallas area, the ease of travel, and the very engaged business community," he said.

• A TD Cowen analyst said that Microsoft had canceled plans for new data center developments.
• The news suggested that the staggering forecasts for data center growth could be overblown.
• Experts say the industry's growth will still be enormous but is now undergoing a reality check.

Microsoft recently appeared to scale back its data center development pipeline, which would be an extraordinary step back after a period of furious growth by the tech giant. But did it really?

The concerns about whether air was coming out of a data center bubble were raised by a TD Cowen analyst, Michael Elias, in a report dated February 21. Elias wrote that Microsoft, one of the largest data center operators in the country, had recently torn up leases for "multiple +100 megawatt deals in multiple markets that were in early/mid-stages of negotiations."

Elias also said the company had let go of more than a gigawatt of preliminary data center commitments it had made and also five longer-term development deals in prime data center markets. The company's decision-making was "tied to Microsoft potentially being in an oversupply position."

Microsoft's pullback sent a shudder through the data center market, which has seen staggering forecasts for growth in the coming years.

For some, however, the news simply reflected a more modest revision of the sector's extraordinary recent projections for growth.

"I can't think of the big five that haven't done this every 12 or 18 months," said one data center development executive, referring to the industry's largest users: Amazon, Meta, Google, Microsoft, and Oracle. The executive did not want to be identified by name because major customers in the industry prize confidentiality.

"This is not new and definitely not the biggest one we've had in the last three years," the executive said, saying that Meta had canceled major data center commitments it had in recent years in order to reset its strategy, including its data center business's customer and technology focus.

After that reevaluation, Meta rebooted its torrid growth, including the recent announcement of a $10 billion data center campus in Louisiana.

A spokeswoman for Meta declined to comment.

Microsoft still has plans for robust growth

Dan Thompson, an analyst at S&P Global who covers the data center industry, said that the magnitude of the data center boom was bound to modulate because the projections included speculative projects and, potentially, the double counting of tenants who may have expressed interest in multiple projects and/or regions for the same requirement.

"Some of these announcements are not going to get built," Thompson said. "I don't see it as a reflection on the data center industry."

Thompson said that there was a need to differentiate between credible growth and the frothiest forecasts for the industry's expansion. He said that S&P plans to reorganize its data center projections this year into buckets that group projects based on their likelihood of coming to fruition.

"We are modifying our reports this year, for every market, basically draw the line in the sand and say, OK, this is the part where we think: that is real," Thompson said.

What is clear, nonetheless, is that a data center boom is afoot across the nation.

S&P Global projects that the industry will grow from about 35.4 gigawatts of capacity today to almost 82 gigawatts by the end of the decade, an 131% increase. There have been even more ambitious projections that more than 90 gigawatts of data centers could be online by 2029.

The development is being undertaken to commercialize and develop artificial intelligence and also cater to society's yawning digital footprint. Data centers provide the computing and storage that power the internet and a host of increasingly vital functions, such as autonomous vehicles, Zoom meetings, and cloud computing.

A spokeswoman for Microsoft suggested in a statement that the company had pumped the brakes on development, while also highlighting the enormous scale of its data center expansion.

"Thanks to the significant investments we have made up to this point, we are well positioned to meet our current and increasing customer demand," the spokeswoman stated. "Last year alone, we added more capacity than any prior year in history."

Microsoft "will continue to grow strongly in all regions," she said, and that the company's plans to "spend over $80B on infrastructure this FY remains on track as we continue to grow at a record pace to meet customer demand."

Elias's report didn't cite specific examples of Microsoft's retrenchment, except for the recent news of its decision to reevaluate a large data center campus it is building in Mount Pleasant, Wisconsin, outside of Milwaukee.

Reached by phone, Elias declined to comment further.

Sean Ryan, a spokesman for the Village of Mount Pleasant, said that Microsoft had "paused in some of its construction work in order to incorporate new data center designs" and that "Village officials have no reason to believe this will affect the overall scope or nature" of the project.

• Bill Gates confirmed he's dating philanthropist Paula Hurd following his divorce from Melinda French Gates in 2021.
• Hurd, the widow of late Oracle CEO Mark Hurd, has worked in sales and service leadership roles in her career.
• The Microsoft cofounder also mentioned Paula Hurd in the acknowledgments section of his new book, "Source Code."

Bill Gates has confirmed he's seeing someone after his divorce four years ago.

The Microsoft cofounder, who was previously married to Melinda French Gates for 27 years until 2021, said on "The Today Show" on Tuesday that he's "lucky to have a serious girlfriend named Paula."

"We're having fun, going to the Olympics and lots of great things," he added.

Gates' girlfriend, Paula Hurd, was formerly married to the late Mark Hurd, whose career included tenures as the CEO of Oracle and Hewlett-Packard. They share two children and were married for nearly 30 years until Mark's death in 2019.

Hurd got her bachelor's degree in business administration from the University of Texas at Austin in 1984. During her career, she held various sales and services leadership roles at NCR Corporation and, in her final position at the company, was its vice president of services, heading up a team managing global strategic partnerships, according to her bio from Baylor University.

Hurd and her late husband, a Baylor alum, made several notable donations to the university, including leading the funding of its new welcome center, which was named the Mark and Paula Hurd Welcome Center.

"Our family is absolutely thrilled with the Hurd Welcome Center," Paula Hurd said in a 2023 ribbon-cutting ceremony. "When we made this gift, Mark had a vision for how a welcome experience could transform this corner of campus, and the outcome has by far exceeded our expectations and dreams. I know he would be proud of not only what has been built but the way in which others in the Baylor Family rallied around the Give Light Campaign to generate more than $1.39 billion in gifts to support the University he loved."

Hurd, who is a member of Baylor's Board of Regents, is also chairperson of the Universal Tennis Foundation, for which she sponsors annual awards for college players transitioning to professional tennis.

Hurd and Gates share a love of tennis. Gates has played with tennis pros, including Rafael Nadal and Andre Agassi, in charity games and previously told Business Insider that his pandemic routine included socially distanced tennis.

Gates and Hurd were publicly spotted together as early as 2022, when they attended the Laver Cup tennis tournament together. They also watched a tennis match at the Australian Open in January 2023. Also that year, Gates accompanied Hurd to a ceremony at Baylor officially opening the university's new welcome center.

Gates and Hurd made their red carpet debut in April at the 2024 Breakthrough Prize Ceremony. As Gates mentioned in his "Today Show" interview, they attended the 2024 Paris Olympics together as well. Representatives for Gates and Hurd did not immediately respond to requests for comment.

Gates also included Hurd in the acknowledgments of his new memoir, "Source Code," which came out this month.

"Early readers of the manuscript included Paula Hurd, Marc St. John, and Sheila Gulati. The close read from dear and trusted friends provided much-needed thoughtful and insightful feedback at critical stages in the writing," he wrote.

• There's an astonishing amount of news coming out of the tech world right now.
• This week, it's DeepSeek. Before that, it was the TikTok ban-then-maybe-sale, huge infrastructure deals, and, of course, Trump.
• I wanted an insider's view on Silicon Valley's hopes and fears, so I talked to The Information's founder, Jessica Lessin.

This week's news about DeepSeek — the Chinese-built AI engine that's supposedly just as powerful as top-of-the-line stuff out of the US but built at a fraction of the cost — shook Silicon Valley and beyond.

But that's just one of a series of issues roiling tech. The big difference: At the moment, DeepSeek isn't intertwined with Donald Trump.

Just about everything else in the tech world right now seems deeply enmeshed with Trump 2.0 — from a potential forced sale of TikTok to the parade of the most powerful CEOs in the world aligning themselves with the new president.

I wanted to get a view of all this from the Bay Area, so I turned to Jessica Lessin, the deeply sourced journalist who has been covering tech for decades — first at The Wall Street Journal and now at The Information, the publication she founded in 2013.

We had what podcasters like to call a "wide-ranging conversation" that went in all sorts of directions, and you can listen to the whole thing on my "Channels" show. Here's an edited excerpt, focusing on DeepSeek's reverberations and the swirl around TikTok.

The DeepSeek news is still shaking Silicon Valley and Wall Street, and people are still sorting out what it means. What is your takeaway?

I think what's happening here is that DeepSeek is striking at a question that is so important: What is the comparative advantage that businesses can have around AI, and where is the money going to be made over the very long term?

We've seen extraordinary growth, extraordinary products coming out of AI. Everyone says we're in early innings, and we are. And so how is this technology going to advance? How easy is it going to be for others to build great models, and what great businesses are going to be built? I've been feeling like that's an open question for many, many months. And I feel like what happened with DeepSeek is — everyone kind of realized, "Wow, this is still a very open question."

Over the last year, there was a conventional wisdom jelling that said the AI winners are going to be the biggest technology companies — the ones that can afford all the chips, all the engineers, all the enormous power these things use. This has upended everything. It's been fascinating to watch the Wall Street reaction.

I think what Wall Street is doing is in some way actually like a meme moment. Not to get too meta here, but DeepSeek kind of had that feeling: The model's been out for a while. They released another model. All of a sudden, Marc Andreessen and Bill Gurley are tweeting about it. It captures Silicon Valley's imagination-slash-paranoia. And so I think Wall Street did what Wall Street does, which is sort of have blunt reactions on all fronts.

You guys had a story, which I assume contributed to the panic, reporting that Meta had spun up war rooms to figure out what they were going to do about DeepSeek. You can imagine Wall Street seeing a story like that coming from your publication and thinking: "Mark Zuckerberg is worried about this. We should be worried about this."

I just think Wall Street is still trying to figure out what to make of AI in general. We are just days after major infrastructure announcements that investors of all stripes are trying to digest. Now we have this evidence that, you know, it may be easier to catch up. So I think so much remains to be seen. But to your earlier point, I still think the largest companies have a lot to gain with AI because they have the distribution; they have the existing business lines that can be turbocharged with it. I don't think that's changed with this announcement.

Donald Trump says he's going to find a non-Chinese buyer for some or all of TikTok. What's your best bet on who that will be?

I go back to the last time this came around [in 2020], and the conversations were furthest along with the consortium that included Microsoft, Walmart, and Oracle. TikTok is strategically valuable to these Big Tech companies because it's one of the largest customers in the world of cloud computing, AI chips, data centers, and so forth. So I sort of believe — and know, to some degree — all of those entities are back at the table. And, you know, you can't count out Elon Musk.

It's surprising how quiet he is about all of that.

It's surprising.

He bought Twitter and, in many ways, it hasn't worked out. You could argue that it has worked for him in other ways. But TikTok is way more influential than Twitter ever was or could be.

Out of his many challenges, selling Teslas with Full Self-Driving capabilities in China is near the top of the list. He is not allowed to do it. It's hurting his position in the China market, which is essential. And so he's looking for every bargaining chip possible. I think his relationship with Trump has been driven in part by this need to negotiate with China. So I think that's important context here. And I'm sure that, like any good businessman, he'd be trying to use this to his advantage. I think we're going to have to just play this out a little bit and see what happens. And all options are still on the table. I mean, you could also have the Chinese government just saying, no way.

Tech leaders are embracing Trump very publicly. How much of that is because they want to get something from him, and how much of it is because they're afraid of angering him?

I think the story is a little more complicated than just wanting to be in his good graces. I think Silicon Valley senses there is a moment to unwind many policies — ranging from tax law to DEI — of the last several years. So tech leaders are really taking this moment. They are aligned with Trump on many issues. They're taking this moment to try and seize that, to get their companies back to where they want them to be.

• DeepSeek, a Chinese open-source AI firm, is taking over the discussion in tech circles.
• Tech stocks, especially Nvidia, plunged Monday.
• Companies leading the AI boom could be in for a reset as DeepSeek upends the status quo.

DeepSeek, a Chinese company with AI models that compete with OpenAI's at a fraction of the cost, is generating almost as many takes as tokens.

Across Silicon Valley, executives, investors, and employees debated the implications of such efficient models. Some called into question the trillions of dollars being spent on AI infrastructure since DeepSeek says its models were trained for a relative pittance.

"This is insane!!!!" Aravind Srinivas, CEO of startup Perplexity AI, wrote in response to a post on X noting that DeepSeek models are cheaper and better than some of OpenAI's latest offerings.

The takes on DeepSeek's implications are coming fast and hot. Here are eight of the most common.

Take 1: Generative AI adoption will explode

"Jevons paradox strikes again!" Microsoft CEO Satya Nadella posted on X Monday morning. "As AI gets more efficient and accessible, we will see its use skyrocket, turning it into a commodity we just can't get enough of."

The idea that as tech improves, whether smarter, cheaper, or both, it will only bring in exponentially more demand is based on a 19th-century economic principle. In this case, the barrier to entry for companies looking to dip their toe into AI has been high. Cheaper tools could encourage more experimentation and further the technology faster.

"Similar to Llama, it lowers the barriers to adoption, enabling more businesses to accelerate AI use cases and move them into production." Umesh Padval, managing director at Thomvest Ventures told Business Insider.

That said, even if AI grows faster than ever, that doesn't necessarily mean the trillions of investments that have flooded the space will pay off.

Take 2: DeepSeek broke the prevailing wisdom about the cost of AI

"DeepSeek seems to have broken the assumption that you need a lot of capital to train cutting-edge models," Debarghya Das, an investor at Menlo Ventures told BI.

The price of DeepSeek's open-source model is competitive — 20 to 40 times cheaper to use than comparable models from OpenAI, according to Bernstein analysts.

The exact cost of building DeepSeek models is hotly debated. The research paper from DeepSeek explaining its V3 model lists a training cost of $5.6 million — a harrowingly low number for other providers of foundation models.

However, the same paper says that the "aforementioned costs include only the official training of DeepSeek-V3, excluding the costs associated with prior research and ablation experiments on architectures, algorithms, or data." So the $5 million figure is only part of the equation.

The tech ecosystem is also reacting strongly to the implication that DeepSeek's state-of-the-art model architecture will be cheaper to run.

"This breakthrough slashes computational demands, enabling lower fees — and putting pressure on industry titans like Microsoft and Google to justify their premium pricing," Kenneth Lamont, principal at Morningstar, wrote in a note on Monday.

He went on to remind investors that with early-stage technology, assuming the winners are set is folly.

"Mega-trends rarely unfold as expected, and today's dominant players might not be tomorrow's winners," Lamont wrote.

Dmitry Shevelenko, the chief business officer at Perplexity, a big consumer of compute and existing models, concurred that Big Tech players would need to rethink their numbers.

"It certainly challenges the margin structure that maybe they were selling to investors," Shevelenko told BI. "But in terms of accelerating the development of these technologies, this is a good thing." Perplexity has added DeepSeek's models to its platform.

Take 3: Considering a switch to DeepSeek

On Monday, several platforms that provide AI models for businesses— Groq and Liquid.AI to name two — added DeepSeek's models to their offerings.

On Amazon's internal Slack, one person posted a meme suggesting that developers might drop Anthropic's Claude AI model in favor of DeepSeek's offerings. The post included an image of the Claude model crossed out.

"Friendship ended with Claude. Now DeepSeek is my best friend." the person wrote, according to a screenshot of the post seen by BI, which got more than 60 emoji reactions from colleagues.

Amazon has invested billions of dollars in Anthropic. The cloud giant also provides access to Claude models via its Amazon Web Service platform. And some AWS customers are asking for DeepSeek, BI has exclusively reported.

"We are always listening to customers to bring the latest emerging and popular models to AWS," an Amazon spokesperson said, while noting that customers can access some DeepSeek-related products on AWS right now through tools such as Bedrock.

"We expect to see many more models like this — both large and small, proprietary and open-source — excel at different tasks," the Amazon spokesperson added. "This is why the majority of Amazon Bedrock customers use multiple models to meet their unique needs and why we remain focused on providing our customers with choice — so they can easily experiment and integrate the best models for their specific needs into their applications."

Switching costs for companies creating their own products on top of foundation models are relatively low, which is generating a lot of questions as to whether DeepSeek will overtake other models from Meta, Anthropic, or OpenAI in popularity with enterprises. (It's already number one in Apple's app store.)

DeepSeek, however, is owned by Chinese hedge fund High-Flyer and the same security concerns haunting TikTok may eventually apply to DeepSeek.

"While open-source models like DeepSeek present exciting opportunities, enterprises—especially in regulated industries—may hesitate to adopt Chinese-origin models due to concerns about training data transparency, privacy, and security," Padval said.

Security concerns aside, the software companies that sell APIs to businesses have been adding DeepSeek throughout Monday.

Take 4: Infrastructure players could take a hit

Infrastructure-as-a-service companies, such as Oracle, Digital Ocean, and Microsoft could be in a precarious position should more efficient AI models rule in the future.

"The sheer efficiency of DeepSeek's pre and post training framework (if true) raises the question as to whether or not global hyperscalers and governments, that have and intend to continue to invest significant capex dollars into AI infrastructure, may pause to consider the innovative methodologies that have come to light with DeepSeek's research," wrote Stifel analysts.

If the same quantity of work requires less compute, those selling only compute could suffer, Barclays analysts wrote.

"With the increased uncertainty, we could see share price pressure amongst all three," according to the analysts.

Microsoft and Digital Ocean declined to comment. Oracle did not respond to a request for comment in time for publication.

Take 5: Scaling isn't dead, it's just moved

For months, AI luminaries, including Nvidia CEO Jensen Huang have been predicting a big shift in AI from a focus on training to a focus on inference. Training is the process by which models are created while inference is the type of computing that runs AI models and related tools such as ChatGPT.

The shift in computing's total share to inference has been underway for a while, but now, change is coming from two places. First, more AI users means more inference demand. The second is that part of DeepSeek's secret sauce is how improvement takes place in the inference stage. Nvidia took a positive spin, via a spokesperson.

"DeepSeek is an excellent AI advancement and a perfect example of Test Time Scaling. DeepSeek's work illustrates how new models can be created using that technique, leveraging widely-available models and compute that is fully export control compliant," an Nvidia spokesperson told BI.

"Inference requires significant numbers of NVIDIA GPUs and high-performance networking. We now have three scaling laws: pre-training and post-training, which continue, and new test-time scaling."

Take 6: Open-source changes model building

The most under-hyped part of DeepSeek's innovations is how easy it will now be to take any AI model and turn it into a more powerful "reasoning" model, according to Jack Clark, an Anthropic cofounder, and a former OpenAI employee, wrote about DeepSeek in his newsletter Import AI on Monday.

Clark also explained that some AI companies, such as OpenAI, have been hiding all the reasoning steps that their latest AI models take. DeepSeek's models show all these intermediate "chains of thought" for anyone to see and use. This radically changes how AI models are controlled, Clark wrote.

"Some providers like OpenAI had previously chosen to obscure the chains of thought of their models, making this harder," Clark explained. "There's now an open-weight model floating around the internet which you can use to bootstrap any other sufficiently powerful base model into being an AI reasoner. AI capabilities worldwide just took a one-way ratchet forward."

Take 7: Programmers still matter

DeepSeek improved by using novel programming methods, which Samir Kumar, co-founder and general partner at VC firm Touring Capital, saw as a reminder that humans are still coding the most exciting innovations in AI.

He told BI that DeepSeek is "a good reminder of the talent and skillset of hardcore human low-level programmers."

Got a tip or an insight to share? Contact BI's senior reporter Emma Cosgrove at [email protected] or use the secure messaging app Signal: 443-333-9088.

Contact Pranav from a nonwork device securely on Signal at +1-408-905-9124 or email him at [email protected].

You can email Jyoti at [email protected] or DM her via X @jyoti_mann1

• Developers filed permits to operate natural gas turbines at Stargate's site in Abilene, Texas.
• The turbines' combined capacity is 360 MW, a fraction of the power Stargate's data centers need.
• That's still a lot of electricity — enough to power 90,000 Texas homes.

The developers of the first Project Stargate data center in Abilene, Texas, plan to build a natural gas plant on the site, according to public filings reviewed by Business Insider.

OpenAI, Oracle, and SoftBank have thus far provided little detail on the potential energy use of Stargate, a joint venture formed to spend $500 billion on AI infrastructure in the US and announced last week by the White House.

The filings said that once operational, the plant could produce up to 360.5 megawatts of power at any given time, up to 24 hours a day, seven days a week, 365 days a year. One MW of electricity can power 250 residential customers during peak hours, the Electric Reliability Council of Texas said. This means 360.5 MW is enough to power 90,000 homes in Texas — roughly double the amount of households in Abilene, according to the US Census Bureau.

These permit applications were filed with the Texas Commission on Environmental Quality by a consulting firm working on behalf of AI cloud startup Crusoe. Crusoe is the owner of Project Ludicrous, the $1.1 billion data center project in Abilene widely thought to be the first Stargate location.

An initial application to permit 10 simple-cycle natural gas turbines under Title V of the Clean Air Act — at a site that shares an address with Project Ludicrous — is under review by the TCEQ.

The application said the facility housing the turbines will be used for "primary and backup power" for "data centers and computing. " Its use will be "onsite only," meaning the power generated by the turbines won't be available to the local grid.

Half of the turbine models come from GE Vernova, and the other half come from Solar Turbines, a subsidiary of Caterpillar Inc. The turbines are designed to be installed and operational quickly, and some GE models can be up and running in as little as two weeks, the company has said in marketing materials.

A permit allowing diesel-fired backup generators to run at the site is already in effect.

Developing the plant is expected to cost half a billion dollars, a city document viewed by BI said.

Andrew Schmitt, a spokesperson for Crusoe, the startup developing Project Ludicrous, declined to comment on the record about the natural gas plant. He directed BI to the company's previous comments about the site's power, which "includes both on-and off-site renewable resources, including surrounding wind developments and a potential future large-scale onsite solar installation."

OpenAI and Oracle did not immediately respond to requests for comment.

Stargate will need lots of power

Morgan Stanley analysts on Friday estimated the entire Stargate project would require about 15 gigawatts.

That amount of power would be spread out across multiple locations, though the number of Stargate sites and how much power each would use is not publicly known. Some reports in the media and from financial analysts have indicated that Stargate sites could have a power demand of 5 GW, an amount drawn from a document said to have been pitched by OpenAI CEO Sam Altman to the Biden administration, first reported by Bloomberg.

If that is the case, then 360.5 MW, the capacity of the natural gas turbines at Project Ludicrous, would make up a tiny fraction — less than 1% — of how much power the site will ultimately need.

Some of that power will come from a substation, allowing Project Ludicrous to draw power from the local electric grid. The data center is located on land owned by energy tech company Lancium, which has built a 200 MW substation on the site, a company spokesperson told Business Insider. The company's website said it plans to increase capacity to five times that amount in 2025.

The race to develop AI infrastructure has caused electricity demand in the US to soar dramatically for the first time in two decades. Data center electricity use could triple in the US by 2028, according to the latest report from the Department of Energy.

The recent launch of DeepSeek has raised questions about AI infrastructure spending in the US, as the Chinese AI model is 20 to 40 times cheaper than OpenAI's and can run more efficiently with fewer chips.

Do you work in or have knowledge of the data center industry and have insight to share? Get in touch with this reporter at [email protected] or reach out via the encrypted messaging app Signal at +1-929-524-6924.

It was apparently a busy weekend for key players involved in Donald Trump's efforts to make a deal to save TikTok.

Perhaps the most appealing option for ByteDance could be if Trump blessed a merger between TikTok and Perplexity AI—a San Francisco-based AI search company worth about $9 billion that appears to view a TikTok video content acquisition as a path to compete with major players like Google and OpenAI.

On Sunday, Perplexity AI submitted a revised merger proposal to TikTok-owner ByteDance, reviewed by CNBC, which sources told AP News included feedback from the Trump administration.

Read full article

Comments

• China's DeepSeek model challenges US AI firms with cost-effective, efficient performance.
• DeepSeek's model, using modest hardware, is 20 to 40 times cheaper than OpenAI's.
• DeepSeek's efficiency raises questions about US investments in AI infrastructure.

The bombshell that is China's DeepSeek model has set the AI ecosystem alight.

The models are high-performing, relatively cheap, and compute-efficient, which has led many to posit that they pose an existential threat to American companies like OpenAI and Meta — and the trillions of dollars going into building, improving, and scaling US AI infrastructure.

The price of DeepSeek's open-source model is competitive — 20 to 40 times cheaper to run than comparable models from OpenAI, Bernstein analysts said.

But the potentially more nerve-racking element in the DeepSeek equation for US-built models is the relatively modest hardware stack used to build them.

The DeepSeek-V3 model, which is most comparable to OpenAI's ChatGPT, was trained on a cluster of 2,048 Nvidia H800 GPUs, according to the technical report published by the company.

H800s are the first version of the company's defeatured chip for the Chinese market. After the regulations were amended, the company made another defeatured chip, the H20 to comply with the changes.

Though this may not always be the case, chips are the most substantial cost in the large language model training equation. Being forced to use less powerful, cheaper chips created a constraint that the DeepSeek team has ostensibly overcome.

"Innovation under constraints takes genius," Sri Ambati, the CEO of the open-source AI platform H2O.ai, told Business Insider.

Even on subpar hardware, training DeepSeek-V3 took less than two months, the company's report said.

The efficiency advantage

DeepSeek-V3 is small relative to its capabilities and has 671 billion parameters, while ChatGPT-4 has 1.76 trillion, which makes it easier to run. But DeepSeek-V3 still hits impressive benchmarks of understanding.

Its smaller size comes in part by using a different architecture than ChatGPT, called a "mixture of experts." The model has pockets of expertise built in, which go into action when called upon and sit dormant when irrelevant to the query. This type of model is growing in popularity, and DeepSeek's advantage is that it built an extremely efficient version of an inherently efficient architecture.

"Someone made this analogy: It's almost as if someone released a $20 iPhone," Jared Quincy Davis, the CEO of Foundry, told BI.

The Chinese model used a fraction of the time, a fraction of the number of chips, and a less capable, less expensive chip cluster. Essentially, it's a drastically cheaper, competitively capable model that the firm is virtually giving away for free.

Bernstein analysts said that DeepSeek-R1, a reasoning model more comparable to OpenAI's o1 or o3, is even more concerning from a competitive standpoint. This model uses reasoning techniques to interrogate its own responses and thinking, similar to OpenAI's latest reasoning models.

R1 was built on top of V3, but the research paper released with the more advanced model doesn't include information about the hardware stack behind it. DeepSeek used strategies like generating its own training data to train R1, which requires more compute than using data scraped from the internet or generated by humans.

This technique is often referred to as "distillation" and is becoming standard practice, Ambati said.

Distillation brings with it another layer of controversy, though. A company using its own models to distill a smarter, smaller model is one thing. But the legality of using other company's models to distill new ones depends on licensing.

Still, DeepSeek's techniques are more iterative and likely to be taken up by the AI indsutry immediately.

For years, model developers and startups have focused on smaller models since their size makes them cheaper to build and operate. The thinking was that small models would serve specific tasks. But what DeepSeek and potentially OpenAI's o3 mini demonstrate is that small models can also be generalists.

It's not game over

A coalition of players including Oracle and OpenAI, with cooperation from the White House, announced Stargate, a $500 billion data center project in Texas — the latest in a quick procession of developments in large-scale conversion to accelerated computing. DeepSeek's surprise release has called that investment into question, and Nvidia, the largest beneficiary of the investment, is on a roller coaster as a result. The company's stock plummeted more than 13% Monday.

But Bernstein said the response is out of step with the reality.

"DeepSeek DID NOT 'build OpenAI for $5M'," Bernstein analysts wrote in a Monday investor note. The panic, especially on X, is blown out of proportion, the analysts said.

DeepSeek's own research paper on V3 says: "The aforementioned costs include only the official training of DeepSeek-V3, excluding the costs associated with prior research and ablation experiments on architectures, algorithms, or data." So the $5 million figure is only part of the equation.

"The models look fantastic but we don't think they are miracles," Bernstein continued. Last week China also announced a roughly $140 billion investment in data centers, in a sign that infrastructure is still needed despite DeepSeek's achievements.

The competition for model supremacy is fierce, and OpenAI's moat may indeed be in question. But demand for chips shows no signs of slowing, Bernstein said. Tech leaders are circling back to a centuries-old economic adage to explain the moment.

The Jevons paradox is the idea that innovation begets demand. As technology gets cheaper or more efficient, demand increases much faster than prices drop. That's what providers of computing power, such as Foundry's Jared Quincy Davis, have been espousing for years. This week, Bernstein and Microsoft CEO Satya Nadella picked up the mantle, too.

"Jevons paradox strikes again!" Nadella posted on X Monday morning. "As AI gets more efficient and accessible, we will see its use skyrocket, turning it into a commodity we just can't get enough of."