Apple has reportedly boosted iPhone production in India as the country faces pressure from the White House over its Russian oil purchases.

“We have planned to up the tariffs on India,” Treasury Secretary Scott Bessent told CNBC’s “Squawk Box” on Tuesday. “These are secondary tariffs for buying the sanctioned Russian oil.”

Bessent accused India of “profiteering” by purchasing cheap Russian oil and reselling it during the Ukraine war, “which is unacceptable.”

Earlier this month, President Donald Trump hiked tariffs on India to 50%. The president said in July that he would impose secondary tariffs “at about 100%” on Russia’s trading partners if a peace deal isn’t reached with Ukraine by September.

His comments came as Bloomberg reported that technology giant Apple has reportedly upped production at five of its factories in India as it readies for the launch of its new iPhone 17 models.

The expansion includes some new plants and factories belonging to the Tata Group and contract electronics manufacturer Foxconn Technology, according to the report, citing people familiar with the matter. Apple is also looking to create a new iPhone 17e in India next year, according to the report.

In recent years, Apple has shifted more production to India as it looks to reduce its reliance on China, especially in the wake of recent trade tensions.

Data from Canalys in May estimated that iPhone shipments from India to the U.S. grew 76% in May as trade restrictions loomed.

At the same time, Apple has committed to investing over $600 billion in the U.S. over the next four years to improve American manufacturing production.

That includes a $100 billion spending expansion this month, which included a $2.5 billion investment to expand iPhone glass maker Corning’s production.

This commitment to building in the U.S. should put Apple in the clear on India-related tariffs.

Nvidia said Tuesday that it is evaluating several products following a report that the company is working on a new artificial intelligence chip for China that is more powerful than the currently available H20.

The new product, tentatively called the B30A, is expected to be based on Nvidia’s Blackwell chip architecture, Reuters reported, citing people familiar with the company’s plans. Nvidia hopes to deliver sample units to Chinese clients for testing as soon as next month, according to Reuters.

“We evaluate a variety of products for our roadmap, so that we can be prepared to compete to the extent that governments allow,” the company told CNBC in a statement. “Everything we offer is with the full approval of the applicable authorities and designed solely for beneficial commercial use.”

Commerce Secretary Howard Lutnick on Tuesday praised Nvidia CEO Jensen Huang and said he wouldn’t be surprised if Huang wants to sell a new chip to China.

“I’m sure he’s pitching the president all the time,” Lutnick said in an interview with CNBC’s “Squawk on the Street.” “I’ve listened to him pitch the president, and the president listens to our great technology companies, and he’ll decide how he wants to play it.”

Earlier this month, Nvidia and rival Advanced Micro Devices agreed to give the U.S. government a 15% cut of their sales in China, in exchange for permission to resume selling chips in the region.

The Trump administration halted the sale of advanced computer chips to China in April over national security concerns. Nvidia built its China-specific H20 chip after the Biden administration implemented export controls on AI chips in 2023, while AMD developed the MI308 chip for the Chinese market.

President Donald Trump said last week that he originally requested a 20% cut of Nvidia’s sales, but that Huang negotiated the number down to 15%.

Trump suggested at the time that he would be open to Nvidia selling a significantly scaled-down version of its advanced Blackwell chip in China.

“It’s possible I’d make a deal” on a “somewhat enhanced — in a negative way — Blackwell” processor, Trump said. “In other words, take 30% to 50% off of it.”

— CNBC’s Kristina Partsinevelos contributed to this report.

Jeff Wang got a big promotion last month. There were lots of tears, but not the happy kind.

The 39-year-old was unexpectedly named interim CEO of artificial intelligence coding startup Windsurf. The company had been in discussions with OpenAI about a potential acquisition that would have resulted in a handsome payday for many employees. But the talks fell apart and, on July 11, several founders and top researchers instead left to join Google as part of a $2.4 billion licensing deal.  

As one of the highest-ranking executives remaining at Windsurf, Wang was elevated to the top job, at least for the time being. His first order of business, he told CNBC, was to break the news at a tense all-hands meeting at the startup’s Silicon Valley headquarters.

“It was a very, very challenging day,” Wang said. “People were crying. It was very, very emotional. I was spending half the time calming down people, because they have families and they got nothing.”

Windsurf is part of a growing crop of AI startups whose founders and top researchers have been poached by megacaps like Meta, Google, Microsoft and Amazon through high-priced talent grabs that are helping the biggest companies skirt regulatory scrutiny. While the deals often produce big payouts for founders and AI leaders, they can leave investors, other employees and the remaining company in limbo.

Samir Kumar, a general partner at Touring Capital, said that what’s left is something resembling a zombie company.

“There’s a big question of what their future prospects are,” Kumar said. “Frankly, you hollowed out the organization.” 

The headline-grabbing deal came in June, when Meta rocked the tech industry by announcing a $14.3 billion investment in data labeling startup Scale AI. As part of the agreement, Meta took a 49% stake in the company, hired its CEO Alexandr Wang to lead a new superintelligence lab and said it would deepen the work it does with Scale.

A month later, Scale cut 200 full-time employees, or 14% of its staff. Meta’s investment had doubled Scale’s valuation from $14 billion last year. But that number only exists on paper.

Microsoft used a similar playbook in March 2024, when it hired Inflection AI’s co-founders and other staffers. Amazon has completed two such deals in the last year, nabbing the founders and top talent away from Adept in June 2024, and from Covariant two months later. Google inked a $2.7 billion licensing deal with Character.AI and hired its founders last August.

For Silicon Valley venture investors, long the lifeblood of risky tech startups, the system isn’t functioning as intended. Companies that would otherwise be on the path to a potential initial public offering or lucrative acquisition are getting pulled apart, with the bulk of the cash ending up in the pockets of the founders and their leading engineers.

“The money doesn’t flow as straightforwardly as it would in just a pure M&A transaction,” said Rob Toews, partner at Radical Ventures. 

The Scale deal was the exception as far as venture returns go, as Meta’s hefty cash investment resulted in big gains for early investors, most notably Accel.

Scale spokesperson Joe Osborne said the company is very much alive.

“Comparing Scale to these companies overlooks major differences in our revenue performance, company size, and deal structure,” Osborne said in a statement. He said Scale has more than 1,000 employees and over $100 million in revenue with enterprise and government clients.

“Meta’s investment benefited our investors and employees, kept us independent, and positioned us for long-term success,” Osborne said.

VCs say one way to try and keep founders and early employees from bailing is through secondary offerings, allowing them to sell a piece of their ownership to investors so they can buy a house, repay debt or just lock in some gains.

But secondary sales can’t compete with the kinds of offers coming from deep-pocketed tech companies that are flinging open their wallets to win the AI battle. Tech investors and startup employees who spoke to CNBC said it’s a trend that threatens to thwart innovation as founders abandon their ambitious projects to work for the biggest companies in the world.

“This is not business as usual,” said Tom Chavez, co-founder of the startup studio Superset. “This is a disruption.”

The Federal Trade Commission, then led by Lina Khan, was seeking to block a number of notable tech acquisitions, including Microsoft’s $69 billion agreement to buy Activision Blizzard (a deal that eventually closed) and Nvidia’s $40 billion bid for Arm (a deal that collapsed). It even tried, unsuccessfully, to stop Meta from completing a much smaller deal — the $400 million purchase of virtual reality studio Within.

Meanwhile, regulators in Europe were able to block Amazon’s planned $1.34 billion acquisition of iRobot last year, shortly after convincing Adobe to walk away from a $20 billion purchase of Figma. In 2022, the U.K.’s Competition and Markets Authority pressured Meta into selling Giphy, which it had bought for $400 million.

The transition from the Biden administration and Khan’s FTC to a second Trump presidency has led to some regulatory loosening, but antitrust concerns remain.

Tech companies, convinced of their need to bulk up in AI, have found a workaround.

“This is now a new playbook that companies are going to run,” said Matt Murphy, a partner at Menlo Ventures. “If it’s not cracked down upon, I don’t really blame them.”

Murphy added that companies are getting more explicit in saying, “Let’s just buy exactly what we want and leave the rest behind.” He described the process as “a bit soulless.”

Maintaining a minority stake means the acquirer could avoid triggering an FTC rule that would require a premerger review.

“They’re coming as close as possible to just getting under a majority stake of a company,” said J.B. Branch, an advocate for consumer rights at nonprofit Public Citizen. “They’re doing just about everything they can do without sort of tripping any alarms.”

Regulators aren’t completely in the dark.

The FTC last year opened probes into Microsoft’s Inflection deal and Amazon’s hiring of Adept employees. An FTC spokesperson declined to comment on the status of ongoing investigations or potential probes.

At Windsurf, the sudden departure of co-founders Varun Mohan and Douglas Chen left employees reeling. Wang was among them.

Many staffers at the all-hands meeting had been expecting to hear an announcement about a deal with OpenAI, and some even started filming the session, Wang said. The companies had been in talks about a potential acquisition since the spring.

Those discussions fell apart after Windsurf began pushing for a deal that looked less like a conventional acquisition and more like Meta’s investment in Scale, according to a person familiar with the talks. OpenAI wasn’t interested, said the person, who asked not to be named because the talks were confidential.

OpenAI’s exclusivity period for the potential acquisition expired on May 1, according to another person familiar with the discussions who also asked not to be named because the talks were confidential. Mounting antitrust concerns over Microsoft’s role in the deal, and debates over whether the tech giant would have access to Windsurf’s intellectual property, also stalled the talks, the person said.

Microsoft is a major investor in OpenAI and has pumped billions of dollars into the startup.

Wang said he wasn’t told why negotiations with OpenAI had ended.

The Google news broke on a Friday. Wang’s life changed in a hurry. He said he spent roughly 12 straight hours on the phone with investors, potential buyers and panicked customers. Employees were also barraged with incoming calls from recruiters, he said.  

At around 5 p.m., Wang received emails and text messages from the team at Cognition, another AI coding startup that’s best known for its software-building agent called Devin. After a 9 p.m. phone call stretched past midnight, Windsurf’s future began to take shape. 

The following Monday, Cognition announced it was buying Windsurf’s intellectual property, product, trademark, brand and talent.

Terms weren’t disclosed, but according to TechCrunch, the price was $250 million, or less than 10% the amount that OpenAI was reportedly in talks to pay.

“There’s definitely more motivation now to win than before,” Wang said. “You just got abandoned, now it’s time to prove to the world that you’re still here.”

Three weeks after Cognition’s announced purchase, that company’s CEO, Scott Wu, had more news to share. His startup was offering buyouts to Windsurf staffers who weren’t sold on their new employer.

“Cognition has an extreme performance culture, and we’re upfront about this in hiring so there are no surprises later,” Wu wrote in a post on X.

Windsurf employees who chose to accept the buyout would receive an additional nine months of pay and health care, Wang said. Some underperforming employees were let go, but would receive the same benefits, he added. 

“You cannot just have employees that came to try to ride a rocket ship,” Wang said. “You kind of need everybody to be all in.”

For now, Windsurf will continue to exist separately within Cognition, but Wang said he doesn’t know what his job will be in the future as the teams integrate more deeply.

“We’re both CEOs of each entity,” he said. “I’m actually not sure when the entities combine, what the role is going to be.”

That sort of uncertainty isn’t unique to Windsurf.

When Google inked its deal with Character.AI, the search giant’s big win was its hiring of co-founders Noam Shazeer and Daniel De Freitas, who had worked at Google in the past.

At Character, Shazeer and De Freitas wanted to do two things at once: build a foundational AI lab and an AI consumer company. A former Character employee, who was at the company when the deal took place, said he thought the agreement could be beneficial to the startup, because it would allow the company to focus solely on the consumer business.  

But many Character employees were drawn to the company because of the founders. Within a month of their exit, up to 10% of the remaining workforce departed, said the former employee, who asked not to be named because the details were confidential.

Dominic Perella, Character’s legal chief who served as the company’s interim CEO after the Google deal, said his company is far from gutted. Besides Shazeer and De Freitas, only a couple dozen researchers left, or about 25% of the company, with 70 employees still remaining, Perella said.

Perella added that the entire product team is still on board, as the company focuses on shipping new features to users rather than training AI models. Consumers use the app to create virtual characters, including for companionship, paying monthly subscriptions along the way.

Character said it now has about 20 million monthly active users, and that, as of June, paid subscribers were up 250% from a year earlier.

“That’s quite different from what you think of when you think of an acquihire,” Perella said.

Perella also said the company set up an escrow account with some of the cash from Google, and will distribute those funds to employees through July of next year, giving them some liquidity from the deal. And, he said, investors and staffers still have the opportunity to profit.

“The company didn’t leave investors or employees high and dry,” he said.

In June, 10 months after the deal was announced, Character named former Meta executive Karandeep Anand as CEO. Anand had been serving as an advisor to the company since the prior summer.

Anand said he views his role as helping Character become the social consumer app of the AI era, comparing it to what Instagram did for mobile and YouTube for the web.

“A founder CEO versus someone who comes in, who takes care of the company – that’s my job,” Anand told CNBC in an interview. “To prove that I can do this better than anyone else.”

A Google spokesperson said in a statement that the company doesn’t agree with the characterization that Character or Windsurf are zombie companies.

Inflection employees learned about their company’s deal with Microsoft at a last-minute all-hands meeting in March 2024. Inflection co-founder and CEO Mustafa Suleyman sent a Slack message instructing staffers to gather at a hotel in Mountain View, California, early the next morning, according to a former employee. 

That’s where Suleyman, who previously co-founded the AI research lab DeepMind, announced the transaction. Microsoft CEO Satya Nadella also made an appearance to answer employees’ questions, said the person, who asked not to be named due to the private nature of the discussions. 

Most of Inflection’s existing employees moved to Microsoft, the person said. But Inflection continued to operate as a stand-alone company, bringing on a new CEO and staff.

Microsoft reportedly paid Inflection around $650 million, according to reports from The Information and Reuters. That capital was used to help pay the startup’s investors and license its technology in a nonexclusive capacity.

The prior year, Inflection was valued at $4 billion in a $1.3 billion investment that included capital from Microsoft and Inflection co-founder Reid Hoffman, a Microsoft board member.

Still, Hoffman said after the sale to Microsoft that the deal was positive for “everyone involved in Inflection AI.” Hoffman is also a partner at venture capital firm Greylock, which was an early investor in Inflection. 

“This agreement with Microsoft means that all of Inflection’s investors will have a good outcome today, and I anticipate good future upside,” Hoffman wrote in a post on LinkedIn.

In March, Inflection named longtime tech executive Sean White as CEO. Since losing Suleyman and most of its staff to Microsoft, the company has hired 50 employees and is now focused on building enterprise AI products, according to its website, which says its first offering called Pi is an “empathetic and conversational chatbot.”

Microsoft and Inflection declined to comment.

Covariant, founded in Berkeley, California, developed AI systems for warehouse robots. Launched in 2020, its backers included leading AI researcher Fei-Fei Li, Meta chief AI scientist, Yann LeCun, and AI pioneer Geoffrey Hinton.    

The startup had also signed up roughly a dozen customers eager to use its software, including a sportswear giant and one of the leading meal kit companies, according to a person familiar with the matter who asked not to be named in order to discuss private information.

On a Friday afternoon last August, everything changed. Some of Covariant’s employees were asked to join an emergency town hall meeting where executives tearfully announced that three of the company’s co-founders and roughly 25% of its approximately 120 workers were joining Amazon. As part of the agreement, Amazon bought a nonexclusive license for Covariant’s core technology.

Everyone on the call was told they would not be moving to Amazon, according to three former employees, who requested anonymity to speak freely about the matter.

Those who were joining included co-founder Pieter Abbeel, who was tapped to help lead Amazon’s recently created advanced AI lab in San Francisco. The other two co-founders, Peter Chen and Rocky Duan, would be put to work at Amazon building foundation models for the company’s more than 1 million robots in its warehouses.

Remaining staffers were offered far less attractive options. They could either leave Covariant immediately or stay on for about a month and a half, at which point they’d be able to receive a severance payment, the former employees said. 

A “skeleton crew” of about 10% to 15% of Covariant’s workforce is still there, one ex-employee said, calling it a ghost company. Covariant hasn’t published any business updates to its site since the Amazon deal was announced last year.

While Covariant’s founders walked away with substantial payouts, rank-and-file employees were left with far less than they expected to earn when they joined, the former employees added.

In a whistleblower complaint filed in January with the FTC, Department of Justice and Securities and Exchange Commission, a former Covariant employee claimed that the transaction was “deliberately and unlawfully structured” by Amazon to dodge antitrust scrutiny.

The company had reportedly been valued at $625 million in a 2023 funding round. Amazon spent more than $400 million to license Covariant’s technology and bring on senior technical talent at the company, according to the whistleblower complaint, which was viewed by CNBC. That figure includes a final $20 million licensing payment, set to be paid out one year after the deal date, which requires “zombie Covariant staying on life support” until that point, the complaint said.

The transaction terms handicapped Covariant’s ability to seek out future deals, according to the complaint, which was previously reported on by The Washington Post.

Ted Stinson, Covariant’s current CEO and operating chief at the time, said the company had a “decent shot” at landing one or two licensing deals with its remaining technology, according to a transcript of a recorded conversation cited in the complaint.

“But they’ll be single-digit millions, maybe double-digit millions,” Stinson said. “I mean they’ll be a fraction of what Amazon paid is my best guess.”

Stinson didn’t respond to a request for comment.

John Tye, an attorney for the former employee and a former State Department whistleblower, said the Covariant deal and other transactions like it deserve stronger government scrutiny.

“It’s not just a technical, legal matter,” Tye said in an interview. “It affects consumers who use these products. Monopolies are not typically good for the American public.”

Similar to Covariant, San Francisco startup Adept has provided few updates on its work building agentic software since its co-founders and much of its research talent were hired away by Amazon. Less than a year after taking over as CEO of Adept, Zach Brock left for OpenAI, and several employees, including the former product head, have departed for competitors like Anthropic.

Covariant and Adept didn’t respond to CNBC’s requests for comment.

Amazon spokesperson Alexandra Miller said Covariant and Adept continue to serve customers independently. She said that because Amazon’s licensing deal with Covariant isn’t exclusive, “Covariant is free to license its technology to other companies.”

— CNBC’s Jordan Novet contributed to this report.

WATCH: Meta approached Perplexity before massive Scale AI deal