Nvidia (NVDA) has reportedly halted production of China-bound H200 chips, the Financial Times reported, as regulatory efforts in Washington and Beijing have restricted imports to the Chinese market.

News that Nvidia would shift its TSMC output to focus on the Vera Rubin platform came after CEO Jensen Huang suggested on Wednesday that the artificial intelligence leader's multibillion-dollar investments in AI startups OpenAI (OPAI.PVT) and Anthropic (ANTH.PVT) could be its last for a while, as he anticipates those startups going public.

At an industry conference on Wednesday, Huang hinted that the investments were finalized, saying Nvidia would invest $30 billion in OpenAI and that investing the full $100 billion was "probably not in the cards."

Huang's statement on the company's investments comes as investors continue to assess the fallout from OpenAI's partnership with the Pentagon, which emerged just as Anthropic's deal with the Defense Department fell apart.

However, on Thursday, the Financial Times reported that Anthropic CEO Dario Amodei has resumed negotiations with the Defense Department to put a military contract back on the table.

Earlier this week, tech watchers were also treated to a slew of product announcements from Apple (AAPL). The product release list included a new low-cost MacBook Neo computer, an entry-level iPhone 17e smartphone, two new iPad Airs, and new MacBook Air and MacBook Pro laptops with more powerful M5 chips.

Follow along for the latest updates on the tech sector.

Nvidia (NVDA) has reportedly stopped production of its H200 chips intended for China, as it no longer expects significant sales there.

According to the Financial Times, Nvidia has asked Taiwan Semiconductor Manufacturing Company to reallocate manufacturing capacity from making H200 chips to making next-generation Vera Rubin hardware.

In its latest call, Nvidia executives said they had not sold any H200 chips to Chinese customers yet. Although President Trump indicated Nvidia would be able to sell its second-most-powerful chip to China in January, the move suggests that the company expects additional regulatory barriers in Beijing and Washington to impede sales.

"While small amounts of H200 products for China-based customers were approved by the US government, we have yet to generate any revenue, and we do not know whether any imports will be allowed into China," Nvidia CFO Colette Kress said on the earnings call.

Nvidia had lobbied the US government to be able to sell the H200, which was marketed as being compliant with export controls. The FT previously reported that Nvidia had expected Chinese clients to order more than 1 million units and had been ramping up supply accordingly.

Read the full story from the Financial Times.

Anthropic (ANTH.PVT) is back at the negotiating table with the Pentagon, the Financial Times reported, after the deal between the artificial intelligence startup and the Department of Defense fell apart in a dramatic fashion last week, with promises on both sides to cut ties completely.

The Financial Times reports:

Read more here (premium).

Prediction market platforms are facing increased scrutiny and regulatory proposals after bets placed ahead of the US strikes on Iran drew suspicion of insider trading.

On Thursday, Democratic Sens. Jeff Merkley and Amy Klobuchar introduced a bill to prevent public officials from trading event contracts, as prediction markets like Polymarket and Kalshi have seen significant growth alongside reports of misconduct. (Disclosure: Yahoo Finance has a partnership with Polymarket.)

At the same time, the US Commodity Futures ‌Trading Commission has kicked off the process to clarify event contract trading rules by sending a rulemaking plan to the Trump administration's budget office.

Two recent incidents in particular raised suspicion over trading activity: the January arrest of Venezuelan leader Nicolás Maduro and the recent attacks on Iran. In the case of Iran, one Polymarket account with the username Magamyman, established in October 2024, made over $500,000 on a bet that the US would strike Iran by Feb. 28.

Polymarket defended its Middle East markets in a statement on its website that read, in part, "The promise of prediction markets is to harness the wisdom of the crowd to create accurate, unbiased forecasts for the most important events to society. That ability is particularly invaluable in gut-wrenching times like today."

However, that value seemed to hit its limit earlier this week when Polymarket took down its market that allowed users to bet on whether a nuclear weapon would detonate this year. The market had drawn more than $650,000 in trading volume before it was archived after facing criticism.

The Trump administration is considering requiring global companies, and in some cases their home countries, to seek US approval before purchasing AI chips from the likes of Nvidia (NVDA) and AMD (AMD).

According to Bloomberg, the permitting process, which is still under deliberation, would depend on how many chips a company is looking to buy. Those aiming to buy 1,000 chips would face a relatively easy review, while those seeking larger shipments may have to agree to site visits by US representatives.

Companies seeking to purchase several hundred thousand GPUs would need to involve their home country's government. Those kinds of shipments could only be made to US allies and would require the companies to make security promises and invest in American AI.

AMD stock fell more than 2% on the news, while Nvidia stock dropped more than 1%.

The rules, if adopted, would be the US's latest step toward controlling the export of its prized AI chips. Nvidia and AMD's processors are among the most sought-after processors in the world, thanks to their high-performance capabilities in both training and running AI models.

The Trump and, before it, Biden administrations have been working to find a way of getting America's chips into the hands of global companies, while keeping them from ending up in rival nations' data centers. But it's been a difficult task.

The White House's repeated back-and-forth with China over the use of Nvidia's chips has pushed that country to develop its own AI chips in the hopes that it can outmaneuver the US's grip on the high-end processor market.

Chinese automaker BYD (BYDDY) revealed its second-generation Blade Battery, which it said will offer "disruptive" charging speeds in extremely ​cold environments.

Reuters explains:

BYD also launched the world's most powerful mass-produced supercharger, capable of 1500 kW output. These chargers will be part of the 20,000-station network mentioned above.

Read more from Reuters here.

Executives of seven major tech firms signed an agreement on Wednesday promising to shoulder the energy costs needed to build out artificial intelligence, President Trump said after a White House tech roundtable.

Chief executives at Google (GOOG), Microsoft (MSFT), Meta (META), Oracle (ORCL), xAI, OpenAI (OPAI.PVT), and Amazon (AMZN) signed the "ratepayer protection pledge," committing to paying for increased energy costs to power data centers and related infrastructure.

Concerns that growing AI demand could substantially raise consumer electricity bills have become a political problem for President Trump and Republicans ahead of the 2026 midterm elections, though energy affordability is a growing bipartisan issue.

According to a White House fact sheet on the pledge, tech companies will "build, bring, or buy the new generation resources and electricity needed to satisfy their new energy demands, paying the full cost of those resources whether by building, or buying from, new or otherwise additive power plants."

While the agreement could affect how tech companies' multiyear power purchase agreements are negotiated, it's unclear how the terms of the pledge will be enforced.

See the video below for Trump's comments on AI electricity use and a breakdown of the pledge from my colleague, Jennifer Schonberger.

Trade Desk (TTD) stock jumped 19% before the bell on Thursday after reports emerged from The Information that OpenAI (OPAI.PVT) has had early talks to partner with the platform, which offers automated services for advertisers to place ads on a large scale.

The Information reports:

Read more here.

The FT reports:

Read more here.

Chipmaker Broadcom (AVGO) reported its first quarter earnings after the bell on Wednesday, beating expectations on the top and bottom lines and offering a better-than-anticipated Q2 outlook.

But that wasn't enough to satisfy an already anxious Wall Street. Broadcom stock was largely flat on the news.

For the quarter, the company saw earnings per share (EPS) of $2.05 on revenue of $19.31 billion. That topped analysts' estimated EPS of $2.03 and revenue of $19.26 billion.

For the second quarter, Broadcom said it sees revenue of about $22 billion. Wall Street was expecting $20.5 billion.

The company also approved a new $10 billion share buyback.

"Broadcom achieved record first quarter revenue on continued strength in AI semiconductor solutions," CEO Hock Tan said in a statement. "Q1 AI revenue of $8.4 billion grew 106% year over year, above our forecast, driven by robust demand for custom AI accelerators and AI networking."

Tan added that AI revenue growth is accelerating and that the company anticipates Q2 segment revenue of $10.7 billion.

The AI trade has had a rocky start to the year with investors questioning Big Tech's massive investments in the technology. But at the same time, announcements from Anthropic (ANTH.PVT) and OpenAI (OPAI.PVT) have rocked software stocks, as analysts raise concerns about whether AI will crush existing industries ranging from software-as-a-service companies like Salesforce (CRM) to cybersecurity firms like CrowdStrike (CRWD).

Nvidia (NVDA) CEO Jensen Huang said that the AI kingmaker has finalized its agreements with OpenAI (OPAI.PVT) and Anthropic (ANTH.PVT), which he said will likely be the last investments in those startups for the foreseeable future.

"We're going to invest $30 billion in OpenAI," Huang said at the Morgan Stanley Technology, Media and Telecom Conference on Wednesday.

"I think the opportunity to invest $100 billion in OpenAI is probably not in the cards. And the reason for that is because they're going to go public. And so this might be the last time we'll have the opportunity to invest in a company like this. And our $10 billion investment in Anthropic, probably will be the last as well," he added.

Nvidia announced its tentative plans to invest upward of $100 billion in Sept. 2025, saying that the partnership would allow OpenAI to deploy at least 10 gigawatts of Nvidia systems, with the first gigawatt originally scheduled to go online in the second half of this year.

But Nvidia later clarified that the $100 billion announcement was a letter of intent and not set in stone.

The company is now joining SoftBank, which will invest $30 billion, and Amazon (AMZN), which will invest $50 billion, to fund OpenAI's growth.

The AI explosion has meant that GPUs (graphic processing units) have dominated data centers across the world. Hyperscalers, neoclouds, and everyone in between have been splashing billions to get their hands on the high-end chips to train and run their AI models.

That left the humble CPU (computer processing unit), which powers virtually everything else in data centers, and thus the applications and services you use every day, out in the cold.

But that's starting to change. Earlier this month, Meta (META) and Nvidia (NVDA) announced an expanded deal that will see Nvidia provide the social media company with the largest deployment of its Grace CPU-only servers to date.

And just last week, AMD (AMD) announced its own deal with Meta, which includes servers running the company’s Venice and next-generation Verano CPUs.

It might sound counterintuitive for CPUs to grab some of the spotlight amid the global AI buildout, but in a world where AI inference and agentic AI are becoming increasingly important, CPUs, it turns out, are primed to shine.

“It's not a zero-sum game," explained Dan McNamara, senior vice president and GM of compute and enterprise AI at AMD.

"CPUs are growing, but GPUs are not slowing down, because there's more and more workloads.”

Read the full story here.

Bank of America reinstated coverage of Tesla (TSLA) with a Buy rating and bullish outlook on the company’s robotaxi and autonomous efforts.

In a long and wide-ranging note, analyst Alexander Perry gave Tesla a $460 price target, based on the investment bank’s sum-of-the-parts analysis. Perry picked up Tesla and BofA’s auto coverage after longtime analyst John Murphy left the firm.

Tesla shares rose 3% in early trade.

“We view [Tesla] as the current leader in consumer autonomy,” Perry wrote. “We expect TSLA to quickly become a leader in robotaxi services, given its ability to scale more profitably than competitors.”

Perry predicts autonomous vehicles and robotaxis will usher in the next era of mobility and be the “most significant change agent” in the Auto 2.0 landscape, with Tesla at the forefront.

Read more here.

OpenAI (OPAI.PVT) is developing its own version of GitHub, the code-hosting platform owned by Microsoft (MSFT), The Information reported on Tuesday.

The project marks the latest example where OpenAI could compete directly with Microsoft, which has a large stake in the AI startup.

The project is still in its early stages and will likely take months to complete, ⁠according to The Information. Once completed, OpenAI is considering making the code platform available to customers for purchase.

The report stated that OpenAI engineers began working on the project after encountering increasing service disruptions that led to GitHub downtime in recent months.

Read more here.

Apple (AAPL) unveiled its MacBook Neo, the company's long-rumored foray into the low-cost computer market, at an event in New York on Wednesday.

Starting at $599, the Neo comes in four colors, including a splashy yellow-green Citrus, and is designed to appeal to consumers, schools, and businesses looking for a more affordable option than Apple's recently refreshed $1,099 MacBook Air.

The Neo features a 13-inch screen, a bit smaller than the standard MacBook Air's 13.6-inch display, and runs on Apple's A18 Pro processor, similar to the chip that powers the company's iPhone 16 Pro and 16 Pro Max.

The Neo also comes with 256GB of storage, with an option for up to 512GB for an extra $100. That upgrade also gets you Apple's Touch ID fingerprint sensor to unlock the laptop and make payments via Apple Pay.

The Neo should serve as a solid alternative for schools, especially those that typically provide students with Chromebooks and, in some cases, iPads.

"This is one of the most important announcements for Apple in the Mac product line and represents a shift in the history of the Mac. Apple has always positioned the MacBook as a premium computing product, with entry prices typically starting near or above $999," International Data Corporation VP of client devices Francisco Jeronimo said.

Read the full story here.

Neocloud company CoreWeave (CRWV) says it has entered into a multi-year agreement with Perplexity (PEAI.PVT), that will see the AI search company run its inference workloads on CoreWeave’s cloud hardware.

CoreWeave builds data centers that run Nvidia (NVDA) chips, which other companies then rent to power their AI services. Under the terms of the agreement, Perplexity will use CoreWeave’s Nvidia GB200 servers to run its AI offerings.

CoreWeave will, in turn, deploy Perplexity’s Enterprise Max software internally.

CoreWeave is closely linked to Nvidia, which has invested billions in the company and become its second-largest shareholder. That’s led to concerns about so-called circular investing, in which one company invests in another only for the second company to turn around and purchase goods from the first.

The AI industry has faced a number of similar moves across a number of companies, including with Nvidia rival AMD (AMD).

CoreWeave has also faced criticism over its massive spending plans and debt load as it continues its AI buildout. Shares of the company have fallen more than 19% over the last six months, despite climbing 85% over the last year.

Anthropic's Claude became the most downloaded free app on both Apple's App Store and Google's Play Store on Tuesday, following the company's showdown with the Department of Defense (DOD) last week.

Anthropic (ANTH.PVT) and the DOD have been sparring over the company's insistence that the agency not use its Claude models for the mass surveillance of Americans or to develop fully autonomous weapons. On Friday, President Trump ordered all of the federal government to phase out Anthropic's technology over the next six months labeling the company a supply chain threat.

That could force firms that work with the government and use Anthropic's models to stop using the company's technology, though Anthropic says that only applies to those firms' government work and not to their private businesses.

But the blowup has seemingly helped Anthropic's public relations, with the company saying that Monday was its largest single day for sign-ups ever.

Read the full story here.

Apple (AAPL) continued its March product rollout on Tuesday with the debut of its latest MacBook Air and MacBook Pro, alongside more powerful M5 Pro and M5 Max chips.

The MacBook Air, Apple's volume seller, now starts at $1,099, a $100 price jump over last year's model, and comes with the company's M5 processor and more storage, 512GB rather than 256GB.

In addition to the Air, Apple announced its M5 Pro and M5 Max processors. The company says the chips use what Apple calls its Fusion Architecture, which combines two dies into a single processor. Both the M5 Pro and M5 Max can be outfitted with 18-core CPUs that include 6 "super cores" and 12 new "performance cores."

Both processors slot into Apple's new MacBook Pro 14-inch and MacBook Pro 16-inch. Apple is leaning into the laptops' AI capabilities, saying that the MacBook Pro with the M5 Pro chip gets up to 6.9x faster LLM prompt processing than the M1 Pro, while the MacBook Pro with the M5 Max offers 8x faster AI image generation than the MacBook Pro with the M1 Max.

As with the Air, Apple is increasing the MacBook Pro's base storage from 512GB in the M4 Pro the 1TB for the M5 Pro. The M5 Max model now gets 2TB, compared to the M4 Max's 1TB.

The Pros will cost you, though. The base MacBook Pro 14-inch with a standard M5 chip starts at $1,699, up from $1,599 last year.

Jump to the M5 Pro, and you'll pay $2,199. The M5 Max version, meanwhile, starts at $3,599.

Opt for a MacBook Pro 16-inch with all the bells and whistles, and you'll end up paying $7,349.

Read the full story here.

Here's another reason OpenAI (OPAI.PVT) may have felt compelled to reassure its users about its Department of Defense contract.

Data from Sensor Tower, reported by TechCrunch, found that uninstalls of ChatGPT's mobile app jumped 295% in the US on Saturday, when news of the AI startup's deal with the Department of Defense emerged.

ChatGPT's typical daily uninstall rate for the past 30 days has been around 9%, Sensor Tower said.

Meanwhile, US downloads for Anthropic's (ANTH.PVT) Claude app rose 37% on Friday and 51% on Saturday after the OpenAI rival rejected a deal with the Pentagon and was labeled a "supply chain risk" by the government.

Read more here from TechCrunch.

From Bloomberg:

Read more here.

OpenAI (OPAI.PVT) CEO Sam Altman said the ChatGPT-maker is working with the US Department of Defense to amend the agreement the two parties reached over the weekend.

"We have been working with the DoW [Department of ‌War] to ⁠make some additions in our agreement to make our ⁠principles very clear," Altman wrote in a post on ​X late on Monday.

Altman outlined two additions to the contract intended to make explicit the startup's commitment not to surveil Americans.

The first addition says "the AI system shall not be intentionally used for domestic surveillance of U.S. persons and nationals," while the second states, "for the avoidance of doubt," that the DOD acknowledges that it is prohibited from deliberately tracking or monitoring US persons or nationals, including through commercially acquired information."

The contract modifications came after OpenAI received some backlash for striking a deal with the Pentagon, just as DOD's deal with Anthropic (ANTH.PVT) fell apart. Anthropic was also pushing for strong guardrails around using AI models for surveillance and autonomous weapons.

"We were genuinely trying to de-escalate things and avoid a much worse outcome, but I think it just looked opportunistic and sloppy," Altman admitted.