The Information AM

Save 25% on an annual subscription to read the most important news about technology and business first. For even more, Save $250 on The Information Pro for unlimited access to our proprietary org charts, databases and surveys.

Apple is preparing to release its Apple Intelligence on-device AI service in mainland China in December, ending a lengthy wait caused by regulatory hurdles, Bloomberg reported Thursday.

The planned release of Apple Intelligence in China, one of the iPhone-maker’s most important markets, comes two years after the company first launched the service in most other countries including the U.S. Apple has had to grapple with Chinese approval processes and mandated collaborations with local companies before it can roll out any AI features.

In July, China’s internet regulator greenlit Apple’s AI service, along with other similar on-device AI services from Chinese smartphone makers, under the country’s generative AI rules.

Apple Intelligence in China will come with guardrails the company developed with Alibaba Group, such as a filter that allows the Chinese government to approve updates to the AI models, according to Bloomberg.

Due to local regulations, the Chinese version of Apple Intelligence will also exclude the new Siri AI powered by Google’s Gemini models, as well as Apple’s in-house image-generation tools such as Image Playground.

Anthropic’s revenue in the third quarter hit $16.8 billion, the company on Wednesday told prospective investors in its initial public offering, according to people familiar with the matter.

That marked a jump of about 45% from the second quarter, and was more than 3.6 times the company’s revenue for all of 2025. But it also represents a slower growth rate than Anthropic’s second quarter, when fevered demand for its Claude Code product drove revenue up by more than 140% from the first three months of the year.

Anthropic also said it had an adjusted operating margin of 11% this year as of the latest quarter, a figure that excludes major costs such as stock compensation but suggests it is having some success in moving toward profitability.

The company held the meeting in San Francisco as part of preparation for an IPO expected to take place next month before the Thanksgiving holiday.

This brief has been corrected to clarify that Anthropic’s 11% adjusted operating margin was for the year through the third quarter, not just for quarter.

Sachin Katti, vice president of compute strategy at OpenAI, said Wednesday he was leaving the role on November 15 after about a year at OpenAI. Katti, who joined OpenAI after four years at Intel, announced that he would be “moving on to build something new” in a post on X.

During Katti’s time in the role, OpenAI released its custom Jalapeno inference chip, pushed a reorganization in its multi-gigawatt Stargate data center project, and signed a $20 billion deal to use Cerebras’s AI chips through the next three years. Katti also oversaw computing agreements with SB Energy and the expansion of OpenAI’s deal with Amazon Web Services. Katti will continue advising the compute unit after his departure for an undefined period of time, an OpenAI spokesperson told The Information.

OpenAI’s chief technology officer of compute Uday Ruddarraju, who was formerly the head of infrastructure engineering at xAI, is set to succeed Katti and will report to President Greg Brockman, the spokesperson said. The compute unit at OpenAI has seen several leadership changes over the past year, including three executives who helped launch Stargate departing around April.

Bloomberg first reported Katti’s departure.

Nvidia and ChipAgents, a startup that helps customers automate chip design, on Wednesday released a benchmark that measures how different AI models perform at the task.

Anthropic’s Fable 5.1 powered the best-performing AI agent in the newly released evaluation but only scored 16.6% on the test, ChipAgents said. According to CEO William Wang, ChipAgents made its benchmark more difficult than existing ones by covering the entire chip design process, not just a part of it, which is how other evaluations work.

The new evaluation comes as AI appears to be speeding up the field. It previously took at least two years to develop an AI chip, said Wang. Now, the latest models have helped bring down that timeline closer to one year by automating away a lot of the manual work (like combing through hundreds of pages of data to find simulation errors) necessary to verify that chip designs can work in practice before producing the physical chips. Wang thinks that in the coming year, AI advances will further compress the process to four to six months.

ChipAgents was founded in 2024 and has raised $134 million to date from B Capital, memory designer Micron and chip designer Mediatek. Wang said the company counts several semiconductor design companies with multibillion-dollar market caps as customers. He declined to name them or say whether Nvidia was an investor or customer.

Taiwan Semiconductor Manufacturing Co. said Thursday it will build another 3-nanometer chip factory in the Taiwanese city of Tainan, with production set for the first half of 2028. The move signals confidence that strong AI chip demand will last for years.

CEO C.C. Wei said strong customer demand prompted the decision to build the new 3nm factory in addition to other planned 3nm plants in Tainan, Arizona in the U.S. and Kumamoto, Japan. TSMC is also checking the infrastructure behind the skyrocketing AI chip orders, sending teams to the U.S. to assess data-center construction and power supplies. So far, Wei said, the feedback has been positive.

TSMC makes chips that other companies design, including Nvidia’s AI processors and Apple’s iPhone chips. That makes TSMC’s capacity plans a real-time gauge of how much AI hardware the industry can and will actually build.

The expansion announcement came as TSMC posted a sixth straight quarter of record profit. The company reported revenue in the third quarter rose 41.2% from a year ago to $46.72 billion, while net income jumped 68.7% to about $23.86 billion. TSMC now expects 2026 revenue to grow about 44% in U.S. dollar terms, up from its previous forecast of slightly above 40% in July.

Wei also said TSMC sent a team to Texas to assess building a new chipmaking facility there as its American customers seek more local production and the U.S. government supports expansion. The company said it is discussing using capacity tied to Elon Musk’s Terafab project, but has reached no agreement so far. TSMC said it would need to operate any facility it uses, but Musk has said Tesla and SpaceX would run Terafab themselves.

Morgan Stanley CEO Ted Pick said on Wednesday that he expects the AI buildout will encounter bottlenecks and periods of overbuilding, but investment will continue at a massive scale over the next several years.

Pick also said during Morgan Stanley’s third-quarter earnings call that the bank’s role is to regulate the flow of AI capital expenditure, slowing down when the market moves too quickly and speeding up as needed. “The quantums are so big, and the growth of the U.S. economy is at some level, and even the global economy, being powered by this AI CapEx cycle,” he told analysts.

Pick added that he foresees an annual run rate of “plus or minus a trillion dollars,” as the baseline for AI spending over the next five to 10 years.

JPMorgan CFO Jeremy Barnum on Tuesday said during the company’s third quarter earnings call that the bank, for its part, is carefully monitoring its AI exposure as the capex demands of the AI buildout continue to rise. Both executives’ comments arrive as investors have grown more cautious of AI-related deals in recent weeks.

OpenAI has accused a former employee of sharing the company’s unreleased device plans with senior leadership at Apple—a charge that echoes those Apple made against OpenAI in a lawsuit the iPhone maker filed in July.

In a filing Wednesday morning, OpenAI claimed the unnamed former employee joined OpenAI from Apple, worked for OpenAI for six months and then returned to Apple. The person took more than a hundred confidential files with him, including detailed images of the unreleased hardware and information about manufacturing and suppliers, OpenAI alleged in the suit.

OpenAI said that the person made a presentation about its product plans to Apple senior leadership, including a vice president of operations in Asia. “Now, armed with that information, Apple is making a new push into AI-enabled products,” OpenAI wrote. (Apple announced its HomeView smart hub screen device on Tuesday.)

“OpenAI is very concerned that its trade secret and confidential information is now circulating inside Apple, where Apple engineers can shape Apple’s upcoming products based on OpenAI’s confidential and innovative work,” OpenAI wrote in the Wednesday filing.

An Apple spokesperson said that the company has been unable to find any evidence supporting OpenAI’s allegations.

In Apple’s original lawsuit against OpenAI, it claimed the AI company made a “systematic effort” to steal Apple’s trade secrets in developing its AI product plans. OpenAI denied the charges.

Anduril Industries is in talks to allow employees to sell their shares in a tender that would value the autonomous munitions maker at about $90 billion, according to Bloomberg. Employees would be able to sell shares at $94.80 each in the tender, which would be capped at roughly $1.3 billion and is set to close in early November, the report said.

The share sale will help employees cash out of some of their stock-based compensation, relieving one pressure on the startup to go public. Anduril CEO Brian Schimpf said in July that the company is “not in a rush” to go public and that “a bad time to do that is in the middle of a hype cycle.”

Anduril last raised $5 billion at a $61 billion valuation in a May round co-led by Thrive Capital and Andreessen Horowitz. Founders Fund—another major Anduril investor whose partner Trae Stephens co-founded the company—invested about $624 million into that funding round, The Information reported in April.

Berkshire Hathaway has agreed to invest $330 million in Italian digital conglomerate Bending Spoons in a private placement of stock, providing a boost to investor confidence in Bending Spoons. Shares of the Italian company leapt 40% to $52.88 on Wednesday.

That share price is well above the $33 a share price that Berkshire is paying for 10 million shares, however, according to a securities filing by Bending Spoons. Berkshire had already invested $600 million in Bending Spoons’ July initial public offering, Bending Spoons said. The new investment will increase Berkshire’s stake in the company to 4%.

The fresh investment will help Bending Spoons, which has been pursuing a steady stream of acquisitions, mostly funded by debt. The company is buying up aging digital companies whose value has declined, with the aim of cutting costs and turning them around. Investor sentiment around the company has been wary, given Bending Spoons’ gr