|
|
|
|
Nov 18, 2026
|
|
|
|
|
Supported by
|
|
|
|
|
|
Happy Wednesday! Nvidia's board says it will not reinstate Jensen Huang as CEO, as the company raises revenue forecast for 2027. An OpenAI director says there's a 'reasonable' chance of an AI disaster within a year.
|
|
|
Jensen Huang resigned from his roles as Nvidia president, CEO and board member after “disclosure of personal behavior inconsistent with the company’s code of conduct, standards and cultural norms,” Harvey Jones, Nvidia’s board chair, said Tuesday during the company’s quarterly earnings call. Jones said the board unanimously accepted Huang’s resignation and does not intend to reinstate him as CEO. Nvidia also hired search firm Heidrick & Struggles to find a new CEO, he said. Nvidia’s board investigation into Huang’s behavior—including his romantic relationship with a woman who has links to firms that smuggled banned Nvidia chips into China—is set to continue even without Huang at the company’s helm, The Information previously reported.
|
|
|
Nvidia Interim CEO Colette Kress said Tuesday that the chip designer’s revenue would rise 75% in the 12 months ending Jan. 31 2028, compared to the prior fiscal year, or 5 percentage points higher than the forecast she gave three months ago. While that figure represents a comedown from the company’s current growth rate—revenue more than doubled to $116 billion in the October quarter—it provides a measure of predictability for a company in crisis after the resignation last week of then-CEO Jensen Huang. However, Nvidia also said it wouldn’t file its quarterly securities filing with the Securities and Exchange Commission for weeks. Such filings, which usually become public around the time of the quarterly earnings discslosure, break out more details about the company’s business, which is now under scrutiny by U.S. authorities and lawyers hired by the company’s board. Huang stepped down after allegations of his extramarital affairs with a subordinate and with a woman linked to Chinese firms that imported banned Nvidia chips into China. Its shares are down about 10% since then and are down nearly 30% over the past month, as investors have gotten skittish about the financial health of some of Nvidia’s customers. Shares rose 4.5% in after-hours trading Tuesday. The company said revenue would rise about 91% in the current fiscal quarter, though Kress said it included an extra week compared to the October quarter. Nvidia has continued to increase its commitments to secure supply chain capacity—from $279 billion last quarter to $390 billion in the October quarter, mostly related to memory capacity. Gross margins rose 1.5 percentage points from a year ago to 74.8% in the October quarter. In other positive news, free cash flow nearly doubled to $43.8 billion, despite the fact that the company’s accounts receivables roughly tripled year over year. The company previously said the accounts receivable increase was “due to extended payment terms on large, multi-quarter agreements with certain investment-grade customers,” a term that typically refers to major cloud providers and large customers such as Meta Platforms and SpaceX. In another encouraging sign, the company said customers were able to get its latest flagship chip system, Vera Rubin, up and running in half the time it took them to use the prior flagship, Grace Blackwell. The Blackwell issues troubled Meta, OpenAI and many other key customers. Notably, Nvidia said it sold $2.4 billion of its older-generation Hopper AI chips to customers in China, though it said it would no longer do so, without elaborating. Such Hoper chips are approved for sale to that country, but Nvidia is coming under scrutiny because its banned Blackwell chips have also ended up in China in droves, according to The Information reporting throughout the year.
|
|
|
Paul Christiano, a prominent AI safety researcher and member of the board that controls OpenAI, offered an alarming assessment of where the technology is heading in an interview with a popular technology podcast, saying the likelihood of humans completely losing control of AI next year is rapidly increasing. “There’s a reasonable chance right now that we’re going to be in a world where in six months there’s a disaster that kills a very large numbers of people or the government is overthrown,” Christiano told podcaster Dwarkesh Patel in a nearly two-hour interview that aired Monday. Patel’s show often draws millions of listeners including many in the tech world. Christiano said whether that happens will depend on the speed at which the technology continues to improve, adding that he thinks there is a “reasonably high” chance the technology will reach that point over the next year because the models are improving themselves using a technique called recursive self-improvement. The comments were an escalation of previous statements from Christiano, who leads the Alignment Research Center, a Berkeley-based nonprofit focused on AI alignment. A veteran AI researcher who has long warned about safety concerns related to the technology, he joined the board of the OpenAI Foundation, which controls the company, in September. Christiano emphasized in the interview that he was speaking in a personal capacity “and not as an OpenAI board member.” When he initially joined the board, Christiano estimated the risk of disaster stemming from humans losing control of AI to be around 4% over the next year, he wrote in a blog post at the time. In the Monday interview, he said models have come along so much in the past few months that he now thinks the likelihood of an AI takeover is “looking more like a 10% risk than a 1% risk over the next year.” He said his estimate would be “significantly higher” if companies don’t start taking more precautions to slow the risk of AI development.
|
|
|
Salesforce on Monday told employees they will have a cap on AI spending that will vary depending on their role, ranging between $100 to $1,000 a month, an employee of the software firm said. The email Salesforce sent to staffers about the change said the caps were a response to some employees spending thousands of dollars per month on AI tools, according to the person. The Information previously reported that Salesforce in recent months had worked to implement team-level AI spending caps, though some employees were not aware of those caps. Details of the new Salesforce spending caps were first reported by Business Insider.
|
|
|
Microsoft debuted a suite of new AI models at its annual Ignite conference in San Francisco on Tuesday and said it will license the models’ weights—the settings that determine how the models answer questions—so customers can create highly customized versions for their businesses. The move differentiates Microsoft’s business model from those of Anthropic and OpenAI, neither of which let customers access model weights. Other firms like Thinking Machines Labs and Meta have previously let customers customize AI models, known as open-weight models. Microsoft also unveiled a suite of new features in its Copilot AI software and Azure cloud service aimed at giving enterprises more options to securely use AI agents while undercutting rival AI providers on cost. “Everything you can do in Claude or ChatGPT, you can do better in Copilot for less,” Microsoft’s commercial chief Judson Althoff said at the event. Microsoft’s new models focus on code-writing, generating images, transcribing audio and rapidly responding to messages, all of which CEO Satya Nadella said would give customers similar performance to the most advanced AI models at a fraction of the cost. Microsoft also announced new features in its Copilot AI software, including software that lets workers collaborate with each other’s AI agents within Office applications, as well as a new feature to keep tabs on what AI agents are doing and control the data they can access. New updates to Microsoft’s Azure cloud servers let companies host their AI agents in isolated servers that aren’t connected to the internet so they are more secure and won’t go rogue. The Information first reported on some details of the new Copilot and Azure features earlier this week.
|
|
|
Walt Disney Co. reported that operating income for its streaming operations more than doubled in the quarter ended Oct. 3, compared with a year earlier, reflecting growth in the number of subscribers as well as the quarter including an extra week compared to a year earlier. But this is the last quarter Disney will report streaming operating profit. The company is changing its disclosures and will only report profit from Disney’s overall entertainment operations, which encompasses its ABC broadcast network, cable channels, film studio and various smaller operations include its Broadway theaters as well as streaming. In the most recent quarter, Disney entertainment’s revenues rose 17% to $11.9 billion, while operating income rose 92% to $1.3 billion. Streaming contributed about half of the revenue, mostly from subscriptions. The new disclosure strategy follows Disney’s decision a year ago to stop reporting streaming subscriber numbers.
|
|
|
Jensen Huang said Tuesday he was willing to return as Nvidia’s CEO if its board decides his return is in the best interest of Nvidia’s shareholders, employees, the company and America. “Since my departure, I’ve seen an outpouring of calls from Nvidia’s employees, investors, business partners, and customers for me to return as CEO,” Huang said in a statement, his first since resigning as CEO last week. An online petition calling for Huang’s return now has nearly 200,000 signatories, including approximately 10,000 employees. Huang reaffirmed that he is cooperating with the board’s independent investigation into his relationship with a suspected Chinese spy and whether national security or Nvidia trade secrets were compromised, which it initiated on November 8 when Huang admitted he had been unfaithful to his wife. “I did not live up to the standards expected of everyone at Nvidia, and I have made serious mistakes in my personal life,” Huang said. “I created problems for the Board that I deeply regret.” On Monday, a person directly familiar with the board’s thinking said the board was not interested in bringing Huang back as CEO or in any other role. It’s unclear if the board’s position has changed since then.
|
|
|
Anthropic’s advisers have been guiding some investors that it could set the valuation of its upcoming IPO at around $1.5 trillion, implying that it could price over the expected range or increase the number of shares it sells. The maker of Claude has received at least six times more demand for its upcoming share sale than available shares. It planned to sell 825 million shares at a price range between $70 to $78. If Anthropic prices its IPO at a fully diluted valuation of around $1.5 trillion, the offering price could fall between $83 to $85 or it can choose to sell more shares. The discussions are fluid and the details could change. Anthropic will stop taking orders at 4 p.m. Eastern Tuesday, and it expects to price its shares late Wednesday. Several of its investors have committed to buying in the IPO. Brad Gerstner’s Altimeter Capital is aiming to invest $5 billion in the IPO, he told CNBC. Business partners including Nvidia, AMD, Broadcom and Google are investing $26 billion in a private placement.
|
|
|
Google has announced updates to Gemini, called Tasks, that will give it more agentic features, comparable to tools already released by Meta with Muse and OpenAI with Dots. The Task features will be free. Gemini will now be able to call businesses, make purchases, and read and send emails, according to a blog post Google published on Tuesday. Users will also be able to text or email Gemini to initiate Tasks. In May, Google rolled out a personal AI agent called Spark, but the company quietly rolled the tool into the Gemini app in October, renaming the feature Tasks. Google has lagged other companies in releasing a personal agent for consumers, which is especially notable given the ubiquity of its consumer ecosystem. Within weeks of its launch, Meta’s Muse agent topped more than 3 million users giving the tool at least one prompt a week according to The Information. OpenAI quickly followed with its own agent, Dots. In a blog post, Google said its access to passwords and payment information made Gemini the “best choice” as a personal agent.
|
|
|
Texas environmental regulators have approved an air pollution permit for Elon Musk’s SpaceX’s 1.456 gigawatt gas-fired power plant for a supercomputer planned for the Rio Grande Valley, according to a news alert from the power regulatory service Halcyon and reports this afternoon from Politico and the Texas Tribune. The high-profile move indicates that outgoing Gov. Abbott’s administration may be speeding along AI data center projects that could otherwise be held up by the new Democratic Gov.-Elect Gina Hinojosa, who takes office January 19. Texas has become the most active spot for new data center construction in the country. Investors, developers and tech companies have hundreds of billions of dollars in investments awaiting resolution at a time when the new governor elect had vowed to pause approvals of not-yet-permitted projects until the Texas legislature could pass new safeguards. Previously, many developers believed they would have to wait until at least December to get clarity on Texas projects. Gov. Abbott had paused approvals for data center grid connections in August and mandated that the Electric Reliability Council of Texas, or ERCOT, gather information on potential community impacts from data centers. Abbott extended that pause to all environmental permits for data centers, whether grid connected or not, until ERCOT completed its review and issued final findings on Dec. 10. This meant data centers building their own power plants were held up until they received air pollution permits. SpaceX had filed for an air permit in September that was on hold. The Texas Tribune reported earlier today that the Texas Commission on Environmental Quality was resuming environmental permits for data centers. According to the Tribune, Abbott’s office notified the Texas Commission on Environmental Quality last Friday that ERCOT had gathered what it required for its audit. According to Politico, the state environmental agency said this meant it could resume its normal permit review process. Environmental groups had filed a complaint with the environmental agency just three days ago, Politico said, alleging SpaceX had begun construction activities at the site without a permit in violation of the Clean Air Act.
|
|
|
|
This week on AI Deep Dive
Check out the most recent episode of AI Deep Dive. A reported deep dive into how frontier AI actually gets built — the models, the money, and the people behind them.
| |
Popular articles
By Phoebe Liu, Leo Schwartz and Michael Roddan
|
|
|
|
Opportunities
Empower your teams to stay ahead of market trends with the most trusted tech journalism.
Learn more
Reach The Information's influential audience with your message.
Connect with our team
|
|
|
|
|