| |  | Mar 09, 2026 | | | | | | Welcome back! OpenAI's head of robotics resigns over the Pentagon deal. Microsoft, Google and Amazon keep selling Anthropic models to customers other than the Pentagon. SoftBank seeks a $40 billion loan to help fund new OpenAI investment.
| | | | OpenAI's robotics chief said Saturday that she was leaving OpenAI over its negotiations with the Department of Defense. "Surveillance of Americans without judicial oversight and lethal autonomy without human authorization are lines that deserved more deliberation than they got," Caitlin Kalinowski wrote in a X post. Last week, OpenAI said it would strike a new deal with the Defense Department and comply with a request by Defense Secretary Pete Hegseth to make its AI available for any lawful use. The move occurred immediately after Anthropic had refused to do the same deal, prompting the White House to blacklist it from federal agencies. The negotiations with the U.S. military have caused turmoil within both Anthropic and OpenAI, and OpenAI CEO Sam Altman has said that OpenAI should not have rushed its initial agreement with the Pentagon. Before joining OpenAI, Kalinowski was the head of augmented reality glasses hardware at Meta Platforms. She also serves on the board of Axon, which supplies tasers, body cameras and other hardware and software to the military and law enforcement agencies. | | | | Microsoft, Amazon and Google will keep selling Anthropic models—and other products powered by its models—to most customers after the Pentagon said it was designating Anthropic a supply chain risk, the companies said. The Pentagon's letter to Anthropic, signed by Defense Secretary Pete Hegseth and dated March 3, states that the agency determined that Anthropic's technology can't be used by the Defense Department specifically, according to a copy of the letter reviewed by The Information. The letter does not say that other federal contractors need to sever ties with Anthropic, an idea Hegseth previously floated publicly. Anthropic said in a statement Thursday that it believes the supply chain risk designation is illegal and is preparing to challenge the Pentagon's designation in court. Microsoft said Thursday that its lawyers determined that the Pentagon's designation only barred Microsoft from including Anthropic products in its contracts with the Pentagon specifically, and that it wouldn't need to sever all business ties with Anthropic. "Our lawyers have studied the designation and have concluded that Anthropic products, including Claude, can remain available to our customers—other than the Department of War—through platforms such as M365, GitHub, and Microsoft's AI Foundry and that we can continue to work with Anthropic on non-defense related projects," a Microsoft spokesperson said in a statement. Google said in a similar statement Friday that the Pentagon's determination "does not preclude us from working with Anthropic on non-defense related projects, and their products remain available through our platforms." And Amazon said in a statement Friday that "AWS customers and partners can continue to use Claude for all their workloads not associated with the Department of War," and that for Pentagon contracts the company is "supporting customers and partners as they transition to alternatives running on AWS." | | | | SoftBank Group is working with four banks including JPMorgan Chase on a $40 billion loan to help fund its new investment in OpenAI, according to a person with knowledge of the discussions. The bridge loan would have a one-year tenor, according to the person, and be the largest U.S. dollar-denominated loan the Japanese company has ever taken out. Representatives for JPMorgan and SoftBank declined to comment. SoftBank said last month it planned to initially fund its new $30 billion investment in OpenAI through bridge loans and "other financing arrangements from major financial institutions." The new investment is on top of $34.6 billion it's already invested in the AI company. S&P Global changed its outlook for SoftBank's credit rating to "negative" following the investment decision, citing a drop in the creditworthiness of its holdings. Bloomberg first reported on the loan discussions. | | | | Top technology banker Niall Cannon is leaving Citi to join Morgan Stanley as a senior investment banker focusing on software, according to people familiar with the matter. Cannon was most recently a vice chairman of Citi's global technology investment banking and also served as its global co-head of software investment banking. Spokespeople of Morgan Stanley and Citi declined to comment. The departure is a loss to Citi's senior ranks. The investment banking arm of Citi has been aggressively hiring from rivals to boost its technology banking business. It hired two veteran tech bankers from JP Morgan, Drago Rajkovic and Pankaj Goel, late last year. | | | | Alphabet granted CEO Sundar Pichai new stock awards with a potential value of $686 million, the company said in a securities filing on Friday, citing his "strong performance" in the top job. It was the largest series of grants Pichai has received since he became CEO of Google in 2015. The Google parent grants Pichai big awards every three years. The last grant, in 2022, was worth as much as $336 million and in 2019 it granted him stock valued at up to $330 million. The latest grant includes several tranches of stock in Alphabet, with a potential value of $336 million, in addition to stock units in its self-driving car unit Waymo and drone delivery startup Wing worth up to $260 million and $90 million respectively. The new grants in Waymo and Wing potentially signal that Alphabet plans to spin those businesses off in the future. In the filing, Alphabet said both businesses "have made strong progress under Mr Pichai's supervision and show strong potential." | | | | Kalshi was sued by users for failing to pay $54 million to people who bet that Iranian Supreme Leader Ayatollah Ali Khamenei would leave office before March 1, after he was killed during the U.S. and Israeli airstrikes last month. After Khamenei's death, Kalshi said it couldn't pay out any winnings because its guidelines include a death carveout provision, which would prevent traders from directly profiting from death. The suit claimed that the provision is a "fine-print mechanism designed to allow [Kalshi] to avoid paying consumers what they were owed," according to a class-action lawsuit filed Thursday in the US District Court for the Central District of California. "Kalshi doesn't allow markets directly tied to death," said a company spokesperson. "Our rules were clear from the beginning, we never changed them, and we settled based on the rules." The lawsuit is the latest wrinkle in the development of prediction markets, which have surged in popularity but face any number of questions over what they can allow on their sites and how they should operate. | | | | Oracle and OpenAI will not expand their data center footprint at their Abilene, Texas campus beyond the initial 1.2 gigawatts, at a site being developed by startup Crusoe, according to people with direct knowledge of the situation. Oracle has leased eight buildings in Abilene for OpenAI, with the capacity to hold around 400,000 Nvidia Blackwell GPUs when completed, although only two of the eight buildings are done so far. Separately, as The Information reported last April, Oracle was trying to convince OpenAI to sign up for an additional six buildings. The firms have now decided against that expansion. The power won't be available for at least a year or so. And by the time it is available, OpenAI does not want to expand the cluster of Nvidia Blackwell chips when Nvidia will be selling its next-generation Vera Rubin chips. OpenAI would rather buy the new generation of chips for another data center, said one of the people with direct knowledge of the deal. Bloomberg first reported on the expansion talks ending. Bloomberg also reported that as a result, Nvidia is helping Crusoe try to lease the rest of the site to Meta Platforms. The talks with Meta are in early stages, according to someone briefed on the deal. | | | | The U.S. government is working on rules that would require American approval for all exports of AI chips, and on the conditions of stringent security promises and "matching" investments in American AI infrastructure, Bloomberg reported. The move would give Washington extensive control over how other countries build their data centers that power AI and on what terms. The level of scrutiny would depend on the size of the chip order. Small purchases would face a basic review with some exemptions, according to Bloomberg. Larger orders, sufficient to run an AI data center, would require government clearance upfront and could come with strings attached, including disclosing business plans or allowing U.S. officials to inspect facilities, Bloomberg reported. The biggest deals would face the highest bar. Any company seeking more than 200,000 of Nvidia's latest chips in a single country would trigger a government-to-government negotiation, with the buyer's nation required to make security guarantees and invest in American AI in return, according to Bloomberg. The proposal is still under review and could change before it is finalized, Bloomberg reported. | | | Popular articles By Ann Gehan and Sri Muppidi
By Stephanie Palazzolo, Erin Woo, Sri Muppidi and Amir Efrati
| | | | | 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 | | | | | |
0 comentários:
Postar um comentário