| |  | Mar 11, 2026 | | | | | | Happy Wednesday! Microsoft backs Anthropic's lawsuit against the Pentagon. Nvidia and Mira Murati's Thinking Machines Lab announce a strategic partnership. Tencent is building a new AI agent for WeChat in a top-secret project.
| | | | Microsoft on Tuesday backed Anthropic's federal court request to temporarily block the Pentagon from cutting business ties with the AI firm, arguing that the Pentagon's action could harm Microsoft's business and the broader tech sector. In a filing Tuesday, Microsoft told a federal judge it agrees with Anthropic's request for a temporary restraining order that would pause the Pentagon's designation of Anthropic as a supply chain risk until the court could determine the legality of that action. AI researchers from Google and OpenAI have also filed a similar brief backing Anthropic's position. "Should this action proceed without the entry of a temporary restraining order, Microsoft and other government contractors with expertise in developing solutions to support U.S. government missions will be forced to account for a new risk in their business planning," Microsoft said in the filing. Microsoft and Anthropic have become close business partners in the past year. The Pentagon last week said Anthropic's technology is a supply chain risk and could no longer be used in any Defense Department contracts after the startup sought special assurances its tech wouldn't be used for fully autonomous weapons and mass domestic surveillance. (Read a full account of the affair here.) Anthropic filed a lawsuit Monday claiming the designation was an illegal retaliation by the Trump administration. | | | | Nvidia and Thinking Machines Lab, the model developer co-founded by former OpenAI Chief Technology Officer Mira Murati, announced a strategic partnership on Tuesday to deploy at least one gigawatt of servers powered by Nvidia's forthcoming Vera Rubin chips, according to a blog post. The companies said that they would aim to deploy those servers early next year. Nvidia has also made a "significant investment" in Thinking Machines, the blog post said, though it didn't disclose any details of the investment. Nvidia has backed a number of "neolabs," or newer AI labs that hope to exploit new approaches to developing AI models and research they say major developers like OpenAI and Anthropic may have overlooked. Those include Reflection AI, Humans& and Periodic Labs. That could help the chipmaker diversify its customer base beyond the major AI labs like OpenAI and Anthropic. | | | | Tencent Holdings is secretly building a new AI agent for its hugely popular WeChat messaging app, in hopes of leapfrogging rivals like Alibaba Group and ByteDance in the race to dominate China's domestic AI market, The Information reported. The new agent would connect with the millions of miniprograms—lightweight apps—running inside WeChat that provide scores of services, from booking taxis to ordering groceries. If it works, the agent could undertake those tasks on behalf of WeChat's 1.4 billion monthly active users. Tencent's move is emblematic of the stakes AI agents hold for tech titans. From Silicon Valley to China, companies are racing to launch AI assistants that can autonomously complete tasks for both workers and consumers, from coding to commerce. Tencent plans to launch the WeChat AI agent feature for a trial among select users by midyear—a process known as gray-box testing in software development—and roll out to all users by the third quarter. But that timeline could still shift as WeChat won't release the agentic features until they are solid. | | | | Oracle shares rose 8% in after-hours trading Tuesday after the firm posted strong cloud-server rental growth in the February fiscal quarter, line with its earlier projections. It also raised its projected revenue for the fiscal year that starts in June by about 1% to $90 billion. The disclosures may allay some concerns about the enterprise firm's ability to quickly transform itself into a provider of AI cloud servers to OpenAI, Meta and other large developers. The firm also said some of its customers are buying their own chips, lessening Oracle's capital requirements as it builds data center campuses. The buildout has depleted Oracle's cash reserves, prompting the company to raise tens of billions of dollars this year. The firm burned $11 billion in the February 2026 quarter, up about $1 billion from its cash burn in the prior quarter. Oracle co-CEO Clay Magouyrk said during a call with investors that the firm delivered 400 megawatts of capacity in the quarter, and that 90% of that capacity was delivered on time. He also said gross margins of that business were 32%, in line with its expectations of its gross margins being between 30% and 40%. Even with the bump following the Tuesday earnings report, Oracle shares are down 50% since September of last year, when it disclosed large expected payments from OpenAI and other AI cloud customers. Investors have since grown skeptical of customers' ability to pay Oracle and Oracle's ability to raise enough capital to develop facilities to achieve its sky-high revenue targets. Oracle's cloud server-rental revenue grew 84% year-over-year to $4.9 billion in the February quarter. That growth rate was 16 percentage points higher than in the November 2025 quarter and 61 percentage points higher than cloud growth in the February 2025 quarter. Oracle's cloud revenue is about six times smaller than that of market leader Amazon Web services. Oracle said it had $553 billion in remaining performance obligations—a metric reflecting expected revenue—up around $30 billion from three months earlier. Oracle said most of the increase between quarters was related to large scale AI contracts and that Oracle won't need to raise additional money to support the projects. "Most of the equipment needed is either funded upfront via customer prepayments so Oracle can purchase the [Nvidia graphics processing units], or the customer buys the GPUs and supplies them to Oracle," the firm said. Oracle said its over revenue in the February quarter grew 22% from the same period last year to $17.2 billion. Oracle still expects its fiscal year 2026 revenue to be $67 billion and for its capital expenditures to be $50 billion. | | | | Google is providing its Gemini "Agent Designer" to the Department of Defense for unclassified work, Google announced on Tuesday. The feature, which will be part of the Pentagon's AI platform GenAI.mil, will allow DOD employees to create AI agents to automate administrative tasks like drafting meeting notes or breaking a project into a step-by-step checklist, Google said. The announcement comes one day after Google's rival AI lab Anthropic sued the Pentagon for declaring it a supply chain risk. The Pentagon recently struck deals with OpenAI and xAI to offer their models on classified systems. Meanwhile, Google is in talks with the Pentagon for putting the agent designer on classified systems as well, defense undersecretary Emil Michael told Bloomberg on Tuesday. Tech companies have broadly become more comfortable working with the military in recent years. Even Anthropic, in its negotiations with the Pentagon, stressed that it wanted to work with them and had just two red lines on mass domestic surveillance and lethal autonomous weapons. In 2018, Google declined to renew its Project Maven deal—which included AI drone targeting—after employee protests and published AI principles banning similar applications. Google rescinded those principles in 2025, and the Department of Defense awarded Google a $200 million AI pilot program contract alongside OpenAI, Anthropic and xAI last summer. | | | | Meta Platforms hired the creators of Moltbook, a social network designed for AI agents, Matt Schlicht and Ben Parr, who will join its AI division, Meta Superintelligence Labs (MSL), the company said. Meta also bought Moltbook itself. Meta said it hopes the integration of the Moltbook team into MSL "opens up new ways for AI agents to work for people and businesses." Moltbook is a Reddit-like forum where AI agents post and interact with one another about tasks they are carrying out for their human owners, as well as other topics related to their workflows. Some discussions also venture into more philosophical territory. A recent popular thread, for example, discussed "artificial self-identity" and consciousness. Moltbook's homepage now states that it hosts more than 194,000 AI agents verified by their human owners, with over 2.8 million agents registered on the platform overall. | | | | A cybersecurity agency in China issued a warning about the security risks of using OpenClaw, the viral AI agent builder that has become wildly popular in the country. OpenClaw's fragile security configuration could allow attackers to gain full control of the system, said the National Computer Network Emergency Response Technical Team/Coordination Center of China, also known as Cncert, in a post published Tuesday on its official WeChat social media account. For key industries such as finance and energy, the use of OpenClaw with insufficient security measures could lead to the leakage of core business data, trade secrets and source code, the agency warned. The agency recommended that users implement additional security measures when deploying OpenClaw, such as limiting the AI agent's access to the system, adopting a more stringent authentication process and tightly screening any plug-in software components. OpenClaw, which has captivated Silicon Valley since late January, has also taken China by storm. Chinese tech entrepreneurs became obsessed with OpenClaw and started building new products or business models that involve AI agents, while the country's biggest tech firms, Alibaba Group, Tencent Holdings and ByteDance, have launched services running OpenClaw on their cloud computing platforms. | | | | Amazon began a corporate bond sale aimed at raising as much as $42 billion, Bloomberg reported, the latest sign of how tech giants are borrowing money to fund their AI expansion. Amazon had filed paperwork a few weeks ago allowing it to quickly sell bonds. The new bond offering follows one last November, when Amazon raised $15 billion, giving it cash reserves of $123 billion (along with total debt of $65.6 billion). Amazon has projected it will spend $200 billion on capital expenditures this year, an increase of 50% on last year, and will invest as much as $50 billion in OpenAI. The projected capex is more than Amazon's operations are expected to generate in cash, meaning it would have to dip into the cash on its balance sheet to fund it. Making the OpenAI investment would further reduce Amazon's cash reserves. Amazon, like other big tech companies, have the capacity to borrow significantly more money, however. | | | | Former Meta Platforms chief AI scientist Yann LeCun has raised more than $1 billion for his new AI startup, Advanced Machine Intelligence Labs, from investors that included Nvidia, Greycroft, former Google CEO Eric Schmidt and Jeff Bezos. LeCun said it was valued at more than $4.5 billion. The startup announced Tuesday that it aims to build "world models"—AI systems that can learn from simulated real-world environments and human interactions. Alexandre LeBrun, the former CEO of French startup Nabla, will serve as CEO. Aside from LeCun, AMI has hired Meta's former Europe VP Laurent Solly as COO and two former directors from Meta's FAIR research lab: Pascale Fung will serve as chief research and innovation officer, and Michael Rabbat as a cofounder. Saining Xie, a former research scientist at Google DeepMind who worked on the GenAI and Nano Banana team, and previously a research scientist at Meta, is joining as chief science officer, according to his LinkedIn profile. | | | | Salesforce plans to sell between $20 billion and $25 billion in bonds as soon as this week, to help fund its recently-announced $50 billion stock buyback, Bloomberg reported. The bond offering, coming as Amazon was also in the market trying to raise twice as much, shows how a flood of tech companies have started borrowing money. Most are doing so to finance their AI expansions, however. Ahead of Salesforce's expected bond offering, credit ratings agency Moody's lowered its rating on the company's debt to A2 from A1. S&P signaled it may do the same, changing its outlook on Salesforce to "negative" from "stable" but maintaining Salesforce's A+ credit rating for now. Moody's Ratings Senior Credit Officer, Matthew Jones said Salesforce's "significantly stepped up share repurchase authorization, together with expected debt funded repurchases, represents a material shift in financial policy." Salesforce announced the $50 billion buyback on Feb. 25 during its quarterly earnings report. Shares of the software giant have fallen 27% so far this year as fears grow that AI will disrupt the software industry. On last month's earnings call, CEO Marc Benioff called the plunge a "great buying opportunity." Salesforce bought back $12.6 billion worth of shares in the year through January 2026, compared to just below $8 billion in each of the two prior fiscal years. | | | 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