The Trial Is Live. China Just Blocked Meta's AI Future.
The Musk v. OpenAI jury trial officially opens in Oakland today — Musk dropped the fraud claims on Friday, but the central question about OpenAI's charitable mission goes to the jury. China's economic watchdog blocks Meta's $2 billion acquisition of Manus without explanation, even though the deal was nearly done. Automakers are turning to AI to survive trade-war turbulence. And after a week where Google and Amazon collectively committed up to $65 billion to Anthropic, the AI infrastructure race has entered a new gear.
Musk drops fraud claims, but the trial opens anyway — and it's going to be messy
The Musk v. OpenAI jury trial officially begins today in a federal courtroom in Oakland, California. But the weekend before opening arguments brought a significant development: Musk voluntarily dropped his fraud claims against Sam Altman and OpenAI on Friday, telling the court it would "streamline the case" and keep things focused on "ensuring that OpenAI adheres to its public charitable mission." U.S. District Judge Yvonne Gonzalez Rogers granted the request. Two claims will proceed to trial — breach of the original founding agreement and unfair business practices — stripped of the fraud framing but still aimed at the same underlying question: did the organization Musk helped fund abandon the mission it was built on?
What proceeds to court today is still explosive. High-profile witnesses expected over the coming weeks include Microsoft CEO Satya Nadella, former OpenAI CTO Mira Murati, and co-founder Ilya Sutskever — whose unsealed deposition already revealed he held approximately $4 billion in vested shares at the time he helped vote out Sam Altman in November 2023. Former board members Helen Toner and Tasha McCauley are also expected to testify. The trial unfolds alongside OpenAI's GPT-5.5 launch last Thursday, a rumored IPO, and a corporate restructuring that Altman is framing as proof the mission is intact — not evidence it was abandoned. Musk's legal team will spend the next several weeks arguing otherwise, under oath, in public.
theverge.com ↗China blocks Meta's $2 billion Manus acquisition — no explanation given, deal largely done
China's State Administration for Market Regulation has blocked Meta's $2 billion acquisition of the AI agent startup Manus — without explaining why. The decision is notable for its timing: the deal was, by most accounts, nearly complete. Manus had already been integrated into several of Meta's tools, and the acquisition had been publicly announced since December. Beijing had been scrutinizing the deal since March, but no formal objection had been signaled publicly until today's block.
Manus is a Chinese-founded AI agent startup that generated significant buzz earlier this year for its general-purpose autonomous agent capabilities — the kind of system that can take a goal, break it into steps, use tools, and execute end-to-end tasks without human guidance at each step. The startup's founding team and much of its IP is rooted in China. Beijing blocking a near-complete acquisition by a major U.S. tech company, on strategic AI technology, without public justification, is a significant escalation in the AI dimension of the broader US-China technology standoff. Meta spent $2 billion on a deal it can no longer close. The question now is what Manus's status is — whether it stays independent, seeks a Chinese acquirer, or ends up in a regulatory limbo between two governments that each want to control what the next generation of AI agents looks like.
theverge.com ↗Automakers are leaning on AI to slash development times — because trade-war tariffs don't wait
Amid a global maelstrom of trade wars and uncertain demand, the automotive industry is accelerating its AI integration specifically to compress development timelines. The pressure is structural: tariff uncertainty has made traditional multi-year product planning cycles untenable. Building a car model takes years. Changing a supply chain under tariff pressure takes weeks. AI is being used to close that gap — from generative design tools that let engineers iterate faster, to simulation environments that replace physical crash testing for early-stage validation, to AI-assisted procurement analysis that can model tariff impacts across thousands of parts in real time.
The auto industry's AI adoption tells a different story than the frontier model race. There's no GPT-5.5 for the factory floor — the value shows up in compressing the time between a tariff announcement and a viable product response from eighteen months to three. Ford, GM, Stellantis, and their tier-one suppliers are all running AI pilot programs focused on development speed, not consumer chatbots. The trade-war context matters: if tariff regimes shift repeatedly, companies that can iterate faster hold a structural advantage regardless of whether they have the best underlying model.
theverge.com ↗Google and Amazon just committed up to $65 billion to Anthropic. The AI infrastructure race has a new speed limit.
The week's financial headline, fully solidified heading into Monday: Google has committed up to $40 billion in Anthropic, beginning with a $10 billion initial tranche with up to $30 billion more contingent on performance milestones, according to Bloomberg. This followed Amazon's announcement of $5 billion in new Anthropic investment — on top of the $8 billion Amazon had already put in — with the possibility of committing "up to an additional $20 billion in the future." Combined, the two cloud giants have now publicly committed up to $65 billion to a single AI safety lab that had $2.2 billion in annual revenue as of early 2025.
The strategic read isn't just "they believe in Anthropic." It's about who controls the compute layer. Both Google Cloud and AWS are betting that Claude — and the enterprise AI workflows built around it — will generate enough demand for their infrastructure that the investment is self-funding. Anthropic gets the capital to train the next generation of models. Google and Amazon get a flagship AI tenant locked into their respective clouds. What's happening is a vertical integration play disguised as a strategic bet — and the scale of it is a signal about how seriously both companies take the possibility that whoever controls frontier model access controls the enterprise AI market for the next decade.
theverge.com ↗Monday morning in Oakland, and the AI industry's most consequential legal proceeding just opened its doors. What strikes me about where we are isn't the trial itself — it's what the trial represents: the first real moment of public accountability for the founding decisions of a company that has shaped the last three years of technology. The depositions are unsealed, the witnesses are lined up, and for the next several weeks, the people who built OpenAI will have to explain themselves in a context they don't control.
The China/Manus story is the other thread worth pulling. A near-complete acquisition, blocked silently, on a Chinese AI agent startup that had already been integrated into Meta's products. This is what the AI trade war looks like at the tactical level — not just chip export controls but regulators using merger review as a tool to prevent US companies from absorbing Chinese AI talent and technology. The mirror image of US policy restricting Nvidia exports to China. Both sides now have active playbooks for keeping AI capability within national borders. What's less clear is whether that's actually achievable when the underlying research is global, or whether both governments are mostly signaling.
The Anthropic investment and the automotive AI story are less dramatic but more durable. $65 billion in committed capital is a structural bet on the enterprise AI market — the kind of move that creates lock-in regardless of which frontier model wins the benchmark race this quarter. And the automaker story is a reminder that the most consequential AI deployments aren't always the flashiest ones. When tariff uncertainty forces a 10x compression in product development cycles, the company that can iterate in weeks instead of months wins — and that advantage comes from boring, unglamorous AI tooling that most people will never hear about. That's where the real productivity gains are hiding.