OpenAI Proposes New Deal, Farrow Probes Altman.

Mazen Al Sabbagh

— 4 min read
Share

TL;DR

  • UK Courts Anthropic -- Britain moves to exploit the Pentagon fallout, pitching London expansion and dual listing. A clear strategic capture. (FT/Reuters)
  • OpenAI's $122B Capital Wave -- A record private raise at an $852B valuation. Amazon, Nvidia, SoftBank fuel the growth, but contingent stakes reveal deeper market plays. (OpenAI, CNBC)
  • China's AI Economy Accelerates -- 140 trillion daily AI tokens processed in March, a 1,000x surge. DeepSeek V4-Lite already live. This is scale redefining the game. (Digital in Asia)
  • China Regulates Digital Humans -- Draft rules mandate labeling, ban virtual companions for minors, and prohibit non-consensual likenesses. Proactive control shaping a new social architecture. (US News)
  • Google Capitalizes on Sora's Silence -- Veo 3.1 price cuts and a new Lite version land as OpenAI's Sora vanishes. A strategic vacuum filled. (Google Blog)

Lead Story: The Anthropic Gambit: UK Seizes on US Discord

A vulnerability exposed. The UK moves. British officials are actively pitching Anthropic on expanded London operations, even a dual stock listing on the London Stock Exchange. This follows Anthropic’s escalating clash with the Pentagon. Reports confirm the push.

Prime Minister Keir Starmer’s office backs the play. The proposals target CEO Dario Amodei directly during his late May visit. This is calculated.

The DOJ's Ninth Circuit appeal, blocking the Pentagon's "illegal First Amendment retaliation" against Anthropic, briefs through April 30. Enforcement remains contested. Emil Michael’s defiance of the injunction lingers. Anthropic faces genuine uncertainty in its US government dealings.

A London listing offers access to European capital, a symbolic hedge against Washington. For Britain, it’s a marquee AI win, positioning London as a credible hub for safety-focused AI firms. The "why" is clear: attract top-tier AI.

This British pitch isn't isolated. California's Governor Newsom already decoupled state AI procurement from federal decisions. If democracies begin competing for Anthropic's presence, the Pentagon’s blacklisting may be achieving the opposite of its intended effect: a global dispersion, not a containment, of cutting-edge AI.

In Other News

OpenAI closes $122 billion round — the largest private raise in history. Amazon committed $50 billion, with Nvidia and SoftBank adding $30 billion each. The $852 billion post-money valuation solidifies OpenAI’s position. Note the fine print: $35 billion of Amazon's stake hinges on a public offering or AGI. Meanwhile, Bloomberg reports ~$600 million in secondary shares struggle to find buyers, even with Goldman Sachs waiving fees. Capital flows are complex, even at this scale.

China's AI machine hits industrial scale. A comprehensive Digital in Asia analysis reveals the velocity: 140 trillion daily tokens processed in March, a 1,000-fold increase since early 2024. Chinese models now drive 61% of global token consumption on OpenRouter. DeepSeek V4 – a 1-trillion-parameter MoE model – will run entirely on Huawei's Ascend 950PR chips. Midday update: V4-Lite is already live on API nodes, showing 30% inference speed gains and 94% context recall at 128K tokens. The full V4 is weeks away. This isn't just growth; it's a structural shift in global AI architecture.

China drafts sweeping rules for AI "digital humans." The Cyberspace Administration of China published regulations requiring prominent labeling, banning digital companions for minors, and prohibiting non-consensual AI likenesses. As Futurism notes, China is preemptively regulating societal impact while the US lags. The public comment period closes May 6. A deliberate shaping of AI's integration.

Google slashes AI video generation costs ahead of Sora's silence. Veo 3.1 Fast sees a price cut effective April 7. Google also launched Veo 3.1 Lite at less than half the cost: $0.05/second for 720p. The timing is stark: OpenAI shut down Sora entirely, burning $15 million daily before the kill switch. Google now commands the AI video generation market by default, a strategic vacuum filled.

X / Social Pulse

The Britain-Anthropic story dominated. The geopolitical absurdity was a common thread: the US blacklists its own AI safety lab, and allies immediately try to onboard it. AnthroPAC's Friday FEC filing also drew skepticism from the right, with Breitbart highlighting 99.84% of past Anthropic-affiliated donations to Democrats, challenging the "bipartisan" narrative. On the China front, the 140-trillion-token stat went viral, with many calling tokens "the new oil." Bloomberg's piece on "vibe coding" FOMO hit a nerve – the fear that AI coding tools are extending, not reducing, work hours.

One to Watch

DeepSeek V4's Huawei debut. The full model, a 1T parameter MoE, is expected within weeks. V4-Lite is already showing significant promise. If the full release meets its leaked benchmarks (81% SWE-bench, 90% HumanEval) while running entirely on domestic Chinese chips, it’s the clearest signal yet: export controls haven't preserved America's AI compute advantage. ByteDance, Alibaba, and Tencent are already bulk-ordering Ascend 950PR chips, driving prices up 20%. A true test of strategic autonomy.

Quick Hits

  • Georgia's legislature adjourns today with three AI bills on Governor Kemp's desk: chatbot disclosure (SB 540), an AI study committee (SR 789), and an insurance AI ban (SB 444).
  • ProPublica published three federal AI cautionary tales — contrasting Biden-era cyber defense partnerships with Trump's $1/seat ChatGPT deals for agencies. A lesson in conflicting strategies. (ProPublica)
  • SpaceX-xAI revised its IPO valuation target above $2 trillion in investor talks, up from the initial $1.75T filing. A June listing would precede both OpenAI and Anthropic. The valuation game intensifies. (CNBC)
  • AI-powered virtual try-on is going mainstream — Google integrating the tech into search results April 30. A direct attack on $850B in annual retail returns. Market architecture shifting. (CNBC)
  • Ledger CTO warns AI is making crypto hacks cheaper and easier — $1.4B in crypto losses over the past year. AI lowers the barrier for attackers, a dark acceleration. (CoinDesk)

The Pentagon's fight with Anthropic is no longer contained. It’s reshaping how allied nations compete for essential AI talent and how China measures its own progress against a fracturing American front. The global market incentives are clear: opportunity favors the adaptable.

Sources

Anthropic / Pentagon / UK

OpenAI Funding

China AI Strategy / DeepSeek / Regulation

Google Veo / Video AI

Federal AI / State Regulation

IPO / Markets / Culture

Lock in. M. mazen@thorterminal.com

Read more