Anthropic の収益化への歩み（3 分読了）

Anthropic is projecting $10.9 billion in Q2 revenue, up more than 2x from $4.8 billion in Q1. Not only that, but it’s expecting to report $559 million in profit, making Q2 the first profitable quarter in the company’s history just in time for its IPO, which is expected to take place in October.

Anthropic’s pace of growth is remarkable; as WSJ pointed out, the company’s quarterly revenue is now growing faster than Zoom’s did during the early part of the COVID-19 pandemic. The biggest driver of that growth is enterprise adoption of Claude’s coding tools. As we noted in a recent perspective, Anthropic’s run-rate revenue had already climbed to roughly $40 billion annualized even before these Q2 projections, with Claude Code alone generating over $2.5 billion.

The unit economics underlying the growth are almost as impressive as the growth itself. In Q1, Anthropic spent 71 cents on compute for every dollar of revenue. In Q2, it expects to spend 56 cents. Not only is Anthropic growing quickly at scale, but its margin is improving. This likely explains the feverish demand for Anthropic shares in private markets, sometimes through less-than-reputable SPVs. Anthropic took aim at such vehicles in a post warning investors against unauthorized stock sales, where it stated in no uncertain terms: “Any sale or transfer of Anthropic stock, or any interest in Anthropic stock, that has not been approved by our Board of Directors is void and will not be recognized.”

Reports emerged this week that the Trump administration is in talks to take equity stakes in quantum computing companies such as IonQ, Rigetti, and D-Wave, with each seeking at least $10 million in federal funding through the Commerce Department. The stocks surged on the news before partially reversing after a Commerce Department official said it was “not currently negotiating with any of the companies,” and a separate source said that equity stakes are “not necessarily something the Trump administration is considering.” All four companies remain unprofitable, and China’s public funding for quantum is roughly double that of the US.

This is part of a broader theme we’ve been covering surrounding the nationalization of critical tech sectors, such as the government taking a 10% stake in Intel last August and the Defense Department taking a 15% stake in rare-earth miner MP Materials the month prior, as the urgency around rebuilding rare-earth supply outside of China increases. Quantum has become increasingly urgent, with Google recently estimating Q-Day could arrive as soon as 2029. Quantum supremacy is inseparable from the race to break and defend encryption. A nation that achieves a meaningful quantum advantage over its adversaries would gain a significant edge in surveillance, intelligence, computing, and cyber warfare, making it a national security concern.

Last week, South Korea’s presidential chief of staff for policy, Kim Yong-beom, shared a 2.5K-word Facebook message arguing that the massive AI-era windfall flowing to Samsung Electronics and SK Hynix should be “structurally returned” to the public as a “national dividend,” writing that “The fruits of the AI infrastructure era are not the result of certain corporations alone…They come from an industrial foundation that the entire people have built together over half a century.” The Kospi fell as much as 5.1% the next day, with foreign investors pulling 5.6 trillion won from the exchange in a single session amid fears that profits at Samsung and SK Hynix would be redirected from shareholders.

South Korea’s president, Lee Jae Myung, later clarified that the post was meant as “a review of a plan to distribute the national excess tax revenue generated from excess profits in the AI sector to the public as a citizen’s dividend,” not a hardened plan. This conversation comes as Samsung union workers have threatened to strike to increase their share of AI-related profits. The scale of the potential windfall is staggering, with the two firms’ combined operating profits projected to reach 600 trillion won this year, roughly a quarter of South Korean GDP. Comparisons to Norway’s sovereign wealth fund and proposals for a windfall tax or a UBI trial have already faced criticism on the grounds that they might alienate engineering talent or investment.

• Lead Bank is a BaaS platform that was valued at $1.5 billion in its Series B late last year. Read our new report here.

• President Trump planned to sign an executive order on AI cybersecurity, inviting tech leaders to the White House, with a voluntary framework requiring AI labs to share covered frontier models with the government 90 days before public release, before the signing was abruptly postponed Thursday afternoon.
• OpenAI prepared to confidentially file IPO paperwork as soon as Friday, targeting an autumn debut at a potentially trillion-dollar valuation; Kalshi prediction-market odds of Anthropic listing first collapsed from 69% to 20% on the news, and CFO Sarah Friar has called it “good hygiene” to “look and feel and act” like a public company.
• Minnesota Governor Tim Walz signed the nation’s first state law banning prediction markets like Kalshi and Polymarket, with the CFTC immediately suing in federal court, joining five other state-vs-CFTC battles over whether prediction markets are “event contracts” (federal jurisdiction) or gambling (state jurisdiction).
• SpaceX plans to complete its $60 billion acquisition of AI coding startup Cursor 30 days after its expected June 12 Nasdaq debut under ticker SPCX, with a $10 billion cash breakup fee if the deal falls through.
• Google expanded SynthID watermarking from Gemini to Search, Chrome, Circle to Search, and Pixel devices at I/O 2026 with OpenAI, Nvidia, Kakao, and ElevenLabs now embedding the invisible per-pixel watermark in their AI outputs; the technology is designed to survive compression, cropping, rotation, and screenshots.
• The SEC readied an “innovation exemption” for tokenized stocks this week, allowing crypto-native platforms to offer on-chain trading of digital equities (including third-party-issued versions of stocks without the underlying companies’ consent).
• Google launched Gemini for Science, a research suite spanning Co-Scientist, Literature Insights, Computational Discovery, and Deep Think mode, with DeepMind CEO Denis Hassabis saying the company hopes to “reimagine the drug discovery process with the goal of one day solving all disease.”
• Blackstone committed $5 billion in equity to a new joint venture with Google. The venture offers Google’s TPUs as a compute-as-a-service alternative to Google Cloud with plans to bring 500 MW of capacity online in 2027, with plans to scale significantly in the future, making Blackstone both the world’s largest alternative asset manager ($1.3T trillion AUM) and largest global data center provider.
• Meta reassigned 7K employees into four new AI-focused organizations using “AI native design structures” with fewer managers per employee, days before it laid off 8K employees, or 10% of its total headcount.
• Armada raised a $230 million Series B at a $2 billion valuation to fund a 400K square foot Arizona factory (Galleon Forge One) that will produce its Leviathan megawatt-scale modular data centers that attach to existing energy sources like solar and gas flares.

*At Contrary Research, we’ve built the best starting place to understand private tech companies. If you're interested in researching and writing about tech, apply here for our Research Fellowship!*