Analytics Drift

Anthropic is closing a funding round this week that would value it above $900 billion, according to Bloomberg, making it the most valuable private AI company in the world and officially pushing OpenAI to second place. For the first time since OpenAI set the benchmark in March 2026, there is a new number at the top of the AI capital stack. The round is expected to exceed $30 billion.

Zoho founder Sridhar Vembu called AI “the biggest investment bubble yet” this weekend. He was careful to add that the technology is real. What he was reacting to is something Wall Street analysts have been quietly calling circular financing, and the numbers in SEC filings make it hard to dismiss. Circular financing is a deal structure where money flows between companies in a closed loop. A tech giant invests in an AI startup.

Anthropic told investors this week that it expects to generate $10.9 billion in revenue in the second quarter of 2026, more than double the $4.8 billion it reported in Q1, and to post its first-ever operating profit of $559 million in the process. The Wall Street Journal broke the story on May 20, with CNBC and Bloomberg independently confirming the figures.

Every major AI lab is racing to ship a personal agent. OpenAI has its ChatGPT Agent. Anthropic has Claude Cowork. And now Google has Gemini Spark, announced at Google I/O 2026 on May 19. But Spark is not just another entrant in an already crowded field. It is the first agent that arrives pre-loaded with the most comprehensive view of your digital life that any company on earth has ever assembled.

Andrej Karpathy joins Anthropic, and the AI industry should pay attention. Not because of the hire itself, but because of what the choice reveals about where the frontier is actually being fought. On May 19, 2026, Karpathy announced on X that he is joining Anthropic’s pre-training team. “I think the next few years at the frontier of LLMs will be especially formative,” he wrote. “I am very excited to join the team here and get back to R&D.” He started the same week.

Two years ago, Cursor was a VS Code fork that called the OpenAI API. Today, it is training its own frontier models on open-source checkpoints, running that training on xAI’s Colossus 2 supercluster, and shipping a model that competes with Anthropic’s Opus 4.7 at $0.50 per million input tokens. That trajectory is the most important signal in AI product development right now, and Cursor Composer 2.5 is the clearest proof of it yet.

A court filing entered into evidence Tuesday in the Musk v. Altman trial put a precise number on something that had long been discussed in the abstract: OpenAI CEO Sam Altman holds more than $2 billion in personal stakes across nine companies that have active commercial ties to OpenAI. The Sam Altman conflict of interest OpenAI case is no longer a matter of speculation. It is now public record, in a federal courthouse in Oakland.

Cisco posted the strongest quarter in its history on Wednesday. Revenue hit $15.84 billion, up 12% year over year. Net income surged to $3.37 billion. AI infrastructure orders from hyperscalers reached $5.3 billion for the fiscal year to date, forcing the company to revise its full-year AI order forecast from $5 billion to $9 billion. Investors loved it: shares jumped 17% in after-hours trading.

Demis Hassabis just raised $2.1 billion for his AI drug design company. The first human clinical trial has not happened yet. That tension is the entire story. Isomorphic Labs, the London-based AI drug design company founded in 2021 as a Google DeepMind spinout, closed a $2.1 billion Series B on May 12, 2026. The round was led by Thrive Capital, which also led the company’s $600 million Series A in 2025. New investors include Abu Dhabi’s MGX, Singapore’s Temasek, and the UK Sovereign AI Fund.

Thinking Machines Lab, the AI startup founded by former OpenAI CTO Mira Murati, released a research preview of what it calls Interaction Models on May 11, 2026. The claim at the center of the release is direct: every major AI lab has built its interaction layer as an afterthought, and the resulting latency and limitation is not a tuning problem but an architectural one.