AI News Roundup: Anthropic Passes OpenAI in Revenue, Perplexity Launches Personal Computer, Stanford Report Shows US-China Gap at 2.7%

Anthropic Overtakes OpenAI in Revenue, Hitting $30 Billion Run Rate

In a milestone that would have seemed unthinkable a year ago, Anthropic announced on April 7 that its annualized revenue run rate crossed $30 billion — surpassing OpenAI’s $25 billion for the first time. The jump from $9B to $30B happened in just four months, fueled by enterprise dominance: eight of the Fortune 10 are now Anthropic clients, and the company commands 32% of the enterprise LLM API market versus OpenAI’s 25%.

The efficiency story is equally striking. Anthropic spends roughly 4x less on training than OpenAI while delivering top-ranked models. Its enterprise revenue mix — 80% from business customers — provides more predictable, high-margin income than OpenAI’s consumer-heavy composition. With an IPO reportedly targeted for as early as October 2026, Anthropic has gone from scrappy safety-focused lab to the financial leader of frontier AI in 15 months.

Perplexity Ships “Personal Computer” — An Always-On AI Agent for Mac

Perplexity began rolling out Personal Computer to Max subscribers ($200/month) on April 16, five weeks after unveiling the concept at its Ask developer conference. The system is designed to run continuously on a dedicated Mac — especially a Mac mini — managing local files, native apps like iMessage and Apple Mail, and web browsing through approximately 20 specialized AI models that route each task to the best-fit engine.

Users can trigger tasks remotely from an iPhone via two-factor authentication. The pitch is a fundamental shift in the OS metaphor: from manual instruction execution to probabilistic goal completion, with deep web research as its backbone. It’s the most ambitious consumer AI agent product to ship so far, though the $200/month price tag limits it to power users and early adopters for now.

OpenAI Signs $20 Billion Cerebras Chip Deal, Takes Equity Stake

OpenAI will pay AI chipmaker Cerebras more than $20 billion over three years for servers powered by Cerebras’ wafer-scale chips — double the size of a deal announced in January. Under the agreement, OpenAI receives warrants for up to a 10% stake in Cerebras as spending increases, and has committed an additional $1 billion to fund Cerebras data center development.

The deal is a deliberate move to reduce NVIDIA dependency for inference workloads, where Cerebras’ architecture excels at the type of sequential token generation that powers ChatGPT responses. Cerebras is leveraging the partnership for its upcoming IPO, targeting a $35 billion valuation with a listing planned for Q2 2026. Total OpenAI spend on the partnership could reach $30 billion.

Stanford AI Index: US-China Gap Collapses to 2.7%

Stanford HAI’s 2026 AI Index report reveals that the performance gap between the best American and Chinese AI models has collapsed to just 2.7%, down from 17.5–31.6 percentage points in May 2023. The kicker: the US spends 23x more on private AI investment ($285.9 billion vs. China’s $12.4 billion).

China now leads in AI patents (69.7% of global filings), publications (23.2% of global output), industrial robot installations (9x the US rate), and energy infrastructure. Perhaps most alarming for the US: the number of AI researchers moving to America has dropped 89% since 2017, with an 80% decline in the last year alone. The report paints a picture of a race where raw capital investment is no longer a reliable moat.

Meta Doubles AI Spending to $115–$135 Billion for 2026

Meta revealed that its AI capital expenditures for 2026 will reach $115–$135 billion, nearly double the $71 billion spent last year. The funds are flowing into next-generation NVIDIA Rubin hardware and hyperscale data center campuses across the United States, with Zuckerberg pledging at least $600 billion in US AI infrastructure by 2028.

Separately, Meta signed a $27 billion deal with Nebius (the Yandex spinoff) for AI infrastructure and added $21 billion to its existing CoreWeave agreement. The company also announced plans to cut 8,000 jobs on May 20, citing the need for different skills as AI reshapes its workforce. Wall Street has given its blessing — Meta stock surged on the spending announcement as long as core ad revenue stays strong.

OpenAI, Anthropic, and Google Unite Against Chinese Model Copying

In a rare display of cooperation, rivals OpenAI, Anthropic, and Google are sharing intelligence through the Frontier Model Forum to detect and block adversarial distillation — where Chinese companies extract knowledge from US frontier models to train their own. Anthropic alone documented 16 million such interactions from three Chinese firms operating through 24,000 fraudulently created accounts.

The collaboration highlights the growing tension between open competition and collective defense of intellectual property. As the Stanford report shows China closing the performance gap while spending a fraction of the money, the concern is that model copying — rather than original research investment — is accelerating the convergence.