Anthropic Raises $65 Billion and Passes OpenAI as Endava Cuts Analysis From Weeks to Hours

01The AI money was right there. A former Meta engineer built a website instead.

Craig Campbell had the resume to raise for an AI startup. He was an engineer at Meta, then a founder who sold his last company, an e-commerce tool for businesses, in 2022. Investors were funding almost anything with a model attached. Campbell built a website instead, according to The Verge.

The choice runs against the prevailing math of the open web. Google increasingly answers queries directly with AI summaries, cutting the clicks it once sent to ranked sites. Publishers call this "Google Zero," the moment search stops delivering traffic. AI tools also scrape and reproduce web content at almost no cost. By that logic, launching a website in 2026 is a way to lose money.

Campbell's bet is paying off anyway, The Verge reports. His wager rests on a simple inversion: when machines can manufacture content for free, content a human can verify gets scarce, and scarce things command a price. Trust becomes the product the AI flood cannot copy.

The cheap end of that flood is already visible on TikTok. The Verge reported that operators are using AI to generate fake people who sell low-cost goods through TikTok Shop. One persona, a light-skinned Black woman named Aliyah in country-western gear, cried to the camera in a March video and begged for views while pitching handmade metal buckles. According to The Verge, sellers like her are synthetic, assembled by dropshippers to move Shein-grade merchandise to viewers who think they are watching a real person struggle.

Those two stories sit on opposite sides of the same shift. AI makes it trivial to fabricate a sympathetic face and a sob story, which drives down the worth of any face and any story online. That collapse is exactly what gives a documented, real, human-run site its margin. Campbell is not selling against AI tools. He is selling the one thing they devalue every time someone spins up another fake influencer.

The open question is whether that premium holds as synthetic sellers get cheaper and harder to spot. For now, the dropshippers and the engineer are running the same experiment from opposite ends, and Campbell's revenue is the early scoreboard.

02Endava says an OpenAI agent cut requirements analysis from weeks to hours

Three customer stories landed on OpenAI's site, spanning a startup, an IT services contractor, and Japan's largest bank. Read together, they describe one move: pushing the agent up the org chart, past the developer who writes code, into the work that comes before it.

Braintrust, the startup, says its engineers route customer requests straight through Codex running GPT-5.5, producing experiments and code with fewer handoffs. The scope is narrow. One company, one engineering team, faster output.

Endava claims the larger structural shift. The IT services firm reports cutting requirements analysis from weeks to hours, according to OpenAI's writeup. That step normally sits before any code gets written: turning what a client asks for into a build specification. Endava calls the result an "agentic organization."

MUFG runs on different software. The bank uses ChatGPT Enterprise rather than Codex, and frames its goal as becoming "AI-native," with reworked workflows and new AI-powered financial services. It reports no before-and-after metric comparable to Endava's.

Stacked up, the three describe the agent moving from a tool that writes code to one that reorganizes who does which job. Two caveats sit on top. Every figure here is OpenAI's account of its own customers, published by the vendor selling the product. And the headline number measures one phase, requirements analysis, not delivery end to end.

What none of the three disclose is the cost side: error rates, downstream rework, or whether the compressed analysis step produced specs that survived contact with the build. "Weeks to hours" describes time saved at one desk. It says nothing about where that work reappeared, or whose. For a contractor like Endava, billing clients by the project, faster analysis changes the unit economics of every bid. The next signal worth tracking is whether any of the three publish a number OpenAI didn't supply.

03Anthropic raised $65 billion and passed OpenAI; in Paris, Mistral stopped calling itself a model company

Anthropic has overtaken OpenAI as the most valuable private AI company in Silicon Valley, according to a Qazinform report. The developer of the Claude assistant said it raised $65 billion in a Series H round, pushing its valuation close to $1 trillion. That figure is nearly three times the roughly $380 billion the company carried in February.

Anthropic reported annual revenue of $47 billion, up from about $10 billion a year earlier. It credits Claude and Claude Code, the coding service used by software developers. Altimeter Capital and Sequoia led the round, which folded in a previously agreed $5 billion from Amazon. Alongside the raise, the company introduced Claude Opus 4.8 and a closed system called Claude Mythos Preview, aimed at corporate cybersecurity.

In Paris, Europe's flagship lab spent its AI Now Summit describing a different shape. One attendee's published notes say Mistral is no longer just a model company; it now wants the full stack of compute, models, platforms, and consultancy. The company owns its compute, running a 40MW data center in Paris with more planned, including one in Sweden. Its pitch, per those notes, is efficient open models that customers own and can run on-prem, set against Anthropic and OpenAI.

The summit leaned on partnerships over product. The same notes describe collaborations with ASML, BNP Paribas, and Amazon's Alexa+, with less attention to new models, which the writer called disappointing. Mistral did launch Vibe for Work, a product the notes compare to Claude for Work.

Its stated strategy now rests on small, specialized models. The notes cite examples where focused models beat larger general-purpose ones on energy efficiency. Where Anthropic sells scale and a near-trillion-dollar balance sheet, Mistral is selling ownership and on-prem control to European enterprises like ASML and BNP Paribas. The summit said little about when its next frontier model arrives.