Microsoft, NVIDIA and Anthropic Hold Hands: Claude Goes GA on Azure's Blackwell Ultra
2026-06-29
“The three companies that spent a decade side-eyeing each other just merged their roadmaps, and the only people who win are the developers paying half price.”
Claude is now generally available on Microsoft Foundry, running on Azure, powered by NVIDIA's GB300 Blackwell Ultra silicon. Read that sentence again. Three of the most valuable companies on Earth, who have spent years politely implying the other two were a threat to civilization, just stapled their roadmaps together and called it a partnership. It is the enterprise equivalent of exes agreeing to co-parent because the kid is profitable.
The comedy is in the loyalties. Microsoft is OpenAI's biggest backer and now resells OpenAI's fiercest rival on its own cloud, on the same invoice. NVIDIA sells the shovels to everyone in the gold rush and pretends to have no favourites. Anthropic, whose whole pitch is being the careful, principled lab, is now shipping on a competitor's infrastructure because that is where the customers and the GPUs live. Nobody blinked. Everybody won. That is either mature capitalism or a very well catered standoff.
Strip away the group hug and the substance is real. Opus 4.8 and Haiku 4 are live, latency drops, throughput climbs, and the token bill falls by about half. The only thing louder than the engineering is the buzzword fog around it. If you can say 'autonomous enterprise agents' three times without laughing, there is a keynote slot waiting for you.
- Anthropic's Claude models reached general availability on Microsoft Foundry, hosted on Azure, on June 29, 2026.
- The deployment runs on NVIDIA GB300 Blackwell Ultra systems, the first Claude models to do so inside Azure.
- Claude Opus 4.8 and Claude Haiku 4 are the two models live at launch.
- Availability opens in Azure East US, West Europe and Southeast Asia, with more regions promised through Q3 2026.
- Anthropic and multiple outlets say the GB300 infrastructure cuts token costs by roughly 50 percent versus the previous setup.
- 01
This is the rare Big Tech press release where the substance outruns the spin. Claude actually running multi cloud is a genuine win, not a slide. Blackwell Ultra makes inference faster and roughly half the price, which means the agents you were already building just got cheaper to run and less painful to scale. Corporate soap opera aside, developers and the enterprises paying their invoices come out ahead. Take the cheaper tokens and do not ask the trio to define the relationship.
- 01
OpenAI, mostly, and quietly. Its single biggest investor now sells its sharpest competitor from the same storefront, on the same Azure bill, with a straight face. Also burned: anyone who still believes cloud loyalty is a real thing rather than a spreadsheet decision that changes the second the margins do. And spare a thought for every comms team now expected to explain, without cracking, why 'autonomous enterprise agents' is a sentence humans say out loud.
No meter. No surprises.
The story isn't Copilot. It's the meter. Here is the calmer way to get your code reviewed.
- Flat price for the Full Suite. No usage meter, no month-end surprise.
- Free CLI: 90 roasts a month, no account needed.
- Privacy-first: a dry-run shows the exact payload, and your secrets never leave your machine.
Works in Claude Code, Cursor and Windsurf via MCP. Open source, and proud of it.
Got something the world should see roasted? Drop it.
A full teardown from €2,99. No mercy.