Nvidia makes a data center CPU now. Called Vera. Released details so customers can evaluate it. This matters because AMD and Intel apparently needed more competition in a market where their chips already work fine.
The company spent decades dominating GPUs for gaming and AI training. Printed money doing it. But that wasn't enough. Had to build a CPU too. Had to challenge Intel and AMD on their home turf. Because when you're winning, the smart move is always to pick a new fight in someone else's weight class.
Prospective customers now have the information they need to fully evaluate Vera. What a gift. They can finally compare clock speeds and core counts and thermal envelopes. They can build spreadsheets. Run benchmarks. Schedule meetings about whether to switch from the chips that already run their f*cking data centers just fine.
The retail crowd will love this headline. They'll read "next-generation" and "AI" in the same sentence and feel something stir. That's not insight. That's Pavlovian conditioning. Ring the bell. Watch them salivate. Tomorrow they'll forget Vera exists and chase whatever new three-letter acronym some VP mentions on an earnings call.
AMD and Intel should be terrified. Nothing scarier than a competitor entering your market after you've spent forty years building manufacturing expertise, customer relationships, and software ecosystems. I'm sure they're drafting surrender terms right now. Preparing white flags. Updating their résumés.
Vera probably works. Probably runs code without catching fire. Probably costs more than the alternatives and requires customers to rearchitect their entire stack to support it. But it has Nvidia's name on it, so analysts will write breathless pieces about disruption and market share battles that won't matter for five years, if ever.
Your portfolio doesn't care which CPU runs the data center. The data center barely cares which CPU runs the data center.
Photo by on Unsplash
Leave a Comment