🤖 OpenAI, Oracle, SoftBank Expand Stargate with 5 New Data Centers
OpenAI, Oracle, and SoftBank just dropped news of five fresh U.S. data-center sites - part of that gigantic (and honestly kind of surreal) $500B Stargate push. We’re talking close to 7 gigawatts of juice lined up now… which is mind-bending when you actually picture it.
They’re saying ~25,000 jobs spin up onsite, and OpenAI is framing this as their “clear path” to hitting 10 GW by year’s end. Feels less like building servers and more like straight-up building power plants to feed AI.
🔗 Read more
💥 Elon Musk’s xAI Sues OpenAI for Alleged Trade Secret Theft
Cue the drama: Musk’s xAI slapped a lawsuit in California on OpenAI, claiming they lured away ex-staffers who allegedly walked out the door with code and infra secrets stuffed in their pockets. Not exactly a minor spat.
The filing even names a recruiter as ring-leader of sorts. Musk basically argues OpenAI is playing dirty pool to trip up his newborn venture. “Messy” might be too polite here.
🔗 Read more
🧱 Nvidia Pledges $100B in Support of OpenAI
Nvidia just went nuclear - $100 billion, yes billion - with both compute muscle and cash lined up for OpenAI’s massive growth binge.
Still, plenty of critics are waving the bubble flag, muttering about circular money loops. It does look a bit like Nvidia juicing the very customer that keeps Nvidia’s valuation so frothy. A snake biting its own tail kind of situation, risky if the hype air leaks out.
🔗 Read more
🇩🇪 SAP & OpenAI Launch “OpenAI for Germany” for Public Sector AI
SAP, OpenAI, and Microsoft (piping it through Azure) are teaming up on “OpenAI for Germany,” a sovereign-cloud setup tailored for public agencies. Imagine German bureaucracy with a GPT upgrade - that’s the pitch.
It kicks off with 4,000 GPUs, scaling up if demand explodes. Selling point is clear: AI that hugs local compliance rules tightly, with zero “oops, your data just hopped across the Atlantic” surprises.
🔗 Read more
📊 Google Releases MCP Server to Access Public Data Commons
Google rolled out something pretty technical but surprisingly important: the MCP Server. It lets AI models pull from Data Commons directly, skipping messy API workarounds.
The idea is to keep models from hallucinating quite as much - grounding them in verified stats instead of, well, vibes and guesswork. A nerdy tool that might quietly become a big deal.
🔗 Read more