🏙️ OpenAI to make London its biggest research hub outside US ↗
OpenAI says London is being elevated into its largest research base outside the US - a fairly direct signal it wants deeper roots in the UK talent pool (and yes, right on DeepMind’s doorstep).
Details are still a bit cagey - no big numbers on headcount or spend - but the tone reads as “serious expansion,” not “cute satellite office.”
🧠 Meta signs multi-billion-dollar deal to rent Google AI chips, The Information reports ↗
Meta is reportedly renting Google’s TPUs on a multi-year, multi-billion-dollar deal - which is both practical and faintly hilarious, given how these companies usually posture like sworn enemies.
It’s part of that bigger scramble to diversify away from Nvidia dependency - not ditching Nvidia, just… hedging, loudly.
🤖 Google takes control of “Android of robotics” project in quest for physical AI ↗
Google is pulling Intrinsic back into the main company, basically saying: robotics isn’t a side quest anymore, it’s core plot. Intrinsic was pitched as an “Android for robotics” platform - a software layer to make building robot apps less of a bespoke nightmare.
Now it’ll sit closer to DeepMind and Gemini, which should speed things up - or at least make the org chart slightly less contorted (or so it seems).
🛑 Anthropic refuses Pentagon's new terms, standing firm on lethal autonomous weapons and mass surveillance ↗
Anthropic publicly pushed back on new Pentagon contract terms, drawing hard lines around mass surveillance and fully autonomous lethal weapons. It’s one of those “we support defense, but not like that” stances - principled, and also a massive business gamble.
The tension here isn’t subtle: government wants broader access, Anthropic wants guardrails, and both sides seem ready to play chicken. It reads like watching two drivers argue over who gets to invent the steering wheel.
🧩 Broadcom expects to sell 1 million 3D stacked chips by 2027 ↗
Broadcom is betting big on 3D-stacked chips - stacking dies to boost data movement and cut power draw, which is basically catnip for AI compute. The company says the demand curve is steep enough that it expects at least a million of these units sold within its forecast window.
What’s quietly intriguing is how this positions Broadcom as more than “just” a networking-and-custom-silicon player - it’s edging into the “serious AI infrastructure kingmaker” lane.
FAQ
What does it mean that OpenAI is making London its biggest research hub outside the US?
It signals OpenAI is aiming for a deeper, long-term presence in the UK, not a modest satellite footprint. Positioning London as its primary non-US research base implies a meaningful concentration of research work and leadership there. The article notes that headcount and spend weren’t disclosed, but the intent reads as a serious expansion.
Is OpenAI’s London research hub expansion just about hiring, or something bigger?
From what’s shared, it’s framed as more than incremental hiring. The announcement casts London as a major research base, pointing to broader investment in teams, projects, and institutional presence. With no specific numbers provided, the safest takeaway is directionality: OpenAI is prioritizing UK talent and proximity to major AI ecosystems, including nearby competitors.
Why would Meta rent Google TPUs instead of relying only on Nvidia GPUs?
The report frames it as a hedging move rather than a clean break from Nvidia. Renting Google’s TPUs can diversify compute supply and reduce reliance on a single vendor during heavy AI buildouts. In many pipelines, mixing accelerators is a pragmatic way to balance cost, availability, and capacity planning, especially while demand for top-end chips remains intense.
What is Intrinsic, and why did Google pull the “Android of robotics” project into the main company?
Intrinsic was pitched as a platform layer meant to make robotics software less bespoke, using an “Android” style approach for robot applications. Google bringing it back into the core organization suggests robotics is shifting from an experimental side project toward a central priority. The move also places it closer to other major internal AI efforts, which can tighten coordination and speed productization.
What did Anthropic refuse in the Pentagon’s new contract terms?
The article says Anthropic pushed back on terms it viewed as enabling mass surveillance and fully autonomous lethal weapons. It framed this as setting firm guardrails while still engaging with defense in other ways. The significance sits in the tension between government requests for broader access or flexibility and an AI company’s desire to limit certain high-risk use cases.
Why are 3D-stacked chips a big deal for AI, and what is Broadcom predicting?
3D stacking can improve data movement and power efficiency by stacking dies, which matters for AI workloads often bottlenecked by bandwidth and energy. Broadcom is betting demand will be strong enough that it expects to sell at least one million of these 3D-stacked chip units by 2027. The implication is faster-moving competition in AI infrastructure beyond traditional roles.