Nvidia teams up with chip rival d-Matrix instead of fighting it
THE SO WHAT
Nvidia pairing its GPUs with d-Matrix inference chips for a joint system is a pragmatic admission that heterogeneous AI compute is the new normal. Buyers should expect more mixed-architecture offerings and design their stacks to be modular at the rack and software layer, not locked to a single accelerator type.
READ THE SOURCE
MORE FROM THE WIRE
Deep & Emerging TechWith EU backing, QuantumDiamonds aims to speed up chip manufacturing
Europe is trying to replicate the ASML effect by seeding upstream process innovation — QuantumDiamonds is a bet that metrology and inspection, not just lithography, are now strategic levers. If you rely on EU fabs for AI hardware, track these tool-chain bets as closely as you track new nodes.
Deep & Emerging TechSqueezed on land, Samsung wants to put data centres out to sea
Samsung Heavy Industries planning floating data centers by 2028 is a signal that compute growth is running into land, water, and neighbor limits—offshore is becoming part of the capacity mix. For infra planners, that means future availability zones may be tied to maritime regulation and coastal grid interconnects, not just real estate.
Deep & Emerging TechA £35bn plan to fill the UK with 14 mini nuclear reactors
A £35 billion plan to deploy 14 BWRX-300 small modular reactors in the UK by 2034 is a direct bet that AI- and chip-driven power demand will outgrow current grids. Large compute buyers and data center operators in Europe should factor nuclear-backed PPAs and SMR-adjacent sites into their 10–15 year location strategy.
Deep & Emerging TechVenus Aerospace raises $91M to build its ‘detonation’ rocket engine
A $91 million Series B for a rotating detonation rocket engine — long viewed as too hard — is a bet that propulsion efficiency is now a strategic edge, not just a science project. Space and defense operators should expect more exotic engine architectures to move from papers into procurement conversations.