PicoJool, which is developing AI data center interconnects based on vertical cavity surface emitting lasers, raised a $27.5M Series A led by Socratic Partners
THE SO WHAT
Optics startups like PicoJool going after AI data center interconnects with VCSEL-based links are a bet that bandwidth, not flops, is the next hard constraint on cluster scaling. If you’re planning multi-rack or multi-site AI deployments, network topology and interconnect roadmap now belong in your capacity planning, not as an afterthought.
READ THE SOURCE
MORE FROM THE WIRE
Deep & Emerging TechGoogle is launching its AI chips into space — Project Suncatcher MVP sports four TPUs and will be delivered by a SpaceX rocket
Running TPUs on-orbit is a step toward space-native AI compute—processing EO data, comms, and autonomy workloads without hauling everything back to Earth. If you build for geospatial, satcom, or defense, start assuming an architecture where inference happens in space and your ground stack is orchestration, not raw processing.
Deep & Emerging TechSebastian Hassinger (The New Quantum Era): why neutral atoms lead the qubit race for now
If neutral atoms are leading the qubit race, the near-term advantage goes to teams architecting around high qubit counts and flexible connectivity rather than just gate fidelity. For operators, the move is to treat quantum like a heterogeneous compute bet—track modality roadmaps now so you’re not locked into the wrong integration path later.
Deep & Emerging TechBitget CEO Gracy Chen says she suspects North Korean attackers exploited a backend system used to process wallet transactions to drain $387M from the platform
A $387M drain via a wallet-processing backend is a reminder that the real crypto risk surface is orchestration logic, not just keys and cold storage. If you touch digital assets, threat-model every internal automation path as if a state actor is probing it daily.
Deep & Emerging TechGoogle’s India AI hub cleared for 2.51GW, not 1GW, the Guardian reports
A single AI campus in Andhra Pradesh cleared for 2.51 GW—over 2.5x the initially announced 1 GW—confirms hyperscale AI is now a grid-scale power planning problem, per [The Next Web](https://thenextweb.com/news/google-india-ai-datacentre-andhra-pradesh-land). If AI is material to your roadmap, treat power, water, and local politics as first-class dependencies alongside GPU supply.