Chips, compute infrastructure, manufacturing, and physical systems.

Sovereign AI funds have bet hundreds of billions on the assumption that compute stays scarce and expensive. A Dutch startup with a 100x efficiency claim and some of the most credible names in semiconductor history just made that assumption look less certain.










Chinese physicists hit 63K in nickel superconductors and showed why the material works. The harder question: whether the world actually needs zero-resistance power lines as much as the pitch suggests.
"The rover was complete. The launch vehicle was gone."
The Artemis II crew said Orion flew better than the simulator. That sounds like a win. The history of NASA programs says otherwise: Apollo, Skylab, and Constellation all died after they worked.
UK scientists bolted fluorescent biosensor worms to the ISS hull for 15 weeks — the real bet isn't the biology, it's whether a streaming radiation readout can replace months of waiting for returned samples.
FCC approved SpaceX satellite phone service calling it a public safety win — the 911 call gets through. But the dispatcher does not know where you are. The infrastructure that was supposed to fix that stopped working in 2020.
China just tested proximity operations with its Qingzhou cargo spacecraft 5 km from a target satellite. No international framework governs that capability — and Beijing is developing it faster than anyone is writing the rules.
TakeMe2Space has 17 employees and a cubesat bolted to its launch vehicle. To make a 50kW orbital data center pencil out, launch costs must drop 17x — from $3,600/kg to $200/kg. Starship is the bet. The $5M seed round is not the capital stack for that bet.
AI labs are hitting a power wall — and the physics of heat is proving harder to solve than the software.
Taiwan wants democracies to trust it with their military communications. The problem: the island’s semiconductor fabs — the reason the world cares about Taiwan — sit within artillery range of the mainland it must defend.
NASA certifies spacecraft materials with Earth fire tests — but nobody has burned anything on the Moon to verify those tests work at 1/6g. An experiment to answer this launches late 2026, after crewed landings. Lunar gravity may let fires burn far longer than on Earth.
ADEOS-II died 10 months after launch in 2003. Nobody knows why. In 2027, a Japanese inspection spacecraft will fly within meters of the wreckage — and the answers have been sitting there for 23 years.