Chips, compute infrastructure, manufacturing, and physical systems.

The EU just bet €211M that a seven-year-old Cambridge spinout can solve the AI data center power problem with graphene — before silicon photonics makers like Ayar Labs and Lightmatter make the question irrelevant.










NASA wants industry to replace the satellite comms network it built for Hubble and the space station. The compatibility specs written into the bid may lock out everyone but the companies that already know the 1980s-era system.
The Space Force paid Turion $32.6M for three satellites before the startup had the capital to build them. Now it has raised $75M more to scale from 8 vehicles a year to 40 — and the government is counting on it.
ESA Proba-3 measured slow solar wind at 250-500 km/s — 3-4x faster than models predicted. The K-corona was supposed to be slow. ASPIICS says otherwise. Space weather forecasting now needs to account for it.
Hyperscalers are designing chips with AI tools at 15-85% higher productivity than traditional EDA vendors can match — and that gap is Nvidia's problem.
A laser pushed on a piece of graphene foam in a lab and barely moved it. Then they did it in microgravity and the thing shot 30 times faster. The reason, researchers say: Earth’s gravity was hiding the real numbers from every lab on the planet.
In agentic AI, the bottleneck is CPU tool processing — causing up to 90.6% of total latency, not GPU inference. Both Arm and Nvidia launched CPUs for this orchestration layer in the same month. The GPU-centric infrastructure model may be structurally incomplete.
SpaceX just test-fired a rocket stage at 9,240 tons of thrust — more than Saturn V produced at liftoff. The V3 architecture targets 200 tons to orbit, vs 35 on the current version. Flight 12 targets May.
VIPER is dead, NASA has no funded replacement for its polar lunar rover, and Zeno Power is building the only radioisotope power system designed for the dark without a NASA contract. PRISM in February 2026 and a 2027 hardware target are on a collision course.
ESA picked Kepler to prove optical relay works around a 10x ground station gap
A galaxy 5 billion light years away bent light from a supernova 9 billion years distant, magnify it 100x, and make it the most studied stellar explosion in cosmic history. Now astronomers need to watch the clock.
The US is building satellites that interpret sensor data and coordinate responses without waiting for ground control. Nobody has written the rules for what happens when they do.