An autonomous laboratory building superconducting computers.

Why superconducting

The deepest layer of the future datacenter is superconducting.

In the datacenter of the future, AI routes its hardest workloads — materials and chemistry, quantum dynamics, optimization — down to a fault-tolerant quantum computer.

Superconducting computers make this possible. They are integrable with silicon datacenters and manufacturable in 300 mm foundries; they run fast, consume ultra-low power due to lossless superconducting interconnect and sub-aJ switching energies. The same platform carries classical superconducting logic and quantum logic together — one integrated substrate from control to qubit.

2026 — a structural shift

Superconducting hardware is going fabless.

Pure-play 300 mm foundries now offer superconducting fabrication as a service, splitting the stack the way TSMC and NVIDIA split silicon. GiL owns design and test; the foundry owns fabrication.

• ✓Faster design cycles — weeks, not months.
• ✓Access to multiple mature processes, not one in-house fab.
• ✓Focus on intelligence and learning, not fabrication capex.

The problem

Scaling is the shared bottleneck.

Superconducting computing is where semiconductors were in the 1960s — artisanal, physics-led, one experimental cycle at a time. A single design → fabricate → cool → measure → diagnose loop takes weeks to months.

Closing the ~1,000× gap to a fault-tolerant machine takes thousands of compounding cycles.

Our solution

A superconducting world model.

GiL builds one world model across the whole stack — materials, junctions, devices, circuits — that learns how fabrication becomes computer performance. Our AI is trained at scale using proprietary cryogenic metrology.

What we're building

Our first target is fault-tolerant quantum computing.

Our first target is a fault-tolerant superconducting quantum computer. Its performance is set by materials and device physics at the nanometer scale — exactly the intelligence GiL is built to learn.