Latest in quantum · updated quarterly, alongside The Intelligence Wars
For ten years the honest answer to whether fault-tolerant quantum computing works was “probably.” In the past eighteen months, neutral-atom groups at Harvard, MIT, QuEra and Caltech have published the core ingredients in Nature: logical magic-state distillation, a universal fault-tolerant architecture running below threshold on up to 448 atoms, and continuous operation of coherent 3,000 and 6,100-atom arrays.
Every ingredient now exists on at least one modality. Nobody has yet put them all in one machine. The question is no longer whether FTQC is possible. The question is how to make enough of it.
The conclusions
Most capital, highest valuations, longest horizon. We hold a strong presumption against it, with breakthrough exceptions only. The chip wars are over. The interconnect wars started in 2025.
Oratomic raised $300M three months out of stealth on a paper and a founding roster drawn from Harvard and Caltech. In this industry the org chart is the cap table, and lab lineage is the least inflated diligence signal left. It is one of our three screens.
Anything that makes fault-tolerant machines cheaper, faster and more resilient gets paid on every modality and every roadmap. The layer that directly tackles today’s bottlenecks is the layer that turns room-sized machines into deployable ones.
Companies solving commercially relevant problems today, on HPC and quantum-inspired methods, already hold the customer and capture FTQC value through a software update. We screen for a defensible approach to adopting quantum computers once they are available.
QuEra’s Libra, announced in June 2026 with AWS, is scheduled for the cloud in 2028: more than 256 error-corrected logical qubits at a 10−6 logical error rate, built from more than 10,000 physical qubits. That is a megaquop machine, about a million reliable logical operations per job. Breaking RSA-2048 is a gigaquop problem, thousands of times larger. The first slots go to chemistry, materials and many-body simulation. The migration deadlines do not wait for the rest.
