Cloud API simulates up to 1,000-qubit circuits on 4 vCPU / 16 GB RAM using tensor networks and KLT Engine routing, not brute-force statevector.
Key Takeaways
KLT Engine auto-routes each circuit to the most efficient representation: tensor network, MPS, Gaussian covariance matrix, nexus graph, or full statevector based on entanglement structure.
Depth limits are tiered by qubit count: 1-20 qubits supports depth 20; 106-1,000 qubits caps at depth 7 entangling layers; single-qubit gates are uncapped.
Benchmarks show exact agreement with statevector on circuits up to 20 qubits, and matches Google Sycamore 2021/2022 results on MBL time crystal and holographic wormhole experiments.
Drop-in backends for Qiskit (QumulatorBackend) and Cirq (QumulatorSimulator) replace AerSimulator and cirq.Simulator in two lines; OpenQASM 2/3 supported natively.
Free tier: 500 CPU-seconds/month, 100 requests/day, 1 req/min rate limit; paid plans start at $99/month for 10,000 CU.
Hacker News Comment Review
Commenters correctly identified the core caveat: the 1,000-qubit claim applies only to circuits with exploitable structure (low entanglement, short depth) where tensor methods work; general high-entanglement circuits at that scale remain intractable classically.
Comparison to Bluequbit’s free simulator was raised immediately, signaling that technically literate readers will benchmark against existing cloud sim offerings before committing.
Notable Comments
@alyxya: “it doesn’t simulate any general purpose quantum circuit with 1000 qubits, only ones where there’s a more efficient strategy than an exponential state” – key accuracy flag on the headline claim.
