QuantumSpark raises $250M to commercialize photonic AI accelerators

AI · 6 min read

QuantumSpark's photonic accelerator uses integrated silicon photonics to perform matrix multiplications with lower energy per operation compared with traditional GPUs. The architecture targets inference and certain transformer workloads where throughput and power efficiency are paramount.

The massive Series C was led by DeepCompute Partners and includes strategic investments from hyperscalers interested in greener AI infrastructure. Funds will expand fabrication partnerships, create SDKs for major ML frameworks, and support pilot deployments inside select data center customers.

QuantumSpark acknowledges remaining engineering challenges — such as programming model maturity and thermal management — but argues that the energy savings and scaling prospects justify rapid commercialization. Early benchmarks show promising energy-per-inference improvements on targeted workloads.