AI training
Transformers, diffusion models, mixture-of-experts. The workload that pulled accelerators from the gaming aisle into the enterprise.
foundation models→
Transformers, diffusion models, mixture-of-experts. The workload that pulled accelerators from the gaming aisle into the enterprise.
Climate, fluid dynamics, molecular docking. Equations that used to need weeks of supercomputer time fit in a single overnight run.
The original use case. Real-time path tracing, neural denoising, virtual production for film and games.
Sequence alignment, protein folding, generative chemistry. Pipelines that produce candidate molecules in hours, not months.
Monte Carlo at scale, derivatives pricing, factor models. Risk teams now re-run portfolios intraday.
NPUs in phones, cars, cameras. Low-power accelerators that run a model in milliseconds without ever leaving the device.