Moxin builds sparse computing acceleration cards that optimize AI workload performance and efficiency.
$140M
Investor undisclosed
A $140M raise for a Chinese sparse computing accelerator in May 2026 signals that inference optimization—not just training—is now a capital-intensive category. Moxin likely uses this to scale manufacturing and build out software/compiler support, which is the real moat in accelerators. If you're building any AI infrastructure that touches model serving or edge deployment, watch whether sparse compute becomes table-stakes for your cost model.