Wafer raises $40M Series A to automate AI inference optimisation
What's the deal? WaferDealroom has a profile for this one. Try Dealroom → has raised a $40 million Series A at a $200 million valuation, co-led by Marathon Management PartnersDealroom has a profile for this one. Try Dealroom → and ChemistryDealroom has a profile for this one. Try Dealroom →. Wing Venture CapitalDealroom has a profile for this one. Try Dealroom →, AMD VenturesDealroom has a profile for this one. Try Dealroom →, Outset Capital, Fifty Years, and Y Combinator also joined, alongside existing investors.
What's the endgame? Wafer builds tools that automate inference optimisation, the process of tuning how AI models run in production. Its system learns from a workload's traffic patterns and constraints, then searches for the best deployment across model, engine, kernels, and hardware.
Why now? Most inference optimisation is still manual, service-heavy, and done once before deployment, according to the company. Wafer says the capital will help it automate more of that loop so each deployment improves performance per dollar continuously.
The angels: The round drew a long list of individual backers, including GoogleDealroom has a profile for this one. Try Dealroom →'s Jeff DeanDealroom has a profile for this one. Try Dealroom →, Vercel chief executive Guillermo RauchDealroom has a profile for this one. Try Dealroom →, DoorDash chief technology officer Andy Fang, Notion chief operating officer Akshay Kothari, and Cloudflare chief executive Matthew Prince.
The M&A angle: Wafer, which positions itself as an inference provider using non-Nvidia chips, has also fielded acquisition offers, per reports tied to its $200 million-plus valuation. The startup counts chipmaker AMDDealroom has a profile for this one. Try Dealroom → among its investors through AMD Ventures.
The signal: Wafer's $40 million round ranks in the 95th percentile of all Series A deals in its sector and geography, drawn from a sample of 5,233 rounds. The size and investor roster point to strong appetite for tools that squeeze more value from AI hardware as inference costs mount.
Read more: linkedin.com
Image credit: Wafer
