Wafer raises $40 million at valuation above $200 million to optimize GPU inference

  • Wafer raised $40 million in a Series A co-led by Marathon and Chemistry.
  • July testing reached 80% of Nvidia B200 throughput on AMD’s MI355X below half-cost.
  • Wafer says production GPUs average 20% utilization and claims 2x to 2.8x speedups.

Wafer has raised $40 million in a Series A at a valuation above $200 million, about five months after closing a $4 million seed round in April 2026. The San Francisco-based startup, led by co-founder and CEO Emilio Andere and co-founder Steven Arellano, uses autonomous AI agents to optimize GPU workloads across chips and model architectures. In July 2026, Wafer fine-tuned Z.AI’s GLM-5.2 model on AMD’s MI355X, achieving approximately 80% of Nvidia’s B200 throughput at less than half the cost. Wafer says its system delivers 2x to 2.8x speedups over stock baselines across various open-source models, while production GPUs average about 20% utilization. The Series A was co-led by Marathon and Chemistry, according to the latest account; existing coverage identified AMD Ventures and Outset Capital among participants and described Marathon Management Partners as a co-lead. Backers include Google’s Jeff Dean and OpenAI’s Wojciech Zaremba, while earlier investors Fifty Years and Y Combinator increased their support. Wafer is targeting the software gap around Nvidia’s CUDA ecosystem through automated optimization for AMD’s ROCm stack and other heterogeneous computing environments.

The information on this website is generated using AI and we cannot guarantee its accuracy. Please use it as reference information only.