Стартап по ИИ-выводам Wafer завершил раунд финансирования серии A на сумму 40 миллионов долларов США с оценкой, превышающей 200 миллионов долларов США.

Beating AI News: Wafer, an AI inference startup with only 8 team members, has completed a $40 million Series A round, valuing the company at over $200 million. Previously, the company turned down acquisition offers from multiple cloud providers and inference service vendors. Wafer achieved $8 million in annual recurring revenue (ARR) from its inference business in just three months. It does not manufacture chips; instead, it helps models run faster and cheaper on existing hardware. Wafer enables AI agents to automatically adjust models, inference engines, kernels, caching, quantization, and scheduling, then find optimal configurations for different hardware such as NVIDIA and AMD. Previously, much of this work required manual tuning by inference performance engineers—Wafer aims to automate this with AI. In its own tests, GLM 5.2 running on AMD MI355X achieved approximately 80% of the throughput of NVIDIA B200 at less than half the cost. Current clients include Vercel and Inworld AI. Vercel has also independently published test results, stating that Wafer delivers approximately twice the throughput of GLM 5.2 compared to other serverless providers. After closing the funding round, Wafer immediately began expanding its team. It has recently opened positions in four areas: engineering, growth, CEO’s office, and GTM—all requiring full-time on-site presence in San Francisco five days a week. Engineering roles offer a base salary of $250,000 plus equity.