Wafer sells what it calls continual inference: a serving stack it keeps re-tuning per workload across kernels, engine and batching, on AMD as well as Nvidia accelerators (site); it raised a $40M Series A on September 1, 2026 (announcement). In this corpus it appears once: the July 4 item in which it served GLM-5.2 on AMD’s MI355X at 2,626 tokens per second per node for less than half Blackwell’s cost, arguing that the kernels agents now write are erasing the software moat around Nvidia, a market-structure reading of the recursive-self-improvement loop that began with a model topping KernelBench.
Organization
Wafer
Inference startup serving open models on AMD accelerators, which argued that AI agents writing GPU kernels are closing AMD's software gap with Nvidia.
Sources
Attribution and reporting linked by this record; links do not imply independent verification.
- Welcome to July 4, 2026 (theinnermostloop.substack.com)
Development timeline
Newest reporting dates first. Reporting dates may differ from event dates.
Knowledge graphopen in browser →
Cite or correct this record
Suggest a correction — opens a GitHub draft for you to review and submit.