Infinity Raises $15M for AI Inference

It's telling that researchers from the big labs like OpenAI and Anthropic are putting their own money into Infinity. Usually, when the people actually building the frontier models invest in infrastructure, it means there's a specific bottleneck in inference that the giants haven't solved internally yet.

While NVIDIA owns the hardware layer, the "inference gap"—the cost and speed of running these models at scale—is where the real war is being fought right now. A $100 million valuation for a company focusing on the plumbing of AI shows that investors are shifting focus from who can train the biggest model to who can serve it most efficiently. If Infinity can actually optimize the inference pipeline, they'll be indispensable regardless of which LLM wins the market.