AMD has agreed to acquire Taalas, a Toronto-based startup that takes a fundamentally different approach to AI inference: instead of building a general-purpose chip and running a model on it, Taalas builds the chip around the model. The result is silicon physically wired to run one specific AI model. Financial terms were not disclosed, but AMD confirmed this is a full acquisition.

How Taalas Chips Work

Taalas builds what it calls "Hardcore Models": processors tailored to a single model's weights, produced by finalizing a small number of a chip's metal layers once the model is fixed. Think of it like the difference between a programmable calculator and a circuit board wired to do only one calculation. The latter is vastly faster and cheaper for that one task.

The approach eliminates the biggest bottlenecks in modern AI serving by hardwiring a model's dataflow between compute elements and burning in the weights. Inference runs extremely fast, but switching to a new model means designing a new chip.

That constraint is less punishing than it sounds, because of how Taalas manufactures. The startup assembles a nearly complete chip and customizes only the final two metal layers per model, so TSMC needs roughly two months to finish a chip versus six months to fabricate something like Nvidia's Blackwell from scratch. Taalas emerged from stealth in February with a demo chip achieving more than 16,000 tokens per second per user on Llama 3.1-8B, built by a team of 24 on just $30 million.

The performance claims are striking, though they come from the company's own data:

• 17,000 tokens per second per user on Llama 3.1-8B, roughly 10x faster than current GPU setups
• 20x lower build cost and 10x lower power consumption versus comparable GPU inference
• First-generation chip uses aggressive quantization to a custom 3-bit data type, which degrades output quality relative to GPU baselines