The world is so AI-crazy right now that software companies are investing in nuclear power to fuel our demand for generating text, images, and videos. But what if they don’t need to? What if we could have our AI cake and eat it too, with 90% higher efficiency so we can do more at lower cost and much lower energy use?
That’s the pitch by Microsoft-backed chip startup d-Matrix. The core idea: building chips that offer much faster, cheaper, more efficient inference, which is what AI companies do when answering your AI queries. If d-Matrix is right, AI’s future may not hinge on who trains the biggest models. It may depend on who can answer fastest and cheapest.
