Is a specialized chip about to challenge Nvidia’s dominance?

Etched is building a chip intended only for a process called "inference." In the world of AI, inference is the moment when the software takes what it has learned and puts it into action, like when it writes a response to your question or classifies an image you uploaded. While most AI today runs on powerful graphics chips traditionally used for video games, Etched claims that its hardware is built exclusively to run AI models—the complex mathematical systems that power tools like ChatGPT—at a much higher speed.

Think of modern AI chips like a luxury sports car that can handle almost anything—office work, racing, or off-roading. They are very powerful, but they are "general purpose," meaning they carry extra parts to handle tasks they don’t actually need for AI. By stripping away those unneeded features, Etched is trying to create a specialized engine that does one thing: calculating the next word in a sentence or the next pixel in an image. Because the chip doesn't have to carry the "weight" of features used for gaming or high-end graphics, the machine can dedicate all its power to the specific math required for AI, potentially making the process faster and cheaper for the companies running these services.

This $1 billion in sales is a signal that the AI industry is entering a new phase. In the beginning, researchers used whatever hardware they had on hand. Now that AI is being built into the tools we use every single day, the cost of running that software—the electricity, the equipment, and the maintenance—is becoming a massive burden for businesses. If specialized chips can handle these tasks more efficiently than the standard hardware currently in use, it could lower the cost of running AI services significantly. The big question is whether one specialized design can keep up with how quickly AI is evolving, or if we will continue to need the "sports cars" that can adapt to whatever comes next.