Could open-source hardware reduce dependency on dominant chip makers?

The Theoretical Case for Reduced Dependency

The logic behind why open-source hardware could reduce dependency on dominant chip makers is fairly straightforward: if chip architecture designs are freely available rather than locked behind proprietary licenses controlled by a small number of companies, more organizations gain the ability to design their own chips without needing to negotiate licensing terms with, or rely entirely on, those dominant suppliers. This could, in principle, foster a more diverse and competitive chip design landscape, reducing the market concentration that currently characterizes the most advanced AI chip supply.

This dynamic has some precedent in other areas of computing, where open architectures have enabled a broader range of companies to participate in chip design without each needing to either license a proprietary architecture or build an entirely separate one from scratch.

Where This Has and Hasn’t Materialized Yet

In practice, the degree to which open hardware has actually reduced dependency on dominant chipmakers varies significantly depending on the specific computing category in question. For certain less performance-extreme applications, including some embedded systems and specialized processing tasks, open hardware architectures have gained meaningful traction and provided real alternatives to proprietary designs, giving companies more flexibility and reducing reliance on any single dominant supplier for those particular use cases.

For the most advanced, cutting-edge AI training chips specifically, however, dependency on a small number of leading proprietary suppliers remains largely intact. These companies have built significant advantages over years of concentrated investment in performance optimization, manufacturing partnerships, and software ecosystem support that make their chips considerably easier to use effectively for large-scale AI development, advantages that open alternatives haven’t yet closed for this particular category of hardware.

Manufacturing Is a Separate, Persistent Bottleneck

It’s important to recognize that even if open chip designs become more capable and widely adopted, this alone wouldn’t fully solve the dependency problem, because chip manufacturing itself is a separate bottleneck. Actually producing physical chips, whether based on an open or proprietary design, requires access to advanced semiconductor fabrication facilities, and this manufacturing capacity is concentrated among a relatively small number of companies globally. Reducing dependency on chip designers doesn’t automatically reduce dependency on chip manufacturers, which is a distinct part of the overall supply chain that open hardware designs alone don’t directly address.

Bottom Line

Open-source hardware could, in principle, reduce dependency on dominant chip makers by enabling more companies to design their own chips using shared, freely available architectures. This has happened to a meaningful degree in some computing categories, but for the most advanced AI training chips specifically, dependency on leading proprietary suppliers remains largely intact, and manufacturing capacity concentration represents a separate, persistent bottleneck that open hardware designs alone don’t resolve.

Go deeper

• What Does ‘Open-Source AI Model’ Actually Mean?
• What’s the Difference Between ‘Open-Source’ and ‘Open-Weight’ AI Models?
• What Are the Advantages of Open-Source AI Models Over Closed Ones?