Listen: What is Nvidia up to buying Hugging Face and Groq?

Nvidia is consolidating its grip on the artificial intelligence landscape with a series of massive acquisitions, including buying the open-source platform Hugging Face and the chipmaker Groq. By bringing Hugging Face into the fold, Nvidia can funnel open-source software distribution directly toward its own hardware and proprietary tools.

With the twenty-billion-dollar acquisition of Groq, Nvidia tackles the physical bottlenecks of generating AI text. Standard graphics processing units, or GPUs, rely on high-bandwidth memory, which requires pulling data back and forth across a bus. Groq’s specialized language processing units place ultra-fast static random-access memory directly on the chip itself. This eliminates the delay, allowing for incredibly fast token generation. Nvidia plans to combine these technologies, using traditional GPUs for initial processing and Groq's hardware for speedy delivery.

Yet, no matter how these chips are designed, the entire industry rests on a highly concentrated supply chain. Nvidia, Groq, and competitors like Google all rely on Taiwan Semiconductor Manufacturing Company, known as TSMC, to fabricate their chips. In turn, that fabrication depends on extreme ultraviolet lithography machines made exclusively by the Dutch company ASML.

Strict trade restrictions prevent ASML from exporting these advanced machines to Chinese chipmakers. As a result, Chinese manufacturers must rely on older, less efficient techniques, while their domestic efforts to build equivalent lithography tools remain years away from commercial reality.