Nvidia's $12.9B Hugging Face bet doubles down on AI ecosystem control

Nvidia announced on September 3 that it would acquire Hugging Face, the platform where millions of developers share artificial intelligence models, for $12.9 billion. The deal is expected to close in the first half of 2027, and Nvidia has committed to keeping Hugging Face open to the entire AI ecosystem—no requirement to use Nvidia's compute to build on it.

The price is striking because it nearly doubles Nvidia's largest acquisition on record. In 2019, Nvidia agreed to buy Mellanox, a data center networking specialist, for $6.9 billion. When that deal closed in April 2020, the total transaction value was $7 billion. Mellanox was a mature, profitable business: in its final year as a standalone company, it generated $1.33 billion in revenue and $205 million in net income. Nvidia paid roughly five times Mellanox's annual sales.

What happened next offers a template for how Nvidia thinks about acquisitions. The company doesn't report Mellanox's results separately anymore, but it discloses data center networking revenue—the line where Mellanox landed. In fiscal 2024, that segment generated $8.6 billion. In fiscal 2025, it hit $13 billion. By fiscal 2026, which ended in January, it reached $31.4 billion. The growth accelerated from 51 percent year-over-year to 142 percent. That single business line now generates more than four times what Nvidia paid to acquire it. Few large acquisitions anywhere have produced returns that clean.

Hugging Face presents a different picture. Founded in 2016, the platform hosts more than 3 million AI models and 500,000 datasets used by over 18 million developers, researchers, and creators. Nvidia is already an investor. Three years ago, Hugging Face was valued at $4.5 billion in a funding round. According to reporting from The Information in August, the company's annualized revenue had recently climbed 50 percent in two months, reaching more than $150 million. At $12.9 billion—which includes up to $1 billion in equity-based retention for employees—Nvidia is paying roughly 86 times annualized revenue. That's a vastly different multiple than the five times sales it paid for Mellanox.

The bet hinges on an indirect payoff, not a direct one. Mellanox succeeded not because it kept growing as a standalone business, but because networking became integral to the AI data center systems Nvidia sells. The theory behind Hugging Face is similar: by owning the platform where developers discover and build with models, Nvidia keeps those developers—and their compute budgets—locked into Nvidia's hardware and software ecosystem. The return comes through chip and system sales, not through Hugging Face's own revenue line.

Context matters here too. When Nvidia announced the Mellanox deal, the company had about $11.7 billion in annual revenue, so the $7 billion price represented nearly 60 percent of a year's sales. Today, Nvidia generated $96.2 billion in revenue and nearly $60 billion in net income in the fiscal second quarter alone. The $12.9 billion for Hugging Face is proportionally smaller—a bet the company can afford to make and absorb if it doesn't work out.

The Mellanox acquisition ultimately looked like a bargain, but only because the payoff ran through Nvidia's own product line, not through Mellanox's independent growth. For Hugging Face to justify its price, it will need the same kind of indirect return. The platform's value to Nvidia lies not in what it earns on its own, but in how it shapes where the next generation of AI developers build and what hardware they choose to build on.