Key Takeaways

  • NVIDIA projects 70% revenue growth through the fiscal year ending January 2028, with compute sales still strictly supply-constrained.
  • The pending $12.9 billion Hugging Face acquisition exists to crush model software margins so customers spend their budget on silicon instead.
  • Rory O'Driscoll sees only one threat to the chip giant: a sudden drop in end-user application demand across enterprise software.
  • Jason Lemkin argues enterprise tech will stay healthy as long as hardware capex numbers remain positive.

The Real Reason NVIDIA Bought Hugging Face

NVIDIA is pulling off classic platform strategy: commoditize your complements. When you sell the world's most expensive compute, every dollar your customer spends on model licenses or proprietary software layers is a dollar they cannot spend on your GPUs.

O'Driscoll captured the logic in plain terms: “man making $120 billion a year selling compute decides to buy a company that helped makes compute more cost-effective so he can sell more compute that is the summary of the deal.” By acquiring Hugging Face for $12.9 billion, NVIDIA gains control of the primary hub for open-weights models. Making open-source models faster and cheaper to run directly shrinks the margins of closed-source model providers while expanding total GPU consumption across the industry.

O'Driscoll was blunt about the economic motive: “If you're selling GPUs, you want everyone else's margin to be lower. So yours can be higher. So it's just exactly right and rational.”

The Only Risk That Can Break the Run

Hyperscalers are pouring hundreds of billions into data centers because they believe application revenue will follow. Lemkin noted the mood among software founders: “broadly speaking, if Nvidia is crushing it, everyone's going to crush it. Like, everything's green.” Hardware sales act as the primary sentiment indicator for the entire venture market.

Yet the supply-demand balance remains precarious. NVIDIA sells every chip it can produce, and O'Driscoll predicts “this intense demand for compute is going to continue for at least another 12 months.” The danger lies downstream. If enterprise buyers fail to see clear productivity gains from their software subscriptions, hyperscalers will eventually halt their cluster expansion.

As O'Driscoll observed: “Really, the only thing if you step back that can go wrong at some point, and it's not today, is end user demand because all this is predicated in the end.” The hardware boom buys the software industry roughly four quarters to generate real user revenue before infrastructure budgets contract.

What to Do With This

Audit your product pricing model by Friday afternoon. If your gross margins depend on reselling access to base models, your margin will trend toward zero as NVIDIA pushes open-source distribution. Rebuild your defensibility around proprietary workflow state, private customer data integrations, and deterministic business logic rather than pure model inference.