Nvidia Agrees to Buy Hugging Face for Nearly $13 Billion

By Carlos Montiel | Enterprise AI Specialist
Leer en español →
Published: 2026-09-02 | By: Carlos Montiel | Reading time: ~5 minutes

According to reports from late August 2026 confirmed by TechCrunch, CNBC, and Forbes, Nvidia agreed to buy Hugging Face — the central hub where thousands of teams download open models, datasets, and libraries — for a figure close to $13 billion. The deal isn't formally signed yet, but it marks the biggest structural move of the week in the AI infrastructure market.

What's known about the deal

According to CNBC, Nvidia agreed to buy Hugging Face for $12.9 billion; Forbes and Slashdot cite the same figure rounded to $13 billion, and TechCrunch reported that negotiations were closing in late August. That's a massive jump from the $4.5 billion valuation Hugging Face carried in its Series D round back in August 2023. Worth noting the caveat that shows up in several of these reports: no signed agreement existed at the time of coverage, and the talks could still fall through.

Why Nvidia wants the hub, not just the chips

Hugging Face isn't a model or a chip — it's the distribution layer where developers publish, discover, and download open models, on top of offering managed inference and deployment infrastructure. Buying it gives Nvidia two things its GPUs alone don't: control over the entry point where millions of developers choose which model to run, and a direct foothold in the hosting and cloud business, territory where it has so far depended on third parties. According to Business Insider (cited in Forbes' coverage), Hugging Face had also held prior talks with Microsoft before this offer.

The deal in numbers: - Reported amount: ~$12.9-13 billion - Prior valuation (Series D, Aug. 2023): $4.5 billion - Status as of late August 2026: reported deal, not signed - Other reported suitor: Microsoft (prior talks)

What this means for a company that runs open models

If your team downloads model weights, uses Inference Endpoints, or hosts datasets on Hugging Face Hub, this is a vendor-concentration risk signal worth watching, not ignoring. Nvidia already dominates the hardware most AI models run on; if it also controls the ecosystem's most-used distribution hub, the neutrality of that layer — today perceived as independent of any chipmaker — comes into question. That doesn't mean access changes tomorrow, but it is a good time to audit how much of your model supply chain depends on a single hub and a single compute vendor.

What to do while the deal gets confirmed: keep a plan B for where to get model weights and where to run inference that doesn't depend exclusively on the Hugging Face + Nvidia combo — your own registries in S3/GCS for the models you run in production, and at least one alternative inference provider evaluated (AWS Bedrock, Vertex AI, or self-hosting with vLLM). This isn't panic, it's business-continuity hygiene in the face of real market consolidation.
Carlos Montiel
Enterprise AI Solutions Architect
LLMs, Agents & Orchestration Specialist
guatemalia.com/en/#contacto · info@guatemalia.com

Need to implement AI at your company?

Carlos Montiel is an enterprise AI solutions architect. He implements LLMs, Agents, RAG, and orchestrators for companies in Guatemala and Latin America. Reach out for a consultation.

Contact Carlos Montiel

info@guatemalia.com