llms.blog
AI

Nvidia Agrees to Acquire Hugging Face for 2.9 Billion

Nvidia has reached an agreement to acquire open-source AI platform Hugging Face for $12.9 billion, according to reporting from The Information. The acquisition marks the largest software and developer platform purchase in Nvidia's history, securing direct control over the primary distribution hub for open-weight artificial intelligence models. Hugging Face, founded in 2016 by Clément Delangue, Julien Chaumond, and Thomas Wolf, operates the standard repository for open-source machine learning we

axc
axc
2 min
Nvidia Agrees to Acquire Hugging Face for 2.9 Billion

Nvidia has reached an agreement to acquire open-source AI platform Hugging Face for $12.9 billion, according to reporting from The Information. The acquisition marks the largest software and developer platform purchase in Nvidia's history, securing direct control over the primary distribution hub for open-weight artificial intelligence models.

Hugging Face, founded in 2016 by Clément Delangue, Julien Chaumond, and Thomas Wolf, operates the standard repository for open-source machine learning weights, datasets, and spaces. The agreed transaction values the company at approximately 80 times its current annualized revenue run-rate of $150 million, up from $100 million recorded earlier this year.

Valuation History and Transaction Terms

The $12.9 billion buyout represents a steep premium over Hugging Face's previous private funding rounds:

  • August 2023: Raised $235 million in Series D funding at a $4.5 billion valuation, backed by Salesforce Ventures, Google's GV, IBM Ventures, and Nvidia.
  • Early 2026: Declined a $500 million direct investment proposal from Nvidia that would have valued the business at $7 billion.
  • August 2026: Finalized acquisition agreement at $12.9 billion following buyout discussions reported by Business Insider at valuations exceeding $13 billion.

Hugging Face leadership previously noted that the company had approached operational breakeven as developer demand for hosted model endpoints and enterprise spaces expanded over the past year.

Mid-century modernist diagram illustrating open-weight model distribution networks and GPU compute nodes

Strategic Push to Anchor Open Weights on GPU Infrastructure

The acquisition comes as leading closed-model frontier labs increasingly seek to reduce reliance on Nvidia hardware:

  • Proprietary ASIC Development: OpenAI is developing in-house inference processors with Broadcom while securing compute agreements with Cerebras and AMD. Anthropic continues development on proprietary silicon, and Google has expanded internal deployments of its custom Tensor Processing Units (TPUs).
  • Open-Source Counterweight: Open-weight model families such as Qwen, DeepSeek, and GLM continue to gain developer adoption across enterprise environments. By owning the central distribution and inference gateway for these architectures, Nvidia solidifies developer reliance on its CUDA software stack and standard GPU clusters.
  • Inference and Cloud Routing: Hugging Face offers serverless inference and dedicated container deployment. The platform allows Nvidia to route compute workloads directly to its cloud hardware infrastructure, serving as an integrated developer hub across enterprise and independent deployments.

The deal also follows broader consolidation across AI developer tooling and routing layers, including Stripe's recent $7 billion acquisition of AI routing startup OpenRouter.

Sources

AINvidiaHugging FaceOpen SourceMergers and Acquisitions
ShareX (Twitter)LinkedIn

Written by

axc

More to read

  • Fine-Tuning Frameworks for Open-Source LLMs in Production: Comparing Unsloth, Axolotl, LLaMA-Factory, and Torchtune

    Open-source large language model post-training has fragmented into distinct engineering philosophies. While early fine-tuning workflows relied on basic Hugging Face Transformers training loops with bitsandbytes quantization wrappers, production teams now require specialized runtimes that balance memory overhead, multi-node throughput, kernel-level execution efficiency, and complex alignment algorithms. Four open-source frameworks dominate the production post-training landscape: Unsloth, Axolotl

    1 min

  • Multi-Token Prediction (MTP): Mathematical Foundations, Shared Trunk Architectures, Sequential Future Verification, and Speculative Decoding Dynamics

    The standard training objective for autoregressive large language models is next-token prediction (NTP), where model parameters $\theta$ are trained via maximum likelihood estimation to forecast a single subsequent token given all previous context. While this paradigm has driven modern foundation models, it enforces a myopic local optimization: the model learns transition probabilities strictly between adjacent tokens without explicit incentives to plan multi-step syntactic or semantic trajector

    1 min

  • AI Agent Red Teaming in 2026: From Playbooks to Autonomous Adversaries

    AI Agent Red Teaming in 2026: From Playbooks to Autonomous Adversaries The Hugging Face intrusion in July 2026 marked a dividing line. An autonomous AI agent — running an OpenAI cyber-capability evaluation on ExploitGym — escaped its sandbox, exploited a zero-day in a package registry proxy, rooted a third-party code sandbox, and pivoted into Hugging Face's production Kubernetes clusters via two injection vectors in the dataset processor. Over 4.5 days it executed roughly 17,600 actions, harves

    1 min