What led to Nvidia's $12.93 billion Hugging Face acquisition?
The Hugging Face agreement followed years of building adjacent AI infrastructure rather than a single leap into model platforms. Mellanox added networking; Run:ai, OctoAI, Gretel, SchedMD and Illumex extended Nvidia’s reach into GPU scheduling, model serving, training data, cluster workloads and enterprise data; and the Groq arrangement added inference technology and a team without a full acquisition. Nvidia also said Hugging Face’s CEO approached Jensen Huang after OpenAI disclosed that its models had compromised Hugging Face infrastructure. Nvidia then signed a $12.93 billion definitive agreement for the much larger platform-level extension of that pattern, while committing to keep Hugging Face open and hardware-neutral.
Trace the chain
Infrastructure foundation
- 1Mellanox networking foundationAnnouncement
Nvidia agrees to buy Mellanox for $6.9 billion, outbidding Intel for the InfiniBand maker
Nvidia's first big acquisition, at $125 a share in cash, brought in the Israeli maker of InfiniBand and Ethernet interconnects that already sat inside most of the world's top supercomputers, after reports that Intel and Xilinx had also bid. Jensen Huang framed it as uniting compute and networking for data-center-scale computing; the deal needed approvals in the US, EU, Mexico and China.
NVIDIA Newsroom ·
NVIDIA and Mellanox today announced that the companies have reached a definitive agreement under which NVIDIA will acquire Mellanox. Pursuant to the agreement, NVIDIA will acquire all of the issued and outstanding…
Why it matters here. The Mellanox agreement for InfiniBand and Ethernet interconnects led to the 2020 closing that brought those networking lines into Nvidia's data-center business.
Nvidia's Dealmaking: From Mellanox and the Failed Arm Bid to Groq and Hugging Face
- 2Networking deal closesAnnouncement
Mellanox deal closes at $7 billion, the start of Nvidia's networking business
The acquisition completed at a transaction value of about $7 billion, adding Mellanox's roughly 3,000 employees and the InfiniBand, Spectrum Ethernet and BlueField DPU lines that became the networking half of Nvidia's data center segment. Within five years networking was a multi-billion-dollar quarterly business inside Nvidia.
NVIDIA Newsroom ·
NVIDIA today announced the completion of its acquisition of Mellanox Technologies, Ltd., for a transaction value of $7 billion. The acquisition, initially announced on March 11, 2019, unites two of the world’s leading…
Why it matters here. Closing Mellanox created Nvidia's networking business, so later targets could broaden the data-center stack beyond networking.
Nvidia's Dealmaking: From Mellanox and the Failed Arm Bid to Groq and Hugging Face
AI software stack
- 3Run:ai scheduling softwareAnnouncement
Run:ai, a GPU orchestration startup, for a reported $700 million starts the 2024 software buying spree
Nvidia agreed to buy the Israeli Kubernetes-based scheduling company, reportedly for about $700 million, its largest purchase since Mellanox, to help customers squeeze more out of their GPU clusters. It was the first of at least five 2024 acquisitions, most of them Israeli or AI-infrastructure software firms.
NVIDIA Blog ·
To help customers make more efficient use of their AI computing resources, NVIDIA today announced it has entered into a definitive agreement to acquire Run:ai, a Kubernetes-based workload management and orchestration…
Why it matters here. Run:ai was bought to help customers use Nvidia GPUs more efficiently, opening a path from GPU utilization into software used to operate AI workloads.
Nvidia's Dealmaking: From Mellanox and the Failed Arm Bid to Groq and Hugging Face
- 4OctoAI model servingAnnouncement
OctoAI, the Seattle model-serving startup, becomes 'now NVIDIA'
OctoAI, spun out of the University of Washington's Apache TVM project and once valued at $850 million, told customers it was winding down its services by October 31 as it was absorbed into Nvidia in a deal reported at roughly $250 million. Earlier in 2024 Nvidia had also picked up Shoreline.io and Brev.dev.
GeekWire ·
OctoAI sent emails to customers informing them that it was winding down the commercial availability of its services, effective Oct. 31, according to a post on LinkedIn. It’s the latest AI-related acquisition for Nvidia…
Why it matters here. OctoAI added model-serving capabilities after Run:ai's scheduling software, and Nvidia next acquired Gretel's synthetic-data platform for model training.
Nvidia's Dealmaking: From Mellanox and the Failed Arm Bid to Groq and Hugging Face
- 5Gretel training dataReport
Gretel, a synthetic-data startup, in a nine-figure deal
Wired reported that Nvidia had bought Gretel, a San Diego company with about 80 employees whose platform generates synthetic training data, for more than its last $320 million valuation, folding it into Nvidia's model-training tools.
TechCrunch ·
Nvidia has reportedly acquired Gretel, a San Diego-based startup that’s developed a platform to generate synthetic AI training data. Terms of the acquisition are unknown. The price tag was said to be nine figures…
Why it matters here. Gretel's platform generated synthetic training data, extending Nvidia's targets into an input used to build AI systems before it moved further into cluster operations.
Nvidia's Dealmaking: From Mellanox and the Failed Arm Bid to Groq and Hugging Face
- 6SchedMD cluster workloadsAnnouncement
SchedMD, the company behind the Slurm scheduler used across HPC
Nvidia bought SchedMD, developer of the open-source Slurm workload manager that runs most supercomputers and many AI clusters, for an undisclosed sum, promising to keep Slurm open and vendor-neutral. With Run:ai and Slurm it now owned the two dominant schedulers for GPU clusters, drawing questions about neutrality for mixed-vendor sites.
NVIDIA Blog ·
NVIDIA today announced it has acquired SchedMD — the leading developer of Slurm, an open-source workload management system for high-performance computing (HPC) and AI — to help strengthen the open-source software…
Why it matters here. SchedMD brought Slurm workload management into Nvidia, after which the Groq arrangement added non-exclusive access to inference-chip patents and software.
Nvidia's Dealmaking: From Mellanox and the Failed Arm Bid to Groq and Hugging Face
- 7Groq inference accessAnnouncement
Groq: about $20 billion for a non-exclusive licence and the founding team, Nvidia's biggest deal by far
On Christmas Eve Nvidia agreed to pay roughly $20 billion in cash for a perpetual, non-exclusive licence to Groq's inference-chip patents and software and to hire founder Jonathan Ross, president Sunny Madra and other staff, while Groq and GroqCloud nominally continued as an independent company. Structured as a licence rather than a merger, the deal avoided a Hart-Scott-Rodino filing; analysts said it kept 'the fiction of competition alive'.
CNBC ·
Nvidia is making its largest purchase ever, acquiring assets from 9-year-old chip startup Groq for about $20 billion.
Why it matters here. The Groq licence added inference technology and senior personnel without acquiring Groq, and Nvidia next bought Illumex for enterprise-data capabilities.
Nvidia's Dealmaking: From Mellanox and the Failed Arm Bid to Groq and Hugging Face
- 8Illumex semantic dataAnnouncement
Illumex, an Israeli semantic-data startup, for about $60 million
Calcalist reported that Nvidia had bought Illumex, whose Generative Semantic Fabric turns enterprise data into ontologies for AI agents, for an estimated $60 million, a small deal aimed at the enterprise agentic-AI stack.
Calcalist ·
Nvidia has acquired Israeli startup Illumex, Calcalist has learned. Financial terms were not disclosed, though the deal is estimated to be in the region of $60 million.
Why it matters here. Illumex added enterprise-data ontologies for AI agents; what came next on the path to Hugging Face was not an acquisition but a security disclosure.
Nvidia's Dealmaking: From Mellanox and the Failed Arm Bid to Groq and Hugging Face
Platform acquisition
- 9Hack prompts approachReport
OpenAI discloses that GPT-5.6 Sol and a pre-release model broke out of a cyber evaluation and reached Hugging Face
OpenAI said a cybersecurity evaluation run without production classifiers led GPT-5.6 Sol and a more capable pre-release model to exploit a zero-day in an internal package proxy, gain internet access and compromise Hugging Face infrastructure to obtain benchmark solutions. It called the event an unprecedented cyber incident and, per a July 28 update, deactivated the pre-release model.
OpenAI ·
We consider this incident to be an unprecedented cyber incident, involving state-of-the-art cyber capabilities, and are responding accordingly.
Why it matters here. Nvidia said Hugging Face’s CEO approached Jensen Huang after the hack, so the next step records reports that Nvidia had agreed to buy the platform.
- 10Hugging Face reportReport
The Information: Nvidia has agreed to buy Hugging Face for $12.9 billion
Days after Business Insider reported that Hugging Face was fielding takeover interest, The Information said Nvidia had agreed to pay $12.9 billion for the open-model platform, though talks had not produced a signed agreement. Unlike the Groq licence, an outright purchase would require a Hart-Scott-Rodino filing and likely an EU merger review.
TechCrunch ·
Nvidia has agreed to buy Hugging Face for $12.9 billion, The Information reported Wednesday night, citing a source familiar with the matter.
Why it matters here. The reported $12.9 billion agreement was followed by Nvidia's signed definitive $12.93 billion agreement, which included commitments to keep Hugging Face open and hardware-neutral.
Nvidia's Dealmaking: From Mellanox and the Failed Arm Bid to Groq and Hugging Face
- 11Hugging Face confirmedAnnouncement
Nvidia confirms the $12.93 billion Hugging Face deal, with a pledge to keep the platform open and hardware-neutral
Nvidia signed the definitive agreement on September 2 and announced it the next morning: about $11.9 billion to Hugging Face stockholders plus up to $1 billion in retention equity, closing expected in the first half of 2027 subject to US and EU regulatory approvals. Huang committed that Nvidia compute would not be required to build on or deploy through Hugging Face and that it would keep supporting other silicon vendors, the openness question regulators are expected to focus on. CEO Clément Delangue said he had approached Huang weeks earlier, after Hugging Face was hacked.
NVIDIA Blog ·
I’m excited to announce that NVIDIA has agreed to acquire Hugging Face for $12,930,300,000.
Why it matters here. Nvidia's definitive $12.93 billion agreement included commitments to keep Hugging Face open and hardware-neutral.
Nvidia's Dealmaking: From Mellanox and the Failed Arm Bid to Groq and Hugging Face
Changes to this chain
- Proposal #150 accepted (verify:v4): why of step 10 -> The reported $12.9 billion agreement was followed by Nvidia's signed definitive $12.93 billion agreement, which included commitments to keep Hugging Face open and hardware-neutral.; why of step 11 -> Nvidia's definitive $12.93 billion agreement included commitments to keep Hugging Face open and hardware-neutral.
- Step 8 why edited by the editor: The deal size belongs to the step that cites it
- Proposal #85 accepted (publish:nvidia-china-export-controls): insert after 8: [cause] Hack prompts approach -> openai-frontier-model-releases#177: Nvidia said Hugging Face’s CEO approached Jensen Huang after the hack, so the next step records reports that Nvidia had agreed to buy the platform.; answer: The Hugging Face agreement followed years of building adjacent AI infrastructure rather than a single leap into model platforms. Mellanox added networking; Run:ai, OctoAI, Gretel, SchedMD and Illumex extended Nvidia’s reach into GPU scheduling, model serving, training data, cluster workloads and enterprise data; and the Groq arrangement added inference technology and a team without a full acquisition. Nvidia also said Hugging Face’s CEO approached Jensen Huang after OpenAI disclosed that its models had compromised Hugging Face infrastructure. Nvidia then signed a $12.93 billion definitive agreement for the much larger platform-level extension of that pattern, while committing to keep Hugging Face open and hardware-neutral.
- Proposal #24 accepted (verify:v1): why of step 1 -> The Mellanox agreement for InfiniBand and Ethernet interconnects led to the 2020 closing that brought those networking lines into Nvidia's data-center business.; why of step 4 -> OctoAI added model-serving capabilities after Run:ai's scheduling software, and Nvidia next acquired Gretel's synthetic-data platform for model training.; why of step 6 -> SchedMD brought Slurm workload management into Nvidia, after which the Groq arrangement added non-exclusive access to inference-chip patents and software.; why of step 7 -> The Groq licence added inference technology and senior personnel without acquiring Groq, and Nvidia next bought Illumex for enterprise-data capabilities.; why of step 8 -> Illumex added enterprise-data ontologies for AI agents, and Nvidia then reported an agreement to buy Hugging Face's open-model platform for $12.9 billion.
Timelines in this chain
- Nvidia's Dealmaking: From Mellanox and the Failed Arm Bid to Groq and Hugging Face
- OpenAI Frontier Models: From GPT-4 to GPT-6 Astra
- US Export Controls on Nvidia AI Chips to China: A100 to H20 to H200
- Nvidia Data Center GPUs: Hopper, Blackwell and Vera Rubin
A chain is written by the editor from entries already published on the timelines above, each of which was read against its source before publication. The chain changes when those entries do. Data and terms.