From Hugging Face to Figure: How Nvidia is betting hundreds of billions on the future of the AI ecosystem
Nvidia is reshaping the AI industry landscape at an unprecedented speed and scale. The chip giant has completed over $140 billion in transactions in the past few months and holds a nearly $100 billion equity portfolio. Its strategic intentions go far beyond simple chip sales. Jensen Huang is trying to use the power of capital to build an ecosystem of thousands of AI models to diversify the risk of dependence on a few top customers.
On October 7th, according to tech media outlet The Information, Nvidia's next investment focus will be on humanoid robots, autonomous driving technology, and AI models that can run on local devices . Several bankers, lawyers, and investors who work with Nvidia have confirmed this direction. Specifically, Nvidia has reportedly begun discussions regarding an additional $1 billion investment in humanoid robot company Figure, whose valuation prior to this funding round was approximately $38 billion.
Behind this investment offensive lies Jensen Huang's continued anxiety about Nvidia's market position. Nvidia disclosed that in the six months ending July this year, three customers contributed 44% of its total sales. This overly concentrated customer structure is the core driving force behind this massive capital deployment. Meanwhile, Nvidia's credit default swap spreads widened in August, prompting some investors to express concerns about its financial risk exposure.
The acquisition of Hugging Face: A dual game of speed and trust
Nvidia's acquisition of Hugging Face is one of the most representative cases in this wave of deals.
Hugging Face, a ten-year-old open-source AI model platform, has long attracted acquisition interest from various parties. In early summer this year, after OpenAI gained access to Hugging Face's system, it immediately began investment negotiations with the company, planning to invest $100 million; competitors such as Salesforce have also expressed their intention to acquire it.
The key turning point came after Hugging Face co-founder Clem Delangue contacted Jensen Huang. According to reports, sources familiar with the matter revealed that Huang quickly pushed forward the deal, promising Delangue that Nvidia was the only partner capable of ensuring the continued operation of the Hugging Face open-source model community. Ultimately, Nvidia completed the acquisition for $12.9 billion, equivalent to more than 80 times Hugging Face's annualized revenue of approximately $150 million .
Delangue stated at the press conference:
“Throughout Hugging Face’s development, we have received many investment and acquisition offers, which we have rejected in the past, but this summer, the time was right.”
Ecosystem Logic: Combating Customer Concentration Risk, Betting on the "Era of a Thousand Models"
Nvidia's investment logic is rooted in a clear understanding of the fragility of its own business model.
Nvidia holds approximately $99 billion in cash and marketable securities and boasts a strong cash flow. Jensen Huang hopes to leverage this financial strength to diversify the AI ecosystem—allowing thousands of models to serve thousands of application scenarios, rather than being dominated by just a few giants.
According to reports, an AI infrastructure investor familiar with Nvidia's deal said, "If I were in Jensen Huang's strategy room, I would do everything I could to tip the scales toward a world with thousands of models serving thousands of purposes."
At the model level, NVIDIA has successively invested in Anthropic, Musk's xAI, and the open-source model company Reflection AI. In addition, NVIDIA completed a $6 billion software licensing and talent acquisition deal with Poolside, partly to advance the development of its self-developed open-source model Nemotron—Poolside employees launched the Laguna open weighted AI model this year.
According to a previous report by The Information, Nvidia is also in talks to invest approximately $2.5 billion in Thinking Machines Lab, the AI lab founded by former OpenAI CTO Mira Murati.
Local AI and Edge Computing: The Next Battleground
With the increasing popularity of AI agents, the demand for local computing is surging, and Nvidia is actively expanding into the end-user market beyond data centers.
This year, AI agents capable of handling multi-step tasks on local devices have proliferated rapidly, driving up user demand for local computing power. NVIDIA has launched the DGX Spark series of computers, specifically designed for running AI agents locally.
In June of this year, engineers from AI search engine company Perplexity demonstrated to NVIDIA how to run its AI Agent software on two DGX Sparks. The demonstration impressed Jensen Huang, which triggered intensive negotiations between the two parties throughout the summer.
According to reports, sources familiar with the matter revealed that Perplexity co-founder and CEO Aravind Srinivas proposed to Jensen Huang that Nvidia acquire Perplexity in its entirety, and the two sides subsequently discussed a technology licensing and talent acquisition plan worth at least $20 billion .
Ultimately, the two parties announced a cooperation agreement at the end of August. Perplexity released a new application, Portable Computer, optimized specifically for DGX Spark, while Nvidia planned to participate in Perplexity's new round of financing with approximately $3 billion , valuing the company at approximately $35 billion prior to the round.
Data Center Bets: Financial Risks Emerge
The scale of Nvidia's credit endorsement in large data center projects has raised risk concerns among some investors.
According to reports, Nvidia initially discussed with SoftBank a $250 billion credit support package for OpenAI's large data center project on federal land in Ohio, developed by SoftBank's SB Energy, to support the data center's operation and training of models. Jensen Huang stated in a blog post that the data center campus could accommodate approximately $600 billion worth of Nvidia's computing power.
However, Nvidia's credit default swap spreads widened in August, and Jensen Huang himself frequently inquired with colleagues about the spread dynamics. Ultimately, Nvidia reduced the first phase of its credit guarantee to $105 billion and phased the project forward to buy time for subsequent decisions. At the same time, Nvidia also committed to investing $3 billion around the time of SB Energy's IPO.
In addition, Nvidia participated in OpenAI's recent funding round, investing $30 billion , with the final $10 billion completed on October 1.
To alleviate financing pressure, Jensen Huang convened six Wall Street institutions, including Blackstone, Apollo Global Management, and Goldman Sachs, in early August to jointly support a $500 billion hardware financing round. Nvidia indicated it might provide a guarantee of up to 25% for some related transactions.
Jensen Huang responded to external risk concerns at the Goldman Sachs San Francisco Technology Conference in September:
“People are starting to realize that wherever I invest, it’s not a bad investment because I’m an informed investor. We’re not taking any risks—what I need is certainty.”
The operating mechanism of the trading machine: Huang Renxun personally oversaw it.
Nvidia's transaction system is coordinated by the corporate development team led by former HPE and Oracle executive Vishal Bhagwati, but Jensen Huang himself is deeply involved in the negotiation details of key transactions.
In high-value acquisitions like Hugging Face, Jensen Huang personally led the pricing and negotiations. He also regularly meets with startup founders, investors, and executives of privately backed firms to understand product usage and potential collaboration opportunities—an informal information-gathering method strikingly similar to that of tech leaders like Microsoft CEO Satya Nadella.
It's worth noting that Nvidia doesn't always manage to seize the initiative. In July of this year, Nvidia was interested in bidding for the AI model marketplace platform OpenRouter, but missed the opportunity due to the need for more time to evaluate the deal, ultimately losing it to Stripe for $8 billion . Nvidia then turned its attention to other targets, completing transactions worth over $140 billion in the following two months.
As of the end of July this year, Nvidia's equity investments were valued at nearly $100 billion , with another $25 billion in future investment commitments yet to be fulfilled. The report indicates that multiple sources familiar with the matter stated that more deals are still in the works .
Risk Warning and DisclaimerInvesting involves risk; please exercise caution. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific circumstances. Any investment decisions made based on this information are at your own risk.