NVIDIA is reshaping the AI industry landscape at an unprecedented speed and scale. This chip giant has completed more than $140 billion in deals over the past several months and holds an equity investment portfolio of nearly $100 billion, with its strategic intent extending far beyond simply selling chips — Jensen Huang is seeking to use capital power to build an ecosystem composed of thousands of AI models, in order to diversify the risk of dependence on a handful of top customers.

On October 7, according to a report by tech media outlet The Information, NVIDIA's next phase of investment will focus on humanoid robots, autonomous driving technology, and AI models that can run on local devices. Multiple bankers, lawyers, and investors who work with NVIDIA confirmed this direction. Among these efforts, NVIDIA has already held discussions about investing an additional roughly $1 billion in humanoid robotics company Figure, which was valued at about $38 billion before this funding round.

Behind this investment offensive is Huang's persistent anxiety about NVIDIA's market position. NVIDIA disclosed that in the six months through July of this year, three customers contributed 44% of its total sales. This overly concentrated customer structure is the core driver behind this massive capital deployment. At the same time, NVIDIA's credit default swap spreads widened at one point in August, with some investors expressing concern about its financial risk exposure.

Acquiring Hugging Face: A Double Game of Speed and Trust

NVIDIA's acquisition of Hugging Face is one of the most representative cases in this deal frenzy.

Hugging Face is a ten-year-old open-source AI model platform that has long attracted acquisition interest from multiple parties. In early summer of this year, after breaching Hugging Face's systems, OpenAI immediately began investment negotiations with it, proposing a $100 million injection; competitors such as Salesforce also expressed acquisition interest.

The key turning point came after Hugging Face co-founder Clem Delangue proactively contacted Huang. According to the report, people familiar with the matter said Huang quickly pushed the deal forward and promised Delangue that NVIDIA was the only partner capable of ensuring the continued operation of Hugging Face's open-source model community. Ultimately, NVIDIA completed the acquisition for $12.9 billion, equivalent to more than 80 times Hugging Face's roughly $150 million in annualized revenue.

Delangue said at a press conference:

"Throughout Hugging Face's entire development journey, we received many investment and acquisition offers and rejected them in the past, but this summer, the timing was right."

Ecosystem Logic: Countering Customer Concentration Risk and Betting on the "Era of a Thousand Models"

NVIDIA's investment logic is rooted in a clear-eyed awareness of the fragility of its own business model.

NVIDIA holds about $99 billion in cash and marketable securities and has strong cash flow. Huang hopes to use this financial strength to push the AI ecosystem toward diversification — allowing thousands of models to serve thousands of application scenarios, rather than being dominated by only a few giants.

According to the report, an AI infrastructure investor familiar with NVIDIA's deal operations said, "If I were in Huang's strategy room, I would do everything I could to tilt the scales toward a world with thousands of models serving thousands of uses."

At the model level, NVIDIA has successively invested in Anthropic, Musk's xAI, and open-source model company Reflection AI. In addition, NVIDIA completed a $6 billion software licensing and talent recruitment deal with Poolside, partly to advance development of its in-house open-source model Nemotron — Poolside employees launched the Laguna open-weight AI model this year.

According to a previous report by The Information, NVIDIA is also in talks to invest about $2.5 billion in Thinking Machines Lab, an AI lab founded by former OpenAI Chief Technology Officer Mira Murati.

Local AI and Edge Computing: The Next Battleground

With the spread of AI agents, demand for local computing has surged, and NVIDIA is actively positioning itself in end-device markets beyond data centers.

Since the beginning of this year, AI agents capable of handling multi-step tasks on local devices have spread rapidly, driving up user demand for local computing devices. NVIDIA has launched the DGX Spark series of computers, designed specifically for running AI agents locally.

In June of this year, engineers from AI search company Perplexity demonstrated to NVIDIA how to run its AI agent software on two DGX Sparks. The demonstration left a deep impression on Huang, opening the door to intensive negotiations between the two sides throughout the summer.

According to the report, people familiar with the matter said Perplexity co-founder and CEO Aravind Srinivas once proposed to Huang that NVIDIA fully acquire Perplexity, and the two sides subsequently discussed a technology licensing and talent recruitment plan worth at least $20 billion.

Ultimately, the two sides announced a cooperation agreement at the end of August. Perplexity released a new application called Portable Computer optimized for DGX Spark, while NVIDIA plans to participate in Perplexity's new funding round with about $3 billion, at a pre-money valuation of about $35 billion.

Data Center Mega-Bet: Financial Risks Emerge

The scale of NVIDIA's credit backing for large data center projects has triggered risk concerns among some investors.

According to the report, in a large data center project developed by SoftBank's SB Energy on federal land in Ohio, NVIDIA initially discussed with SoftBank a plan to provide up to $250 billion in credit support to help OpenAI lease the data center to run and train models. Huang said in a blog post that the data center campus could accommodate about $600 billion worth of NVIDIA computing power.

However, NVIDIA's credit default swap spreads widened in August, and Huang himself frequently asked colleagues about spread movements. Ultimately, NVIDIA reduced the first-phase credit guarantee to $105 billion and advanced the project in stages to buy time for subsequent decisions. At the same time, NVIDIA also committed to investing $3 billion around SB Energy's IPO.

In addition, NVIDIA also participated in OpenAI's recent funding round with a $30 billion investment, with the final $10 billion completed on October 1.

To spread financing pressure, Huang convened six Wall Street institutions in early August, including Blackstone, Apollo Global Management, and Goldman Sachs, to jointly support $500 billion in hardware financing. NVIDIA said it may provide backstop guarantees of up to 25% for some related transactions.

At Goldman Sachs' San Francisco tech conference in September, Huang responded to external risk doubts:

"People are beginning to realize that wherever I invest, it is not a bad investment target, because I am an informed investor. We have not taken on any risk — what I need is certainty."

How the Deal Machine Operates: Huang Is Personally Involved

NVIDIA's deal system is coordinated by a corporate development team led by former HPE and Oracle executive Vishal Bhagwati, but Huang himself is deeply involved in the negotiation details of key transactions.

In high-value acquisitions such as Hugging Face, Huang personally led pricing and negotiations. He also regularly meets with startup founders, investors, and executives of private-equity-backed companies to understand their product usage and potential cooperation opportunities — an informal information-gathering approach similar to that of tech leaders such as Microsoft CEO Satya Nadella.

It is worth noting that NVIDIA does not always seize the initiative. In July of this year, NVIDIA was interested in joining the competition to acquire AI model marketplace OpenRouter, but missed the opportunity because it needed more time to evaluate the deal, and Stripe ultimately acquired it for $8 billion. NVIDIA immediately shifted its attention to other targets and completed more than $140 billion in deals over the following two months.

As of the end of July this year, the market value of NVIDIA's equity investments was close to $100 billion, with another $25 billion in future investment commitments still pending. The report noted that multiple people familiar with the matter said more deals are still in the pipeline.