In July, NVIDIA (ASX: NVDA) learned that AI model marketplace startup OpenRouter was about to be acquired. The three-year-old company had become a popular trading hub for AI models, hosting numerous cost-effective open-source models. Databricks and Stripe were already in talks with the startup about a sale, and Nvidia had entered the fray too late. Upon hearing the news, the chip giant's CEO Jensen Huang wanted to get involved in the acquisition. A person familiar with the negotiations said Nvidia executives expressed acquisition interest to OpenRouter's deal team and were prepared to make a generous offer, but needed more time to evaluate the transaction before formally bidding. However, OpenRouter's founder was unwilling to wait. Ultimately, Nvidia never submitted a formal acquisition offer (no media outlet had previously reported Nvidia's interest in acquiring this startup), and Stripe secured OpenRouter with an $8 billion bid. Nvidia's deal team quickly pivoted to other acquisition targets.

Over the following two months, Nvidia finalized various deals worth over $140 billion in total, including a $105 billion credit guarantee. As of late July, Nvidia held nearly $100 billion in equity investments, with an additional $25 billion in future investment commitments, and more deals potentially on the horizon. Although investors worry that Nvidia and other tech giants are expanding their AI-related financial commitments too aggressively, multiple investment banks, lawyers, and investors working with Nvidia said the chip designer is expected to continue this wave of large-scale dealmaking in the coming months. Nvidia is scouting startups to invest in or acquire in robotics, autonomous driving technology, and AI models that can run directly on local devices such as smartphones and home computers.

A person familiar with the matter revealed previously undisclosed information that Nvidia has discussed an additional $1 billion investment in humanoid robot company Figure. Figure was valued at approximately $38 billion before its current funding round, and the company is in the midst of a new fundraising effort. (Nvidia is already an existing investor in Figure, which completed a funding round of over $1 billion at that valuation a year ago.) People familiar with the matter said Nvidia may also acquire startups to help advance its self-developed Nemotron open-source large model while expanding into other AI application sectors. Nvidia agreed to pay $6 billion to license Poolside's software and absorb its team, partly to advance the Nemotron project. The Poolside team launched the Laguna open-weight AI model this year.

This series of investment moves is led by a leader who has always maintained a high sense of crisis about Nvidia's industry position. Huang has previously mentioned that he constantly worries Nvidia's lead could erode, whether due to cooling AI demand or the emergence of competitors whose performance could replace Nvidia chips. With $99 billion in cash and marketable securities, plus a steady stream of cash flow, Huang wants to leverage Nvidia's substantial financial resources to build an industry ecosystem featuring thousands of quality AI models, rather than having just a few model makers dominate. This can avoid the risk of a handful of large customers like OpenAI having outsized influence over Nvidia's revenue. Nvidia disclosed that during the six-month period ending in July, three major customers contributed 44% of the company's total sales.

An investor familiar with Nvidia's deal strategy and its investments in AI infrastructure and applications said: "If I were in Jensen's war room, I would do everything possible to create a situation with thousands of models serving thousands of different scenarios." This explains why Nvidia's deal engine is running at full speed. According to reports, Nvidia recently discussed investing approximately $2.5 billion in Thinking Machines Lab, an AI lab founded by Mira Murati, former OpenAI chief technology officer. If the deal materializes, it would be another addition to Nvidia's series of model company investments, following its investments in Anthropic, Musk's xAI, and open-source model company Reflection AI.

Nvidia has also become a core financial backer of large data center projects, investing $3 billion in SB Energy. SB Energy is building a massive data center for OpenAI, and Nvidia is providing a $105 billion guarantee for OpenAI's lease agreement on the project. Huang wrote in a blog post that the data center campus developed by SB Energy and OpenAI can host approximately $600 billion worth of Nvidia computing power. Nvidia chose to support OpenAI's project because "frontier AI labs have enormous demand for training and inference computing power, but many companies are expanding faster than their balance sheets and long-term credit profiles can support."

Although Nvidia missed out on OpenRouter, Huang acts quickly when competing bidders emerge for other targets. Take the recent acquisition of Hugging Face as an example. Investment banks and people close to the company said this decade-old startup, which has become a well-known open-source AI model hosting platform, has received acquisition offers for years. People familiar with the matter said Nvidia's corporate development team had long been interested in investing in Hugging Face. A person familiar with the plans revealed that in early summer this year, after OpenAI's intelligent agents launched an attack on Hugging Face, OpenAI held preliminary talks with Hugging Face about investing $100 million. Around July, Hugging Face's existing investor Salesforce and other competing companies also expressed acquisition interest. Hugging Face co-founder Clem Delangue approached Huang to inform him of these potential acquisition offers.

People familiar with the matter said Huang quickly advanced the deal, assuring Delangue that Nvidia would be the only trustworthy partner capable of ensuring Hugging Face's open-source model community continues to operate. Nvidia made a $12.9 billion acquisition offer that was hard to refuse, equivalent to more than 80 times the startup's $150 million in annualized revenue. Delangue said at a press conference: "Along the way, Hugging Face has received numerous investment and acquisition offers, and we declined them all in the past. But this summer, everything was in place and the timing was right."

Multiple people working with Nvidia in the local computing sector said that in the coming months, Nvidia plans to pursue more deals to push GPUs out of data centers and into home devices, where users will increasingly run AI locally on personal computers and small devices. This year, as AI agents (software capable of executing multi-step tasks such as booking flights and organizing email) became widespread, demand for local AI surged. These agents drove users to purchase Apple Mac minis in large numbers, as the device is well-suited for running such tasks. Nvidia launched new products like DGX Spark specifically targeting local AI agent operation. Nvidia's recent deals also reflect its intention to expand its product matrix in this sector.

People familiar with the matter said startup Perplexity, which began as an AI search engine and later developed the AI agent Perplexity Computer, had its engineers demonstrate to Nvidia's team in June that their software could run on two DGX Sparks. Huang was very interested after hearing about the demonstration, and the two companies negotiated a potential deal throughout the summer. People familiar with the matter said Perplexity co-founder and CEO Aravind Srinivas proposed to Huang that Nvidia could directly acquire Perplexity. The two sides then shifted to negotiating a technology licensing plus talent acquisition arrangement, with Nvidia paying at least $20 billion, possibly more, to acquire Perplexity-related technology. The two sides officially announced the partnership in late August. Under the agreement, Perplexity launched a new application called Portable Computer, custom-optimized to run its AI agent better on DGX Spark devices. A person familiar with the matter said Nvidia plans to invest approximately $3 billion in Perplexity through an equity funding round, valuing the company at $35 billion before the investment.

Like most tech companies, Nvidia has a corporate development team responsible for closing deals. The team is led by Vishal Bhagwati, who previously worked at Hewlett Packard Enterprise and Oracle. Even so, Huang is often deeply involved in deal details, personally finalizing deal prices and leading negotiations on large acquisitions like Hugging Face. Huang also frequently meets with startup founders, investors, and executives of private equity-controlled companies to understand how they use Nvidia products and what support Nvidia can provide. Executives like Microsoft CEO Satya Nadella also use such informal research methods to grasp technology trends.

In recent months, a new reality has confronted Huang that could limit Nvidia's acquisition ambitions: Nvidia may be approaching the upper limit of funds it can safely deploy for investments, acquisitions, and large project guarantees. The financing plan for SoftBank's SB Energy massive data center project on federal land in Ohio reflects this market concern. In early summer this year, SoftBank and Nvidia initially discussed the chip giant providing up to $250 billion in credit support for OpenAI, which planned to lease the data center and use Nvidia chips to train and run models. (Nvidia also participated in OpenAI's recent fundraising, investing $30 billion, with the final $10 billion completed on October 1.) In August, Nvidia's credit default swap spreads widened, reflecting some investors' concerns that the company had taken on excessive risk. Huang noticed this indicator and frequently asked colleagues about spread changes.

Nvidia ultimately provided only a $105 billion credit guarantee for the project's first phase, still enormous but less than half the initially discussed amount. The project's phased construction gives Nvidia several years to decide whether to continue funding the second phase. Meanwhile, Nvidia continues to pursue another investment, planning to invest $3 billion in SB Energy before and during its IPO. Huang emphasized that continued AI progress requires more companies to participate in funding chips, data centers, and power projects. In early August, he convened six Wall Street institutions including Blackstone, Apollo, and Goldman Sachs to jointly raise $500 billion for purchasing Nvidia hardware. Nvidia said it could provide backstop guarantees of up to 25% of total financing in some related deals. But Huang downplayed the risks of such massive investments, disagreeing that cash-rich companies like Nvidia are over-investing. In September, Huang told the audience at Goldman Sachs' annual technology conference in San Francisco: "People are gradually realizing that wherever I invest, it usually turns out well, because I have an information advantage. I don't take blind risks. We're not as clever as everyone thinks. What I want is sure wins."