Nvidia, the company on a crazy investment spree, has transformed from a chip seller to the "Federal Reserve" of AI.
Throughout the history of business, it's hard to find a trillion - dollar company that has achieved such a remarkable market - value leap in such a short period as NVIDIA.
Its rise is astonishing, but behind this amazing growth, there has always been a sense of tension among NVIDIA's senior management:
What if one day, AI no longer relies on NVIDIA's GPUs?
To hedge against this potential uncertainty, NVIDIA has chosen the most direct and radical approach: binding the entire AI ecosystem to its chariot. And its tool is investment.
In the past two years, NVIDIA has made 83 investment moves, surpassing Alphabet's 73 and Microsoft's 40 during the same period. This year, the pace has even accelerated. In just one week in mid - September, NVIDIA officially announced five major investments, with a total amount exceeding $9.2 billion, almost equivalent to the technology budget of a medium - sized country.
From large models to cloud computing, from AI applications to embodied intelligence, NVIDIA's presence can be seen at almost all the key nodes of popular AI tracks.
But this is not all. What really sets NVIDIA apart from other tech giants is that it is trying to establish an AI order dominated by itself.
It holds the issuance right of the "computing - power currency" like a central bank. On one hand, it releases the "liquidity" of GPUs to support the rise of new players such as CoreWeave, Lambda, and Together AI; on the other hand, as the "lender of last resort", it steps in to support key companies like OpenAI when they encounter difficulties in financing, stabilizing the entire ecosystem.
In this newly emerging AI world, NVIDIA is no longer just a chip - selling company. It is more like a macro - regulatory machine, a super - role with systematic influence.
At this point in time, perhaps we should get to know NVIDIA anew.
01
All - in on Large Models, Optimistic about Four AI Application Directions
Judging from the data, NVIDIA not only makes many investment moves but also makes large - scale ones.
Since 2023, there have been 31 companies in which NVIDIA participated in financing with a single - round financing exceeding $100 million. Among the companies it invested in 2025, there are 7 companies with a single - round financing exceeding $1 billion (equivalent to about 7.1 billion RMB), 8 companies with financing of hundreds of millions of dollars, and 5 companies with financing just over $100 million.
In terms of industry distribution, among these large - scale financings, the most are related to large models, with a total of 14 companies. Next are those related to AI infrastructure, with a total of 7 companies. There are 3 companies related to autonomous driving and 2 companies related to humanoid robots respectively. The remaining 5 companies are related to quantum computing, nuclear fusion, etc.
Model companies are the type that NVIDIA bets the most on, without a doubt.
On one hand, NVIDIA has invested in almost all leading models. There are top - tier model companies such as OpenAI, Anthropic, and xAI, as well as differentiated model manufacturers, such as Cohere, which focuses on enterprise - level models, and Together AI, an open - source model manufacturer.
On the other hand, out of the consideration of "sovereign AI", NVIDIA also pays particular attention to betting on regional models. For example, NVIDIA has successively invested in Mistral in Europe, Sakana AI in Asia, and Israeli model company AI21 Labs.
The logic behind NVIDIA's desperate investment in models is very simple. Model manufacturers are NVIDIA's most direct customers. By investing in model companies, NVIDIA can more directly lock in future hardware orders and system deployments.
It's like if you run a flour mill, instead of just selling flour to bakeries, you simply invest in some promising bakeries so that they will use your flour every time they make bread in the future.
In addition to the model layer, infrastructure is another area where NVIDIA invests the most.
In the past three quarters of 2025, the number of AI infrastructure companies invested by NVIDIA is almost on a par with that of model manufacturers, each accounting for 31%. In addition to its "affiliated company" CoreWeave, NVIDIA has also invested in Lambda Labs, Crusoe, and Nscale.
These companies all have one thing in common: they did not start from cloud computing.
For example, CoreWeave originated from miners, Lambda Labs sells deep - learning workstations, and Crusoe engages in power recovery in oil and gas fields. Now they have all transformed into AI cloud service providers, all thanks to NVIDIA's promotion.
Why does NVIDIA do this? To put it simply, it is "buying customer growth". Through investment, these cloud service companies can develop faster, and the sales, leasing, deployment of chips, and software usage will all increase accordingly.
More importantly, these platforms are not self - developed, so they are more flexible. NVIDIA can control the ecosystem without directly conflicting with large customers such as AWS and Azure, making money while avoiding suspicion.
Looking at the application layer, NVIDIA seems to be betting on the future. It is optimistic about two directions in the software layer: vertical AI and multi - modality. Harvey and Hippocratic AI are representatives of the former, while Runway, Luma, and SoundHound belong to the latter.
Behind these choices is Jensen Huang's judgment. He believes that models must be able to "see, hear, and reason" simultaneously. From text, pictures to videos and audio, multi - modal fusion is the key to understanding the world.
Autonomous driving and robots are the hardware directions that NVIDIA is betting on. In 2019, it invested in Wayve, and in 2024, it added more than $1 billion. Figure AI is the company in the robot field that NVIDIA bets the most on, with a $675 million investment in Series B and a 5 - year GPU lease agreement.
Jensen Huang has never hidden his enthusiasm for robots. At the shareholders' meeting, he clearly stated that robots will be NVIDIA's "second growth curve" after AI, and autonomous driving will be the first to be implemented and take the lead.
02
The Essence of NVIDIA: An AI Central Bank that Provides Liquidity and Acts as a Backstop
Although NVIDIA makes many investments, there are still two counter - intuitive phenomena:
First, its investments are concentrated in the mid - to - late rounds, often after Series B.
Second, despite its frequent investment moves, it rarely acts as the lead investor.
This "not taking the leading role" attitude is particularly evident in Reflection's latest round of financing.
In Reflection's new $2 billion financing round this month, NVIDIA's investment pool accounts for as much as 40%. However, when the outside world called it the "lead investor" of Reflection, NVIDIA quickly issued a statement, emphasizing that it is only "one of several investors".
People familiar with the matter revealed that NVIDIA deliberately avoids the role of "negotiation leader". Even if it invests the most money, it is not willing to join the board of directors, set terms, or lead the company's development.
For it, returns are not the focus; the ecosystem is.
Rather than making more money, NVIDIA cares more about participating in shaping the future AI market landscape through investment. What it wants is not just financial returns, but intelligence, influence, and influence.
By locking in emerging trends in advance, influencing technological directions, and binding companies that may explode in the future, it encloses the most potential customers in its GPU system. Investment is a way for it to build an ecosystem and extend its control.
This strategy has made NVIDIA far exceed the role boundaries of traditional technology companies and is becoming more and more like a macro - regulator.
To some extent, NVIDIA is playing a role similar to that of an AI government, or even a central bank.
On one hand, when GPUs have become the hard currency in the AI era, NVIDIA is the one with the issuance right.
Like a central bank controlling currency, it promotes the development of the AI ecosystem in a very direct way by issuing currency, and at the same time controls both the release and restriction of liquidity.
For example, emerging AI cloud companies such as CoreWeave and Lambda have taken off rapidly with NVIDIA's investment and support. Since its listing in March, CoreWeave's stock price has risen by about 2 - 3 times (the company's customers have expanded from NVIDIA and Microsoft to OpenAI, Meta, etc.), and Lambda is also about to go public.
They often get the latest H100 chips earlier than traditional large - scale enterprises, and there is even a cooperation form of "getting the chips first and then sharing the profits later".
Meanwhile, NVIDIA has also begun to exhibit characteristics similar to those of a government regulating the economy. At certain critical moments, it plays the role of the "lender of last resort" in the AI industry.
The most typical example is OpenAI. OpenAI itself said that it won't achieve positive cash flow until 2029, which means it will need to burn money for another 5 years, a total of $115 billion. However, its valuation has soared to $500 billion, and traditional investors are becoming more and more hesitant.
Just at this critical moment, NVIDIA stepped in and invested $10 billion. Because NVIDIA clearly knows that at this time, nothing can more determine the future demand trend of GPUs than OpenAI.
OpenAI's story is not an isolated case. According to two people familiar with the matter quoted by foreign media The Information:
Recently, some large - scale data center builders have encountered difficulties in project financing, and traditional financial institutions have become more cautious. Just at this critical moment, NVIDIA stood up.
It is reported that the amount of funds injected by NVIDIA into AI companies such as OpenAI has approached the level of government stimulus plans.
The reason why NVIDIA is willing to intervene and provide financing support for data center companies is that it helps attract other project funders to join.
In other words, NVIDIA is using its balance sheet to support the liquidity of the entire AI world. And this is exactly the key role that traditional central banks play in the financial system - preventing the occurrence of systemic risks.
This approach has also raised concerns. Harvard Business School professor David Yoffie pointed out that NVIDIA is artificially creating demand. Downstream companies receive NVIDIA's money to place orders for chips before getting orders from end - customers, creating demand in advance.
However, from NVIDIA's perspective, this is still a very clear - cut deal.
With abundant cash on hand and limited by regulations, it cannot conduct large - scale mergers and acquisitions. In front of it is a newly emerging market, while behind it is a typical cyclical industry. In such a situation, investing money in the ecosystem is not only about locking in demand in advance but also about building a moat with capital. When the cycle really comes, NVIDIA can ensure stable shipments, stable customers, and a stable rhythm.
To some extent, this style is very typical, paranoid and pragmatic, highly reflective of Jensen Huang's personal qualities.
At first, almost no one believed that GPUs would become the mainstream computing power, but only Jensen Huang held on tightly. Facing grand propositions such as "whether AI will exterminate humanity", he never participates in debates and just focuses on doing the right things in front of him.
In Jensen Huang's view, rather than being at the mercy of the cycle and facing the risk of being disrupted, it is better to deeply bind all the key participants in the entire ecosystem to its chariot.
He always adheres to one principle: Only the paranoid can survive.
This article is from the WeChat official account "Crow Intelligence Talk", author: Intelligent Crow. Republished by 36Kr with permission.