Alles nur Tricks. NVIDIA investiert hundert Milliarden US-Dollar in OpenAI. Altman nimmt das Geld und kauft Grafikkarten, während Oracle noch Profit aus dem Preisunterschied macht.

NVIDIA and OpenAI have reached a century - long cooperation!

According to the latest news, NVIDIA plans to invest up to $100 billion in OpenAI. In return, OpenAI will use this money to build at least 10GW of AI data centers, all using NVIDIA's systems.

10GW is equivalent to 4 - 5 million GPUs. Moreover, Jensen Huang said in August that the cost of building a 1GW data center is approximately $50 - 60 billion.

It is reported that the first $10 billion of this cooperation will be invested when the first 1GW data center is deployed (on the Nvidia Vera Rubin platform), which is expected to be in the second half of 2026.

Subsequently, every time OpenAI deploys a 1GW system, NVIDIA will make an investment, with a maximum total investment of $100 billion.

NVIDIA and OpenAI are now more deeply intertwined. But who still remembers Oracle in this relationship...

OpenAI spends $300 billion on Oracle's cloud services → Oracle buys NVIDIA's GPUs → NVIDIA invests in OpenAI. Here comes the "real triangle relationship".

The three parties have mastered the billion - dollar cycle in the AI circle.

The "Wealth - Computing Power" Triangle Cycle

Let's first talk about the basis of this cooperation: NVIDIA will invest money in stages, and OpenAI will fully deploy NVIDIA's systems.

Although NVIDIA plans to invest $100 billion in stages, it's not for free. The first $10 billion will be given only after OpenAI deploys its first 1GW NVIDIA system. It is said that the investment will also be calculated based on the valuation at that time, which is somewhat like paying according to the results.

However, OpenAI is also decisive. It promises to build at least 10GW of AI data centers, all using NVIDIA's systems. 10GW is equivalent to 4 - 5 million GPUs.

Moreover, according to Jensen Huang, the cost of building a 1GW data center is approximately $50 - 60 billion.

OpenAI said that the first - stage infrastructure based on the Nvidia Vera Rubin platform can be completed as early as the second half of 2026.

△

The billion - dollar strategic cooperation does seem to deeply bind the two parties. But what's more interesting is this "triangle drama" with Oracle as the middleman.

A few days ago, OpenAI's $300 - billion spending on Oracle's cloud services caused Oracle's stock price to soar by 36% in a single day, and its market value increased by $251 billion.

However, for Oracle's cloud services to run smoothly, it still depends on NVIDIA's GPUs, so Oracle has to place orders for chips from NVIDIA.

Calculated in this way, the $100 billion that NVIDIA plans to invest in OpenAI (has planned but not actually invested) may first become the cloud service fee that OpenAI pays to Oracle, and then become the chip procurement payment that Oracle pays to NVIDIA.

After going around in a circle, the money may even return to NVIDIA's pocket... But OpenAI will have sufficient computing power, and Oracle and NVIDIA's positions in the AI supply chain will also be more stable.

This triangle wealth cycle doesn't seem like a bad deal.

△

Of course, all three parties are smart and have their own plans.

For OpenAI, ChatGPT has a weekly active user base of up to 700 million. Without sufficient computing power, model iteration and operation and maintenance services are just empty talk.

On the other hand, by using Oracle's cloud services to implement infrastructure, it's like holding both hardware and cloud computing firmly in hand.

For NVIDIA, not only does it lock in OpenAI as a core customer through investment, but it also expands its sales through Oracle's GPU procurement. Holding both the leader in AI models and the giant in cloud computing, it ensures its important position in the AI computing power supply chain.

Oracle also doesn't lose out. The $300 - billion cloud service order has directly boosted its stock price, and it can also ensure the computing power of its cloud services with NVIDIA's chips, at least securing a popular seat in AI cloud computing.

Back to this cooperation between NVIDIA and OpenAI, both Jensen Huang and Sam Altman are full of confidence.

Regarding this, Sam Altman said:

Everything starts with computing.

Computing infrastructure will be the cornerstone of the future economy. We will use the results of our cooperation with NVIDIA to both create new AI breakthroughs and empower individuals and enterprises on a large scale.

Jensen Huang also said:

NVIDIA and OpenAI have been driving each other forward in the past decade, from the first DGX supercomputer to the breakthrough of ChatGPT.

This investment and infrastructure cooperation mark the next leap in (AI) — the next - era intelligence driven by a 10GW system.

One More Thing

NVIDIA is not only making a huge investment in OpenAI but also investing in companies such as Intel and Nscale recently.

It even spent over $900 million to recruit the CEO and core technologies of an AI Infra startup, making an acquisition for cost - reduction and efficiency - improvement.

However, compared with $100 billion, $900 million is just pocket money for Jensen Huang.

Reference Links:

[1]https://openai.com/index/openai-nvidia-systems-partnership/

[2]https://www.cnbc.com/2025/09/22/nvidia-openai-data-center.html

[3]https://x.com/SullyOmarr/status/1970176527137718654

This article is from the WeChat official account "QbitAI", author: Wen Le. It is published by 36Kr with authorization.