Jensen Huang has finally taken sides.

Jensen Huang posted his first-ever tweet in his life.

The man in charge of a company with hundreds of billions of dollars in market value has not spoken on social media for decades. In the past, we saw Jensen Huang through GTC keynotes, leather jacket signings, night market visits, and high-level meetings with national leaders.

And then? On July 24, he suddenly appeared on X and made a statement.

What's even more surprising is the content. The tweet didn't talk about Blackwell, didn't mention GB300, didn't promote any hardware, but publicly stood up for open-source AI.

The CEO of a GPU company, in his first tweet ever, didn't talk about GPUs.

This is quite unusual.

At a time when the conflict between open source and closed source is at its peak, this chip giant, which has always remained neutral, has clearly taken a side for the first time.

I. After Decades of Silence, He Suddenly Speaks Out

In the past few years, Jensen Huang didn't need to take sides at all.

The reason is simple. All paths in AI ultimately require purchasing GPUs, whether you are closed-source or open-source. The fiercer the competition, the more money NVIDIA makes. By staying neutral, no matter who wins, everyone will eventually come to NVIDIA.

This is NVIDIA's biggest structural advantage. It doesn't need to take sides or make public statements, just focus on being a supplier, and it can make money effortlessly.

But this premise is starting to falter.

Why?

One direct factor is that today's Washington is discussing restrictions on open source.

A group of people are pushing to tighten open-source AI policies, with a plausible reason: making frontier model weights open may lower the threshold for the development of biochemical weapons and cyberattacks. In congressional hearings and defense-related policy discussions, some policymakers are trying to directly classify open-weight models as high-risk technologies, requiring pre-review or even administrative freeze on open-source models with parameter sizes reaching certain thresholds.

That sounds reasonable.

But if open source is completely stifled, fewer people will train models, fewer people will deploy them, the pace of innovation will slow down, and GPU demand will decline accordingly. This is equivalent to forcibly setting an upper limit on NVIDIA's most proud computing consumption market through top-down intervention.

Once this policy takes further effect, it will directly disrupt NVIDIA's growth model, and Jensen Huang cannot stand by and watch.

II. No Talk About Hardware, Only Open Source

The reason why Jensen Huang publicly took a stand for open source.

There is another layer of factor behind it.

In the past, when Meta launched the open-source Llama, everyone thought it was just a defensive move by a Silicon Valley giant. European and American regulators did not see it as a fatal threat. But after DeepSeek, coupled with the recent sensation caused by the release of Kimi K3, the situation has become different.

This is the first time that instead of Silicon Valley giants like Meta, Google, or Microsoft, two Chinese startups, with lower costs and open-weight approaches, have successively pushed open-source models into the world's top tier.

In the past, everyone thought that only closed-source giants using arrays of tens of thousands of H100/B200 GPUs to stack computing power could open the door to AGI.

But DeepSeek and Kimi K3, through MoE architecture innovation, attention mechanism optimization (MLA), and extreme engineering capabilities, have proven that even in an environment with limited computing power, open-source weights can still match top closed-source models in algorithm efficiency.

This incident has directly changed the global power structure of AI.

What really changed the situation is not that DeepSeek and Kimi went open-source, but that they proved something no one believed in before:

Open source does not have to come from Silicon Valley.

In the past, when the US discussed open source, it was about technical routes. Today, when the US discusses open source, it is about industrial competition.

Once the world's most powerful open-source models begin to appear outside the US, regulation is no longer just regulation, it has become a tool for competition.

First of all, it breaks the closed-source myth. It proves that without burning billions of dollars or being monopolized by a few Silicon Valley giants, the open-source ecosystem can still produce top-tier models.

Secondly, it has alarmed Washington. US regulators realize that the power of open source may be in the hands of overseas teams, and they are in a hurry to put a "safety supervision" spell on open source.

DeepSeek and Kimi have proven that open source can not only survive, but also go through a complete growth path: open weights attract global developers, developers contribute improvements, improvements drive enterprise deployment, and deployment translates into computing power orders.

There is a counterintuitive business logic hidden here. The higher the algorithm efficiency and the cheaper the model, the more exponentially the scenarios for AI applications will explode, and the total computing power market that is ultimately driven will multiply. This is the famous Jevons Paradox.

Once this flywheel starts spinning, the biggest beneficiary of computing power will still be NVIDIA.

When Jensen Huang talks about open source, he is not doing charity. It is because the path that Chinese startups have broken through with open source is exactly supporting NVIDIA's core lifeline.

He just takes this opportunity to amplify the potential of the open-source path.

III. Pulled More Than 20 Companies to Sign a Joint Statement

Making a statement alone is not enough. Jensen Huang directly gathered more than 20 companies to sign a joint open letter titled "Open Weight Models and U.S. AI Leadership".

A hardware seller suddenly took on the role of an industry organizer. This in itself is quite unexpected.

Logically, Jensen Huang could have spoken out directly, and his influence would not be small. But he chose not to fight alone. The only explanation behind this is that the pressure from both closed-source giants and political regulation is too much for him to bear alone.

Looking at the signatories of the letter, you will find that these companies have been fiercely competing with each other in private. Meta and Microsoft are fighting for customers in the cloud market, companies invested by a16z have overlapping businesses with IBM, Perplexity and Microsoft Bing are competing for search traffic, and Mistral and Meta are vying for the top spot in open source.

Looking closely at this list, there is actually a deep business tacit understanding behind it.

Top venture capital firms like a16z and Y Combinator have invested in hundreds of AI application startups. If all these startups have to pay rent to closed-source APIs, their profits will be depleted, so they must rely on open source to gain technical autonomy. For community platforms like Hugging Face, the open-source ecosystem is their very foundation.

And companies like Mistral AI, Perplexity, and major enterprise cloud service providers need to use open weights to penetrate customers' private clouds and on-premises data centers, breaking the blockade of data access by closed-source giants.

Today, these groups of people have been pulled to the same table by Jensen Huang.

Jensen Huang's move is very deliberate. He has elevated the interests of open source to an ecological alliance. Every word he says is fighting for legitimate space for the entire alliance.

Jensen Huang has successfully tied NVIDIA's hardware demand to the entire Silicon Valley midstream and downstream ecosystem, packaging his own business calculations as the public interest of the entire industry.

IV. The Absentees Reveal More Than The Attendees

Looking at the list of absentees is more interesting than looking at who signed.

OpenAI, Anthropic, and Google, the three major closed-source frontier labs, are all absent.

This reveals the most essential part of this struggle: closed source is just a technical choice, while being closed is a form of business monopoly.

OpenAI and its peers want to maintain a structure where they control top models, traffic entrances, and distribution channels, and others can only pay per use under their API framework.

If this closed ecosystem succeeds, all application developers will become laborers, and the added value of the entire AI industry chain will be concentrated in a very small number of upstream entrances.

For NVIDIA, the most fatal crisis is hidden right here. OpenAI is working with Broadcom to actively develop its own ASIC chips, Google has a very mature TPU iteration array, Microsoft is promoting Maia, and Amazon has Trainium.

Once the top closed-source giants gain dominance through entrance monopoly and completely take over the underlying computing power with self-developed chips, NVIDIA will change from an indispensable infrastructure to a negotiable accessory supplier. At that point, NVIDIA's gross margin myth will also be shattered.

Jensen Huang formed this alliance precisely to prevent AI from becoming the private territory of a few giants.

NVIDIA is not afraid of having too many models, it is afraid of having too few entrances. As long as the industry remains multipolar, with dozens of companies running open-source models, NVIDIA will always be the only supplier.

V. From Selling GPUs to Setting Rules

From Jensen Huang's tweet, to forming the alliance, and then to the absence of closed-source giants, it seems that NVIDIA is strengthening its own position.

But the deeper impact is that this tweet marks an inflection point.

In the past, GPUs determined AI. Whoever had more GPUs could train faster and stay ahead. Today, the opposite is true: rules are beginning to determine GPUs. How regulations are set determines who can start training. How open-source policies develop determines how many people buy computing power.

NVIDIA's CUDA moat, which it relied on for survival in the past, is essentially a moat of technology and ecology. But when administrative supervision, legislative compliance, and antitrust reviews are forced into this industry, a purely technical moat is no longer sufficient.

He must build a moat of alliances and rules. To this end, Jensen Huang has pushed himself from a needed supplier to a rule-maker who defines what is needed.

He knows very well that if he doesn't sit at the table, others will write him into the menu.

VI. Large Language Models Leave Their Adolescence

At this point, the core truth has basically emerged.

The era when people only relied on papers, model benchmarking, and technical passion to impress the world is completely over. AI has officially bid farewell to its adolescence and sailed into the cruel deep waters of commercialization.

In the past, Microsoft relied on Windows to monopolize desktops, forcing the entire industry to open up gaps through open-source Linux and browser alliances. In the mobile internet era, Apple used the App Store to build a closed-source empire, forcing the entire industry to form an Android open-source camp to fight back.

The current confrontation between closed-source APIs and open-source weights in the AI industry is just a repeat of these historical events.

The moment Jensen Huang sent his first tweet marks that AI has officially ended its adolescence. It is no longer a simple technical toy, but has become a mature industry that requires competing for entrances, building ecosystems, and playing games with regulators.

Here, no one can just be a pure technical developer. Everyone must take sides between rules, power, and ecology.

Words Beyond the Page:

To be honest, when the person selling shovels starts discussing how to manage the mine, it's hard to tell whether he is maintaining fairness, or ensuring that he can always have shovels to sell.

But on the other hand, when an industry needs the shovel seller to take the lead in maintaining market fairness, it precisely means that the golden age of diverse innovation and technology supremacy is truly gone forever.

Over the past two decades, Silicon Valley has changed the world through products. In the next two decades, it may first need to learn to change the rules.

So did Windows, so did Android, so did the App Store.

AI will be no exception.

Jensen Huang is not the first person to realize this, but he may be the first person to publicly admit it.

This article is from the WeChat public account"Beyond the Page", written by Huahua, and published with authorization from 36Kr.