CEO Column | AI’s Next War Is Not a Model War but an Ecosystem War

Recently, something that seems unremarkable but may change the direction of AI development for the next decade has happened.

NVIDIA CEO Jensen Huang posted, on X (Twitter), his first public post related to AI policy, reposting an open letter co-signed by more than twenty heavyweight global technology companies, including NVIDIA, Microsoft, Meta, AMD, Palantir and CrowdStrike.

The core of the letter is a single sentence: the United States should support Open Weight AI rather than restrict it. Many people read it as a technical debate between “open source and closed source,” but I think the truly important point lies elsewhere entirely.

What this open letter is really telling the world is: AI competition has officially entered its second era.

The First Era: The Model Race

For the past three years the whole world has been discussing: whose model is the largest? Who has the most parameters? Whose reasoning is the strongest? GPT, Claude, Gemini, Llama, DeepSeek… The market has focused on the Foundation Model, as if owning the strongest model were the same as winning AI. But this is only the first stage of AI’s development. A model is only an engine. What truly decides whether a country or a company wins has never been the engine, but the entire industrial ecosystem.

The Second Era: The Ecosystem Race

Today’s real competition is no longer about models, but about who can build the largest AI ecosystem. Think back twenty years. Google’s success was not only a search engine.
Apple’s success was not only the iPhone.
Microsoft’s success was not only Windows.
What truly established their dominance was the complete Ecosystem.

Operating systems, development tools, applications, marketplaces, developers, hardware, services and the cloud form a positive cycle with one another, making more people willing to join and more innovation possible. AI is heading down exactly the same road. The real competition of the future is not whether GPT or Claude is better, but: who can make more companies, more developers and more industries turn AI into infrastructure that runs every day.

AI’s True Value Is Not in Chatting but in Working

There is a line in the open letter I strongly agree with: the United States should put AI into factories, hospitals, farms, classrooms and every local business. That line matters more than any model leaderboard, because what truly creates GDP is not chatbots but whether AI genuinely enters enterprise workflows: can it improve manufacturing efficiency? Can it lower medical costs? Can it shorten decision time? Can it build a company’s own knowledge system? AI should not only answer questions; it should begin to take on work. In the future every company will need an AI of its own, rather than depending forever on someone else’s AI.

Why Does NVIDIA Support Open Weight?

Many people instinctively think: if models go open source, doesn’t NVIDIA earn less? The reality is exactly the opposite. NVIDIA has never made money from models; what it really sells is compute. When AI can be deployed in every company, every factory, every hospital, every government agency and every data center, what truly grows is demand for GPUs. The easier models are to deploy, the easier it is for companies to adopt them, the easier it is for AI agents to land, and the higher overall compute demand becomes. So Open Weight is not a threat to NVIDIA but NVIDIA’s market for the next ten years.

What the US Truly Wants to Protect Is Not a Single Company but a Civilization-Level Capacity for Innovation

This is also, I think, the deepest layer of the open letter. If only a few companies in the future own the most advanced AI, then the pace of innovation across the entire world will actually slow down. True innovation always comes from millions of engineers, researchers, entrepreneurs and companies taking part together. When more people can study models, modify models and deploy models, innovation will happen at an exponential pace. This is exactly what Linux, Apache and Python have proven in history: most of the infrastructure that supports the global internet today is built on open ecosystems, and AI will very likely follow the same road.

True Security Does Not Necessarily Come from Being Closed

Many people think that locking models away is safer, but forty years of information-security development tells us otherwise: truly mature systems are often not that way because no one can see them, but because enough people can see them. Openness lets more researchers test, more red teams verify, more companies patch vulnerabilities and more developers build new defense mechanisms. Transparency does not mean there is no risk, but transparency often makes it easier to build trust. AI safety should not rely only on the promises of a few companies, but be built on the ability of the whole community to inspect it together and keep improving it.

For Companies, the Real Question Is: How Do You Build Your Own AI Sovereignty?
This is also why I have been promoting CEO AI HUB and the PSF AI ecosystem. A company’s real competitiveness in the future will not be merely procuring a few sets of AI tools, but whether it owns its own knowledge assets, its own RAG, its own Knowledge Graph and its own AI Agents; whether it can turn years of accumulated experience into an enterprise-specific intelligence system. The greatest asset of the AI era is no longer just data, but a knowledge infrastructure that can keep learning, keep accumulating and keep evolving. This is also what I believe is the most important investment every company can make in the next ten years.

Finally, I Want to Share One View

Many people ask me: is the future of AI about models or about applications? I think the answer is neither. What truly determines the future is the ecosystem. Models will get cheaper, compute will become more widespread, and AI agents will mature. What truly widens the gap is not who owns the strongest large model, but who can build an AI ecosystem in which talent, data, knowledge, compute, processes and capital continuously circulate and keep amplifying one another. Over the past decade we competed on digital transformation; over the next decade we will compete on AI infrastructure. And the real AI leader is not necessarily the one with the strongest model, but the one who can build the entire ecosystem, because what truly changes the world has never been a single model but an ecosystem that can keep giving birth to innovation.

By CEO Yang (Chun-yu Yang)