Views: 0
I recently read with some surprise that certain US AI leaders (Microsoft, Meta, Nvidia) appeared to have switched sides, championing open-source AI (or adapted open source/open weights). This shift stems not from sudden altruism, but from a purely economic, competitive strategy. They now seem to be advocating for the very approach Chinese companies have long employed.
It is not that they have changed their minds; rather, they have realized their business model lies elsewhere. By taking this step, they can undermine competitors who are currently ahead in AI development, such as Anthropic and OpenAI.
Let’s look at the difference between the two AI models:
- Closed AI models (Anthropic and OpenAI) protect their commercial advantage and charge usage licensing fees.
- Open models (open-weight models)—backed by companies like Microsoft, Meta, and Nvidia—aim to create ecosystems dependent on their hardware, cloud services, and platforms.
A clear objective is to disrupt the current leaders. When OpenAI and Anthropic took the lead with massive closed models (such as GPT-4 or Claude), companies like Meta fell behind in that direct market. By releasing competitive open-source models (like Meta’s Llama family), these companies effectively commoditize their competitors' products. If any company can use an excellent open model for free or at a low cost, the value of paying for an exclusive OpenAI subscription drops drastically—a tactic previously used only by Chinese companies.
For some, the real model isn't the AI itself, but the infrastructure supporting it. For companies like Nvidia or cloud providers (such as Microsoft’s Azure or Amazon’s AWS), the true business lies not in charging for chatbot usage, but in selling the chips (GPUs) and computing power required to train and run AI. Open source encourages thousands of developers, universities, and startups to build their own systems, exponentially driving up sales of Nvidia hardware and the consumption of cloud server resources.
On the other hand, developing and refining a closed AI system is extremely costly. By releasing model weights to the global community, millions of independent developers optimize the code, identify security flaws, and create derivative tools—all for free. Later, major tech companies integrate these community-generated improvements back into their commercial cloud platforms.
There is an intense debate in Washington, where the North American administration and companies favoring closed models have suggested restricting open AI for national security reasons, citing China's advancements. In response, firms such as Meta, Microsoft, and Palantir have signed manifestos arguing that open source is vital for the U.S. to win the technology race. Their argument is that locking technology within a closed ecosystem would weaken Western cyber defenses and slow down innovation compared to Asian open models.
Let’s not kid ourselves: everyone is pursuing their own economic and commercial interests. And it is clear that Chinese AI models are closing the performance gap with their North American counterparts.
AI is still in its infancy—despite its rapid development—and open models are well-suited for deployment on personal computers with high computing power and memory capacity, a capability Nvidia has already begun bringing to market.
I won't comment on Google, as it is positioned to play both sides. It possesses a powerful closed system (Gemini) as well as specialized chips that can be used in open systems. It could play both sides.
Ultimately, AI continues to evolve at breakneck speed, with all players positioning themselves to their best advantage—and, of course, without overlooking Chinese companies.
Time will tell who stands to benefit the most.
EN 
ES