Non-U.S. AIs Seek Open-Weight Advantage in Market-Share Matchup

Chart of the Day

Good morning!

The French AI startup Mistral has just released its latest large language model, Mistral Large 3. Mistral has chosen to make the model open-weight, hoping this will encourage enterprise adoption and thus help it gain market share. 

It’s not an unreasonable strategy. After all, it seems to be working for China’s AI companies. The Financial Times, citing an MIT study, reported last week that Chinese open-weight AI models accounted for a 17% share of downloads in the 12 months ended in August 2025, compared to 15.8% for downloads of U.S. models, such as those from OpenAI, Google, and Anthropic. Though it’s obviously hard to beat “free,” it turns out that open AI models provide advantages to enterprises that go beyond mere cost.  

Think of a closed-source AI model like a high-performance car that you can drive, even though its hood is sealed shut so you cannot examine the engine. An open-weight AI is like a car with an engine that can be examined, and to a certain extent, modified and customized if so desired. (An open-source AI would be like a car that comes with a complete set of blueprints and instructions for how to manufacture and assemble each of the components to make another car from scratch.)

Mistral’s decision to mimic Chinese competitors in adopting an open-weight strategy means that enterprises can more easily and affordably modify them to their individualized requirements. Furthermore, because Mistral is French, the company’s adoption of such a strategy offers the added advantage of largely mitigating the geopolitical concerns (security, bias, censorship, etc.) that might bar some U.S. companies from adopting Chinese AI models. For now, OpenAI, Google, and Anthropic appear to remain committed to a relatively closed-source business model that monetizes through customer subscriptions and enterprise deals. 

This sparks a strong sense of déjà vu in the technology sector, where open-source software has long spurred a more rapid rate of adoption and ultimately a greater market share, despite a less direct and less immediate route toward monetization.

We can see this in Google’s Android mobile operating system, which is open source. This has encouraged smartphone manufacturers to use Android as the foundation for their devices, resulting in Android becoming by far the most widely used mobile OS in the world (72% versus 27% for Apple’s iOS). Yet despite Android being free, Google has found a robust, if roundabout, way to make money from Android, most visibly through the integration of Google Play Store and Google apps such as Gmail and YouTube. 

There is also precedent to be found in the pre-smartphone era. Outside the world of IT professionals, Linux is mostly known as a niche, open-source operating system favored by hardcore amateur tech enthusiasts. Yet Linux enjoys a market share on enterprise servers conservatively estimated at 63%, versus 27% for Microsoft. (Smaller alternatives like Unix account for the remainder.) Every single one of the world’s top 500 supercomputers (as ranked by performance metrics) runs Linux. Why? In large part it’s becausethe ability to easily tailor a tech solution has great appeal to businesses and institutions. (Fun fact: Android is partially derived from Linux.)

Like Android, Linux itself does not make money. However, Red Hat, a provider of Linux-oriented services and customization solutions, was notching annual revenues of $3.4 billion before being acquired by IBM in 2019, and it remains a major revenue driver for Big Blue’s software division. (IBM does not disclose Red Hat’s contribution to its top and bottom lines.) Oracle and Microsoft also include Linux-oriented services in their offerings – in fact, Linux is a core component of its Azure cloud platform. 

Obviously that’s not to say that a closed-source business model is a dead-end street. Apple isn’t exactly hurting for money despite iOS being a closed system, and obviously Windows has been a major success for Microsoft. Still, the growing adoption of open-source AI models, particularly those developed in China, has potential implications for tech investors – though not likely in the near term. 

For example, because Chinese AI models are generally less reliant on bleeding-edge processors, widening adoption could someday start to erode demand for the best, most expensive chips made by Nvidia, AMD, and other nascent competitors. On the other hand, such models could spark opportunities for companies who are capable of customizing them, similar to what Red Hat does for Linux. If the trend we just described sustains or continues, these are considerations that investors with long time horizons might wish to keep in mind. 

Share your thoughts

Are U.S. AI companies making the right choice in focusing on a relatively closed-source business model? Click here to send us your response.

📧✍️Here’s what a reader commented📧✍️

Q:  Should Big Tech and AI companies be subject to more antitrust scrutiny, or less?

A: Less! Larger companies have the resources-funds and talent- to be competitive, not only against established players, but also against China or others that do not have the constraints of an FTC. Bigger is better. 

Catch up with FS Insight

We believe the bearish sentiment that we have seen for much of 2025 reflects the fact that many investors sold at the tariff lows and missed out on many of the 10 best days. By the way, many of the 10 best days happen in the final weeks of the year.

Technical

The U.S. dollar has begun to turn lower in the last week, and it’s thought that both U.S. Treasury yields and the U.S. dollar should begin to weaken with greater velocity post this month’s Fed meeting. Overall, given the market’s snapback from late November, technicals have improved to match some of the bullish seasonality thought possible for this month.

Crypto

Bitcoin led yesterday’s rebound on strong volume. This suggests better odds of a sustained move and provides confirmation that the idiosyncratic seller is gone. Meanwhile, U.S. spot demand appears to be returning as the Coinbase–Binance spread continues trending higher.