The idea that France finally beat the United States in the AI race has been making the rounds for days. Before writing that sentence, I went through the benchmarks one by one, and the honest answer is: yes, but not the way you think, and not on the metric you’d assume.
What’s actually real
On October 6, Mistral —the French startup that’s become Europe’s main counterweight to OpenAI, Google and Anthropic— launched Mistral Large 4, nicknamed “Le Chonk”: a trillion-parameter model that will go open-weight once it releases its weights later this month (TechCrunch). And yes, there are two concrete tests where it beat closed American models. On satellite-image object localization (DIOR-RSVG), it scored 73% against GPT-6 Astra’s 68%. And on the vulnerability-patching test in Artificial Analysis’s Cyber Index, Claude Opus 5.5 and GPT-6 Astra —the two American models in question— scored near zero, because their own safety guardrails made them refuse the task, while Large 4 actually completed it (Trending Topics).
What nobody tells you
Both wins are real, but they’re niche, not frontier-wide. On Artificial Analysis’s aggregate Intelligence Index, Large 4 scores 38 —roughly tied with GPT-6 Luna and below Gemini 3.8 Flash’s 40.9. On coding, in a blind evaluation, it placed second of five, behind Claude Opus 5 (Trending Topics). And since the weights still aren’t public, there’s no independent verification of any of these numbers yet: everything we know today, we know because Mistral said so. In more than 15 years of reading corporate earnings releases, I learned to distrust exactly this pattern: the generous headline, the fine print still pending.
Europe’s real scorecard —and the Latin American question
None of this takes away from the effort. Brussels pushed through InvestAI this year (€200 billion, including €20 billion for gigafactories), a Chips Act 2.0, and a Cloud and AI Development Act aimed at tripling Europe’s data center capacity. France announced €10.9 billion in its own AI infrastructure, and Mistral raised roughly €3 billion at a €21 billion valuation (Servimedia). The money is real. So is the fact that 75% of European executives surveyed by Nokia say their current infrastructure limits their AI plans, and that Mistral’s own share of researchers on staff fell to 29% in the first half of the year. My read: Europe is in time —but to compete in niches like vision, cybersecurity and enterprise sovereignty, not to lead the general frontier this cycle.
And Latin America? This is the turn I find more interesting than the French one. The region isn’t competing to build the next leading model, but it is showing up forcefully as an infrastructure host: 443 megawatts of AI-dedicated capacity in 2026, projected to reach 1.6 gigawatts by 2031. Mexico signed a $5 billion AWS cloud region in Querétaro; Chile is structuring close to $4 billion in its data center strategy, leaning on submarine cables and renewable energy; and Huawei, citing JLL, names Brazil, Chile, Mexico and Colombia as the four countries set to concentrate most of the region’s digital infrastructure investment in the coming years (Bloomberg Línea). Could we ever be competitive? Not in the sense of building the leading model —that race is decided by compute capital, not willpower. But yes, in the sense of being indispensable to the race existing at all: energy, connectivity and land are as much a part of AI’s value chain as the algorithm.
The closing line I owe you straight
This is where geography stops being a postcard advantage. An XDI study ranks Mexico as the fourth most climate-exposed country among those with planned data centers worldwide —and Querétaro, the very hub AWS picked for its AI region, is flagged as a hotspot for that risk. In the southern United States, 51% of the $670 billion in planned data center investment sits in states at high risk from severe convective storms, like Oklahoma and Louisiana (XDI; MS Amlin). I couldn’t find an equivalent study for Chile, and I’m saying so plainly rather than making up the number.
So, will we be a safe haven? Not by default. The same renewable energy and connectivity drawing capital today coexist, in some of the same hubs, with real physical exposure that investor euphoria is underpricing worldwide, not just here. Nature doesn’t decide this race for us —but it does send the bill to whoever leaves it out of the spreadsheet. Before celebrating that Latin America is the next great AI hub because it has sun, copper and sea, it’s worth asking yourself: is your judgment taking in the full context, or just the half that’s convenient to believe?
If this kind of analysis helps you think through your own decisions, the Daily Briefing shows up every day with the same discipline. Subscribe here →