A few days ago, Dario Amodei, CEO of Anthropic, published an essay arguing that AI is advancing so fast that we need to slow down — not stop, but ensure safety keeps pace with capability. Within hours, Sam Altman of OpenAI responded: “I agree with Dario that we need to pace the frontier.” Elon Musk followed: “Dario is right.”

When those three agree on something, you pay attention.

Amodei’s warning is specific: “In 6-12 months, a swarm like the one in the OpenAI-Hugging Face incident could be capable of taking over the entire internet with a persistent botnet, potentially causing hundreds of billions of dollars in damage — and the scale of damage would continue to increase if AI becomes more powerful without the necessary guardrails.”

His proposed solution has three steps: embedded third-party evaluators inside AI companies, democratic coordination among allied countries on safety standards and eventually, global coordination. The phrase “democratic coalition” appears repeatedly in his framework.

That’s where I can actually say something useful.

U.S. Treasury Secretary Scott Bessent has warned that “nothing would matter if China wins the AI race.” Amodei agrees. A Chinese lead in AI, he argues, would pose grave danger, economically, but most importantly: militarily, through AI-driven weapons systems and geopolitical dominance. At its core, this is a national security conversation dressed in technology clothing.

And national security conversations have physical requirements. AI runs on hardware: servers, chips, data centers, power, cooling systems, specialized cables, transformers, processing units. It runs on supply chains. It runs on energy. It runs on engineers.

The United States needs all of those things faster than it can build them domestically, and it needs them from partners it can trust.

Which brings me to Mexico.

In the first half of 2026, Mexico’s computer-related exports reached US $82.9 billion, a 172% increase from the same period a year earlier, putting computing equipment ahead of automobiles and auto parts ($74.8 billion) for the first time in thirty years.

Mexico doesn’t export algorithms and code. It is building the physical layer that makes AI possible; the unglamorous, essential plumbing of the AI data center buildout. The United States accounted for 93.9% of those exports, which tells you everything about the strategic nature of this relationship.

This new nearshoring wave is not a buzzword; it is an AI supply-chain strategy embedded in a global security race. BTW: I’ve written about this before.

U.S. data center investment reached $102.2 billion in 2025, growing nearly 30% year over year. Mexico’s export growth tracks that figure almost perfectly — the architecture of an integrated continental co-production system responding to a common demand signal. Folks, this is actually okay! Not just that, it is strategic and necessary.

Amodei calls for cutting off powerful AI chips and semiconductor manufacturing equipment from China and cracking down on smuggling operations. He’s right. But that strategy only works if you have an alternative. Mexico — integrated into North American supply chains, operating under USMCA digital trade rules, with a manufacturing base that pivots fast — is a large part of that alternative.

Now, here’s the number that makes Washington uncomfortable: Mexico’s imports from Taiwan surged 220% in the first half of 2026. Computer equipment imports from Asia into Mexico are booming. Since those components are assembled into goods that cross north into the United States, this shows up, mechanically, as a larger U.S.-Mexico trade deficit.

Mexico in Numbers: How Taiwanese imports are fueling Mexico’s record tech exports

I can already hear the reaction in certain corners of Washington, D.C.

But this is precisely the kind of number that demands context before it demands a tariff. What is actually happening is that North America — through Mexico’s manufacturing base — is progressively substituting Asian-sourced inputs and redirecting them through a regional production platform that delivers the servers, components, and infrastructure American firms need to build AI data centers faster, cheaper and without the geopolitical risk of direct dependence on Taiwan or China.

A tariff on Mexican electronics would tax precisely the substitution away from China that U.S. policy is trying to achieve. It would be, in effect, a tax on America’s own AI infrastructure — and on America’s own capacity to win this race.

The deficit is not a sign of a broken relationship. It is a sign of an integrated continental production system doing exactly what it was designed to do.

The “democratic coalition” Amodei calls for doesn’t need inventing. It needs to be deepened — starting with the country that shares 2,000 miles of border with the United States, a trade agreement that explicitly covers digital trade, deep supply chain integration, and a talent base that is the envy of the world.

Every day the North American neighborhood spends fighting over tariffs, USMCA renewal mechanics and trade deficits that don’t mean what people think they mean is a day China runs faster on AI infrastructure, critical minerals, energy policy and diplomatic reach.

The question is not whether Mexico is relevant to the AI race. The question is whether the people making policy decisions in Washington are looking at the right numbers and the right strategic angle.

You don’t win a race by slowing down your most integrated and reliable partner.

Pedro Casas Alatriste is the Executive Vice President and CEO of the American Chamber of Commerce of Mexico (AmCham). Previously, he has been the Director of Research and Public Policy at the US-Mexico Foundation in Washington, D.C. and the Coordinator of International Affairs at the Business Coordinating Council (CCE). He has also served as a consultant to the Inter-American Development Bank. Follow his Substack here.