The tiger is in the library
Can Mexico get in on the AI gold rush?
“This time around, there will be no need for a great conflagration. The barbarians don’t have to burn the books. The tiger is in the library.”
—Susan Sontag, from a letter to Jorge Luis Borges (1996)
Welcome to The Mexpatriate.
Today I’m going to talk about AI, which as writer Freddie DeBoer recently put it, is now “in the metadiscursive phase, when half of the discourse is about the discourse.”
As you may have guessed from the opening quote (you can, and should, read the entire Sontag letter here—it is eerily prescient), I lean skeptical on AI and yet, the staggering amount of blood and treasure being thrown at it cannot be ignored. In fact, it appears we’re already living in an AI investment-driven economy.
As journalist Derek Thompson wrote yesterday: “This is either the biggest tech-infrastructure project since the 1960s (since the beginning of the computer age) or the 1880s (the heyday of the railroad age).”
My hunch (grounded on zero expertise) is that the use cases of AI, particularly Large Language Models (LLM) and chatbots, that are dominating the discourse today will lose relevance in the future. OpenAI launched GPT-5 yesterday and while promising to be faster and more accurate than its predecessors, it continues to boldly bullshit (ahem, hallucinate), according to users. In the presentation itself there was a “mega chart screwup”—in Sam Altman’s words—that showed GPT-5 performing worse than OpenAI o3 on “deception evals” though the bars also didn’t match the percentages. Maybe someone forgot to say “please” when asking the stupid clanker to generate the chart. To err is human…right?
Meanwhile, Altman said an “eye-watering” amount of investment is needed to continue to advance. Of course.
Just a glance at recent headlines is enough to make you wonder if there is any hope for artificial intelligence to become less flawed, error-prone or malicious than the human kind:
“Google’s healthcare AI made up a body part”
“The good, the bad and the completely made up: Newsrooms on wrestling accurate answers out of AI”
“Perplexity AI acts like North Korean Hackers, Ignores Scraping Blocks”
“Google’s Gemini AI has an epic meltdown when it fails to complete a task and calls itself a ‘disgrace to all universes’”
Putting aside concerns about the environmental implications, job destruction and dystopian uses in government surveillance for now—at a minimum, this miracle product isn’t quite mission-ready.
In the last week, two AI announcements have been made by the Mexican government: one with Nvidia and the other with Amazon. Now that I’ve aired my own bias, I’ll take a look at what the country is doing to try to keep up with the AI boom.
But first: a favor, dear readers. Please fill in the polls below to help me decide what comes next for The Mexpatriate. Thank you for reading and enjoy your weekend!
Looking for unicorns
“AI is the modern alphabet. Either we hurry up and master that alphabet, or it will be a huge disadvantage for us in the new order emerging today.”
Flanked by the head of Nvidia Latin America and the national business council (CCE), Secretary of Economy Marcelo Ebrard announced the government’s plan to build Mexico’s own LLM last week, promising that AI will be a “key motor” for the country’s economic development. While Nvidia is not investing in the project, the company is collaborating with the government on training and development.
Other examples of government-backed national LLM projects include China’s Deepseek, Chile’s Latam GPT and France’s Mistral—whose chatbot Lucie was taken offline at the beginning of the year after starting to blabber about “cow’s eggs” as a “healthy and nutritious food.” Hopefully Mexico’s chatbot won’t start recommending huevos de vaca as part of a balanced diet.
“We are organizing efforts so that in two years, Mexico has visible and tangible results in this technological transformation,” said Ebrard. In November, Nvidia and the government are hosting an AI conference in Mexico City and will provide more details about the LLM. But does Mexico have what it needs to develop it?
If you’d like to support The Mexpatriate, please consider upgrading to a paid subscription.
Building Latam GPT, which has been in the works for over two years, has required “ultra-high-capacity infrastructure, specialized talent and relevant data sets” according to a Chilean AI company CEO quoted recently in Rest of World.
Mexico is second only to Brazil in the number of data centers in the region, and the industry forecasts installed capacity will reach 1.5 GW by 2030 (coming with around US $18 billion in investment). Brazil currently has 1.2 GW installed capacity; the United States houses a whopping 44% of global capacity (53.7 GW).
While there is palpable public-private sector enthusiasm for data center expansion—expected to create over 20,700 direct jobs by 2030–there are also concerns. In March, President Sheinbaum announced that Mexican company Fermaca Dreams will invest US $3.7 billion in a 250 MW hyperscale data center with its own power plant, fertilizer plant and gas pipeline in Durango as part of Plan México. Data centers are notorious for their water consumption—massive capacity centers can use at least 500,000 gallons a day—and Durango has suffered from extreme drought in recent years, as have other states where massive data centers have been built (such as Querétaro and Guanajuato). While there are cooling system alternatives that use less water, they tend to be electricity-intensive and therefore require cheap and reliable power. The Mexican Data Center Association (MexDC) estimates that to meet the energy demands of data centers in Mexico will require US $8.8 billion in electrical infrastructure investment by 2029.
And what about the talent?
At the AWS Summit 2025 in Mexico City this week, Amazon said it plans to train 495,000 Mexicans in AI skills over the next three years, as part of the 15-year US $5 billion investment in the country already announced at the beginning of the year.
“Looking for unicorns for Plan México” read one El Financiero headline, calling out the invitation by the director of AWS Mexico, Rubén Mugártegui, for Mexican startups and small businesses to participate.
According to survey data published by AWS this week, Mexico is at a “defining moment in its AI-driven digital transformation” and 38% of Mexican businesses today are “AI-adopting firms” (defined as “a business that consistently uses at least one AI tool”), up from 29% last year. The survey also found 68% of business owners think AI will transform their industry in the next five years and only 19% say they have a strong AI skillset.
I’ve just scratched the surface on this topic, so stay tuned for more about it soon.
Questions, criticism or tips? Email me: hola@themexpatriate.com. And if you enjoyed this newsletter, please share it.