The Coatlicue Supercomputer And Mexico's AI bet

Mexico is making it clear that artificial intelligence (AI) is key to the country's economic growth and development in the coming years. The Claudia Sheinbaum administration has announced a series of measures and projects aimed at fostering a technological revolution, including a supercomputer and training center for this technology.

In November, Sheinbaum presented the Public Training Centre for Artificial Intelligence, a strategic initiative which is part of the "Mexico, Country of Innovation" project that seeks to establish itself as the largest public AI school in Latin America and strengthen the country's digital talent pool.

Weeks later, she unveiled the Coatlicue supercomputer project. It marks a turning point for technological development in Mexico, as the government seeks for it to become the most powerful supercomputer in Latin America and one of the most ambitious public investments into artificial intelligence.

Supercomputers -- essentially thousands of powerful processors operating in parallel to conduct complex calculations at high-speed -- have existed since the 1960s and today, the most powerful often reside in countries like the United States, China and European countries part of the EU.

In Latin America, only Brazil and Argentina have supercomputers listed in the TOP500 project, which ranks the most powerful supercomputers in the world.

For the Mexican government, Coatlicue will mark a turning point in its capability. With an estimated power of 314 petaflops, its performance will be seven times more powerful than Pégaso, Brazil's largest private supercomputer, and will have 100 times more processing capacity than Yuca, currently Mexico's most advanced system.

While still far behind other world AI leaders, the move could place Mexico within the conversation of nations with supercomputing capabilities. Already more multinationals have been looking at the country, with Ness Digital Engineering, which is owned by KKR, opening its new LatAm headquarters in Guadalajara that will be focused on AI-powered intelligent engineering.

With a public investment of 6 billion pesos (around $330 million USD), Coatlicue reflects a strategic decision that goes beyond rhetoric and translates into critical infrastructure for the digital economy.

"Coatlicue represents a declaration of intent by Mexico to the world: to compete in the knowledge economy with its own talent," Fermin Muela, leader of the Lottus AI Program, an educational platform in Mexico focused on AI, told Latin Times.

"Its success will depend on strategic and transparent management that, with a long-term vision, promotes AI and a new era of technological competitiveness."

Mexico's strategy to promote AI

Speaking from the country's National Palace in October, Sheinbaum told executives from 17 World Economic Forum member countries that the "Mexico, country of innovation" project would include four main pillars:

• Firstly, greater scientific, technical and humanistic training from childhood through to higher education;
• Second, a National Innovation Driver with the Development Bank to finance small and medium-sized enterprises, enabling greater innovation;
• Third, engineering services;
• and lastly, the National Artificial Intelligence Laboratory to create all the conditions necessary for the use of the technology by the government and private sector.

'With these initiatives, Mexico sees the possibility of positioning the country as a technological leader in Latin America as a strategic opportunity," Belén Ortega, an AI specialist from Argentina, told Latin Times.

She noted that "a plan of this magnitude not only sets a clear direction, but also sends a very powerful signal to the business, academic and government ecosystems: Mexico is serious about innovation."

For some in academia, however, the announcement of Mexico's supercomputing plans are met with cautious optimism.

"Mexico's decision to invest in supercomputing could have a major impact on the innovation ecosystem in the country if all key players are included in its orbit," Odille Sanchez, leader of the Tech and Scientific-Based Entrepreneurship Center at the university Tecnológico de Monterrey, told Latin Times.

"Academia, private companies, government, service providers, capital sources, and other ecosystem connectors stand to benefit from access to advanced AI models generated inside the country and not abroad," she said.

Warnings about Coatlicue

Coatlicue has not only sparked excitement within the scientific and technological community, it has also caused debate.

Luis Areán, a research engineer at the French National Centre for Scientific Research (CNRS), outlined problems he sees with the Coatlicue supercomputer on LinkedIn, pointing out four main challenges that jeopardize its implementation and real impact.

The first problem stems from the planning stage. Areán questioned the project's advance consideration of potential applications without a formal approval protocol. In scientific research, he explained, projects require prior evaluation by a specialized committee. Without this filter, Coatlicue lacks a solid methodological foundation to guarantee relevance, feasibility, and scientific value.

The second challenge relates to the use of the supercomputer to detect tax fraud -- a use case that was touted by the Mexican government. Areán clarified that this task does not require supercomputing, but rather an efficient data processing platform like the one already used by the country's Tax Administration Service (SAT).

The third problem arises from Coatlicue's application for data mining -- essentially analyzing massive databases for new insights.

Areán argues that this type of task is not usually part of traditional supercomputer projects.

"Data mining doesn't demand as much computing power as the ability to process massive volumes of information in a very short time using mapping and reduction algorithms," he wrote.

The final concern is related to who will use and benefit from Coatlicue. More transparency into its financing and access to its benefits need to be published, he said.

It's also lacking "participation of researchers from universities and other institutions in the planning process," he argued.