The National Polytechnic Institute (IPN) is a strategic link in the Coatlicue supercomputer project. The training of talent in high-performance computing is undoubtedly one of the most important contributions of this institution.

In November 2025, the Mexican government announced the Coatlicue project, which within two years will give Mexico the most powerful supercomputer in Latin America. The project aims to accelerate data-driven public policies and boost startups and corporations with computing capabilities previously inaccessible in the country.

To understand the significance of this announcement, it is worth noting that during the pandemic emergency (2020–2023), approximately seven million deaths were reported worldwide; while part of the population stayed at home, vaccines were developed in just under a year, even though their normal development takes approximately a decade.

This technological feat was made possible by high-performance computing, which was essential for understanding the virus’s behavior and accelerating testing. Supercomputers such as Summit (United States), Fugaku (Japan), and MareNostrum 4 (Spain) performed simulations that allowed for the analysis of billions of molecular configurations in a matter of days, to identify the most promising ones and test how different compounds could bind to the virus’s proteins to disable it.

Coatlicue is a project led by the Agency for Digital Transformation and Telecommunications (ATDT) and the Secretariat of Science, Humanities, Technology, and Innovation (Secihti), and among the institutions strengthening this initiative is the IPN.

SUPERCOMPUTERS

A typical computer has a CPU (Central Processing Unit) with between four and eight cores to perform tasks. A supercomputer has thousands of GPUs (Graphics Processing Units), each of which has thousands of cores to carry out tasks in a coordinated manner and much more quickly.

On this topic, in an interview with the Conversus News Agency (AIC), Luis Alfonso Villa Vargas—who, along with Marco Antonio Ramírez Salinas, both from the Center for Computing Research (CIC), coordinates the technical contribution of the polytechnic to the project—explains: “A supercomputer can be understood as thousands of laptops connected in such a way that a task is assigned to them to be executed in a distributed manner across all the computers—or with one part running on some processors and another on others—which allows a problem to be solved in much less time”.

“Imagine a pool of thousands of processors ready to work and connected in such a way that they can communicate with one another. To send a task to that vast pool of processors, there are models known as execution queues (job queues). When it’s your program’s turn to run, the queue dispatches it to those thousands of processors, but it is the programmer who defines how to use the processors, and when it finishes, you receive a result—it could be very quickly, or it could take weeks or even months of execution”, explained Villa Vargas. It’s important to understand that a supercomputer isn’t for writing documents or opening images; that enormous computing power is used for tasks that require processing thousands or even millions of data points and performing a similar number of calculations—tasks that would take a conventional computer months or years to complete.

They are currently widely used in the energy industry, weather disaster prevention, and the search for new materials; in the healthcare sector, they are used to develop new vaccines and drugs and to build models to better understand Parkinson’s and Alzheimer’s, to name just a few.

ABOUT COATLICUE

According to the information that has been shared, the Mexican supercomputer will have 14,480 GPUs—equivalent to about 375,000 conventional computers operating simultaneously—and will reach 314 petaflops, meaning it will be capable of performing 314 quadrillion operations per second.

The TOP500 is an international ranking of the world’s most powerful supercomputers published twice a year, in which, for Latin America, only Brazil appears with 10 machines and Argentina with one. Coatlicue’s capacity would even surpass Harpia, Brazil’s most powerful computer, which only began operations in October 2025.

To tackle a challenge of this magnitude, the ATDT and Secihti formed an alliance with the Barcelona Supercomputing Center (CSB), which hosts three Spanish supercomputers in the TOP500. Regarding this collaboration, Luis Villa noted that “to define the architecture of this computer, the ATDT formalized a strategic partnership with the CSB, which is undoubtedly an international leader in the field, so the architecture and operating model are being defined within the framework of this partnership”.

For a supercomputer to function properly, a significant amount of electrical power is required, as well as a water supply required for the cooling system. The project is currently in a strategic phase regarding the decision regarding its location, as the supply of the approximately 20 megawatts it requires and a water system for efficient cooling must be taken into account; issues such as connectivity and cybersecurity are also being considered.

Regarding these last points, Villa Vargas explained that “educational institutions have attractive connectivity contracts, but they are not a critical service like Amazon or the commercial operations of an airline, since if their systems lose connectivity for five minutes, tens of millions of pesos can be lost in just five minutes. For a supercomputer, connectivity can become critical, depending on the importance of what is being processed, so a service must be in line with those needs. Regarding security, while secure connectivity is necessary, the connection with other computers inside and outside Mexico for data sharing follows an international standard for IT security”.

This year, infrastructure issues will be resolved and contracts will be signed so that the components arrive in Mexico next year. “There is enormous global demand for processors; if Mexico is to acquire a computer the size of Coatlicue, it must place the order about eight months in advance. At this point, the technology to be acquired is not yet on the market because it evolves so rapidly”, he explained.

After the components arrive, several teams of experts will need to work on their installation and testing, a process that takes several months because it is a meticulous task.

“The processing units come pre-mounted in racks, which are similar to cabinets. Each unit has dozens or hundreds of processors, and if it has GPUs, it has thousands of processors, and they must be connected so that they can all communicate with each other. Companies with certified personnel follow an architecture to connect all those processing units to central connection points, which in turn connect to other connection points, and this is how the networking process is carried out. It takes months of working cable by cable to make the connections; electrical connectivity is a process that requires a separate procedure, and the water-cooling system is connected by another group of specialists”, noted the polytechnic researcher.

THE MAJOR CHALLENGE With Coatlicue, the Mexican government is undertaking one of the most significant scientific and technological projects for Mexico. However, this significance comes with major challenges, such as achieving results that positively impact society and ensuring that, at the national level, there is a full understanding that the financial investment made was worthwhile.

The IPN plays a fundamental role in ensuring that this valuable tool yields results that impact citizens’ lives. “Using the supercomputer will require qualified personnel; the Polytechnic is the institution that conducts the most research in engineering and computing in the country, and it trains young people with hight-level skills in computer science. We have 50 researchers working in the field of supercomputing”, he emphasized.

The need for supercomputing in research in Mexico is an issue that used to be addressed with resources from abroad countries such as the United States or France. In this regard, the Polytechnic Institute has had a long-standing partnership with the CSB, and there is also a plan in place to maximize Coatlicue’s potential. There is an agreement for talent development between the Barcelona-based institution and SECIHTI, in addition to strategies already established at the IPN.

Villa Vargas noted that “the partnership with Barcelona has two main components: two-year postdoctoral training at the CSB and doctoral training at the Polytechnic University of Catalonia. The Polytechnic implements a model that allows undergraduate students to specialize in supercomputing, primarily through courses in supercomputing programming and algorithm analysis. We are not new to this field; we have had projects with the Mexican Petroleum Institute (IMP), where the objective was to train specialized human resources in high-performance computing”.

THE CROSSROADS

In Mexico, with a population of over 130 million, enormous amounts of data are generated that can be analyzed and provide valuable insights for public decision-making. Which areas could benefit from this powerful tool? Petróleos Mexicanos (Pemex) has been a strategic company for Mexico for decades and would be one of the main beneficiaries of Coatlicue, as the use of this technology is fundamental in oil and gas extraction.

“I worked at IMP for five years, and the head of the exploration department would tell us, ‘If you give me a model that lets me increase the probability of finding oil by 10% when we drill, that alone would pay for all the research you would ever need to conduct. Huge amounts are invested in drilling—especially in deepwater—and sometimes you don’t find any oil, or you only find a very small amount. But if you do it using a model that guarantees an increased probability, using supercomputing, the savings are enormous—enough to pay for dozens of machines of this capacity”, he stated.

Some potential applications of supercomputing for Mexico include the development and consolidation of wind energy models, the optimization of electricity use, healthcare, materials science, and disaster prevention efforts. It is no coincidence that Brazil has 10 supercomputers in the TOP500 and has been able to diversify its economy and consolidate its domestic market to emerge as a regional leader.

Following the initial phase of acquisition, installation, and commissioning of Coatlicue, it will be essential to continue implementing an upgrade plan, as this would help drive the country’s development, grounded in research and technological development.

Dr. Villa stated that “if the project progresses in a similar way to what has occurred in other countries, Coatlicue will need to be updated every five years. In 10 or 15 years, we could have nodes with significant computing capabilities in different parts of Mexico, since Coatlicue 1 could be distributed across nodes across the country, so that when the upgrade to Coatlicue 3 takes place, after 15 years, all those machines would strengthen the country’s computing infraestructure”.