In a cutting-edge laboratory nestled in Santiago, Chile’s vibrant capital, dedicated researchers are racing against time, eager to secure their nation’s place in the rapidly expanding artificial intelligence revolution.
Meanwhile, in the bustling streets of Cerrillos, a neighborhood skirting Santiago’s southern edge, community activists are vehemently opposing the construction of massive data centers—the very infrastructure that fuels AI’s capabilities.
Within the walls of the presidential palace, government officials are carefully strategizing how to elevate Chile’s technological standing, aiming to expand its role in the AI sector with limited funds, all while safeguarding vital resources and maintaining public support.
Across Chile, these intense political debates have ignited over artificial intelligence. This has transformed the arid South American nation of 20 million people—a country seldom at the forefront of global tech discussions—into a compelling case study of a society striving to navigate the complex trade-offs inherent in the AI race.
Chile has actively sought investments, nurtured local talent, and is diligently building its capacity for AI. These initiatives hold the promise of significant economic growth but also pose threats to the environment and risk deepening dependence on dominant U.S. tech giants. Chilean officials have put forth a plan to regulate new data centers, a proposal that has sparked widespread protests and, most recently, heated debates in Parliament.
Many Chileans, who often view artificial intelligence with skepticism or indifference, are questioning whether the benefits truly outweigh the costs.
“AI is being fetishized, treated as a new obsession,” remarked Rodrigo Cavieres, a key figure in the Socio-Environmental Community Movement for Water and Land (MOSACAT), which has actively protested against major tech corporations. “Data centers are being prioritized over the well-being of our population.”
In Santiago, protesters have demonstrated against data centers, forcing Google to withdraw plans to build a second site in Chile.
These tensions mirror similar conflicts unfolding worldwide. Many nations, from the United Arab Emirates to the Netherlands, face the challenging decision of risking excessive investment, environmental strain, and public disapproval in the pursuit of AI, or the grim alternative of being left behind.
These debates often originate from pivotal moments, much like the one experienced in 2023 by Álvaro Soto, director of the Chilean National Center for Artificial Intelligence. That year, he realized Chile risked being marginalized in the AI landscape when he experimented with an early version of the ChatGPT chatbot, querying it about Chilean literature.
ChatGPT predominantly attributed Chile’s rich literary heritage to Pablo Neruda, the celebrated 20th-century poet. This, Mr. Soto noted, underscored a significant issue: AI models were not being developed to accurately reflect the diverse cultures and languages of regions like Chile.
Today, Mr. Soto and his research team are meticulously training their own AI model using previously overlooked data from across Latin America. In June, Chilean President Gabriel Boric declared in his state of the union address that the nation must embrace AI. His administration is actively working to simplify the process for foreign companies to establish data centers and to integrate AI tools into daily governance.
However, despite strong political will, communities directly impacted by AI data centers remain profoundly dissatisfied. In northern Santiago, a community group is protesting an Amazon site, perceiving it as environmentally destructive gentrification. Nearby, another group is demonstrating against a Google data center, fearing its potential impact on a local wetland. A third group, operating in Santiago’s southern outskirts, successfully compelled Google to retract its plans for a second data center in Chile.
In response, President Boric’s government now intends to redirect data center construction away from Santiago toward the more sparsely populated northern regions. Many environmentalists, however, are deeply concerned about the potential impact on the ecologically sensitive Atacama Desert, an area already scarred by extensive mining operations.
“There are moments in Chile when it feels like we’re glimpsing the future,” observed Marina Otero, an architect and Harvard lecturer specializing in data centers. “The battle over AI is far from over. It’s a harbinger of what’s to come.”
The Ecstasy and Agony
Matías Muchnick, the chief executive and founder of NotCo, at the company’s Santiago laboratory, which uses precise measurements of food preparation to feed an A.I. recipe engine.
In a state-of-the-art kitchen lab in Santiago, employees at the food-tech startup NotCo were intensely focused one morning on a critical mission: leveraging artificial intelligence to solve a pressing issue for major food and snack brands—how to replace food dyes recently banned by U.S. health secretary Robert F. Kennedy Jr.
NotCo employs an AI-powered engine, meticulously trained on molecular data, to reinvent ingredients for some of the world’s largest food brands. To develop this groundbreaking technology, the company, now valued at $1.5 billion, required computing power far beyond what was locally available. Consequently, Matías Muchnick, the chief executive and founder, reached out to Google, which, in 2018, granted NotCo access to specialized microchips designed for AI applications.
“We aspired to build a research and development powerhouse, and that access made our journey significantly easier,” Mr. Muchnick affirmed.
NotCo has emerged as a shining example of how Chilean businesses can effectively harness AI. However, the very act of constructing the necessary infrastructure to support such companies has triggered considerable public outcry.
NotCo’s offices in Santiago.
NotCo has become a sign of Chilean technological knowhow.
In 2015, Google established its first Latin American data center in Quilicura, a community on the outskirts of Santiago, adjacent to a vital wetland. According to environmental records submitted to the government during the project’s planning phase, the site consumes 50 liters of water per second—an amount roughly equivalent to the usage of 8,000 Chilean households—to cool its vast array of computers. A Google spokeswoman later clarified that the site used considerably less water last year, roughly the amount consumed by a typical golf course.
Historically, older data centers often rely on water evaporation to cool overheated computers. More recently, companies have engineered advanced technologies to conserve and recycle water, though environmental advocates caution that many data centers still utilize large amounts of water.
Rodrigo Vallejos, a dedicated local activist, shared video footage depicting the area’s past: lush marshlands and serene lagoons. Today, much of it lies parched, even during the rainy season.
For Mr. Vallejos, the trade-off has been profoundly unequal. The data center provides few local jobs, he argued, and the designated community “offset”—a park situated beside a cemetery—sees minimal use.
“Ultimately, we risk becoming little more than an artificial intelligence warehouse for the rest of the world,” he lamented.
Rodrigo Vallejos, an activist and a law student, has taken a leading role in the protection of the Quilicura wetland.
In many respects, Chile is already serving this role. The country boasts 33 data centers, making it a regional AI hub, and this number is projected to double by 2030, according to Chile Data Centers, an industry association.
Many Chileans are only just beginning to grasp the functions of these facilities and their broader effects. When Google unveiled plans in 2019 for an additional data center nearby in Cerrillos, many residents initially anticipated new job prospects. However, environmental filings soon revealed a sparsely staffed operation that would consume an estimated 228 liters of water per second—an amount roughly equivalent to the consumption of 40,000 households.
Tania Rodríguez, a determined resident, embarked on a door-to-door campaign to voice concerns about the project, successfully mobilizing her neighbors to oppose it. Last year, Google announced it would withdraw its plans.
Despite this victory, Ms. Rodríguez remains cautious. Representatives from a local Google subsidiary recently initiated new community outreach efforts regarding yet another potential data center.
“We are not against artificial intelligence—it’s a technology we absolutely need to develop,” Ms. Rodríguez stated. However, she intends to continue protesting against Google because “we must use our natural resources efficiently, as that is fundamental for human survival.”
The Astronomy Model
The Vera Rubin Observatory in Cerro Pachón, Chile. In the 1990s, foreign astronomers vied to build telescopes in the country’s deserts.
In 2024, Sebastián Howard, an official within the Ministry of Science, Technology, Knowledge and Innovation, discovered that tech companies were planning an additional 30 new data centers over the next four years. Almost all of these were slated for the Santiago area, which is already severely impacted by drought.
“We lacked the energy to support them,” Mr. Howard stated, referring to the capital region. “Most critically, we simply didn’t have the water.”
Mr. Howard spearheaded a government initiative to redirect data center development elsewhere. He and his team devised a tool to identify locations where these facilities would incur the least environmental and social impact. Their research led them to Antofagasta, a northern desert city abundant in solar energy.
Their strategy draws inspiration from Chile’s successful astronomy model. In the 1990s, as international astronomers competed to construct telescopes in the country’s famously clear-skied deserts, the government implemented a pivotal rule: 10 percent of telescope observation time had to be allocated to local researchers. This policy propelled Chile to become a global leader in astronomy.
Mr. Howard and his colleagues envision a similar future for AI. “If these companies are willing to invest here, we need to ensure that their infrastructure also benefits our universities and local businesses,” he explained.
In many parts of the world, countries concerned about AI access have chosen to build their own data centers. Under Chile’s proposed plan, however, local companies and universities would instead gain access to the computing power established by foreign corporations.
Selling this vision has proven difficult. Many perceive the government’s plans as an attempt to favor large corporations at the expense of public interests. Activists have highlighted a controversial environmental rule change that could diminish transparency regarding water and electricity consumption.
At a workshop last year, intended to explain AI development to community leaders, Mr. Howard faced vocal opposition after stating, “It’s a privatized country. These companies can do whatever they want.”
“The moment you lose the capability to understand how your machine is working, or the ability to even build your own machine, that’s the moment you lose,” said Aisén Etcheverry, Chile’s former minister of science, technology, knowledge and innovation.
Whether tech giants will adopt the plan to relocate data centers north remains uncertain. Felipe Ramírez, who oversees Amazon Web Services in Chile, raised concerns that placing data centers nearly 680 miles from Santiago could lead to significant internet lag—a critical issue for voice-based AI, where even a millisecond’s delay can alienate users.
“It seems logical on paper, but in reality, I’m not sure if we’ll end up training global models in Antofagasta,” he commented.
Nevertheless, the government is forging ahead. Aisén Etcheverry, a presidential adviser and former minister of science, technology, knowledge and innovation, confirmed that discussions with major firms are ongoing. She expressed hope that a final plan would be announced by the end of the year.
If Chile fails to influence how AI systems interpret its language, history, and institutions, she warned, it risks being excluded from the future.
“The moment you lose the ability to understand how your machine functions, or even the capacity to build your own, that’s when you lose everything,” she asserted. “And we absolutely do not want that to happen.”