Earlier this year, the city of Rio de Janeiro announced its plan to transform the Olympic Park — home of the 2016 Summer Olympics — into a 3 gigawatt AI data center campus, leveraging Brazil’s existing network of high-speed fiber optic cables and abundant clean energy resources.
It’s a massive undertaking that the city hopes will put it on the map as a destination for the largest AI companies in the United States, where they are grappling with long interconnection queues and access to power. But attracting these global tech giants, who come with high uptime requirements, requires proving the stability of the local grid. In the last few weeks, that work has gotten underway, with the help of Tapestry, the Google X grid management project.
Elea Data Centers, the company developing the Rio AI City project, is the project’s first data center customer, but Tapestry general manager Page Crahan is sure it won’t be the last. “We hope that Rio is one of several cities that Tapestry can be a part of demonstrating what being AI-ready looks like,” Crahan told Latitude Media. The effort is twofold, she added: improving the quality of power for the customers of the local distribution grid, and ensuring the level of reliability that will allow data centers to move forward with confidence.
The project is a shift for Tapestry, which has to date been focused on partnerships with grid operators.
“Things have changed in terms of how we meet partners, customers, projects, and frankly, that has helped us rethink what we should prioritize,” Crahan explained. “The change in the narrative in inbound interest has been stunning…specifically around countries and communities who want to make sure that they are not left behind.”
When Tapestry first emerged from Google X, most of the inbound interest was from utilities looking for new tools to help them plan for future expansions of their grid. But Crahan said that has changed in the last two years: “It’s kind of a fever pitch narrative of ‘how do we build our grid to meet this AI moment?’”
Two-track deployment
Rio AI City is more complex than Tapestry’s prior deployments, in part because it brings together many more stakeholders: a data center developer, a distribution utility as well as a transmission utility, and elected officials. It was the city’s mayor, Eduardo Paes, who first reached out to Tapestry earlier this year. And the fact that Paes already had buy-in from the utilities and developers on the project was a major draw for the Tapestry team, Crahan said.
It’s an “atypical” introduction to a project, she added: “But I also think, given all that’s changing around us, and the kind of societal connection between energy and AI and technology, it wasn’t altogether that surprising.”
As of this month, Tapestry is already starting the Rio rollout of its AI asset inspection tool GridAware. That tool pulls images of grid assets from a range of sources including satellite and street-view imagery, as well as photos taken by field teams, and combines them into a single view to simplify inspections.
That’s a newer tool for Tapestry — the team first deployed GridAware in New Zealand, in partnership with the country’s largest distribution utility, and announced the results of that project over the summer — but Crahan said it’s the right place to start in Rio. “Regardless of what day the data center gets plugged in, Tapestry can start helping those distribution utilities deliver more reliable power…starting now,” she explained.
Some of the insights about the network that Tapestry gathers through deployment of GridAware will then inform the second tool being leveraged in Rio: the Grid Planning Tool, already being tested and deployed in Chile for the last several years and more recently in PJM, which enables large-scale, long-term simulations of transmission grids on an hourly basis. GridAware will give Tapestry better data about the existing infrastructure in Rio, and help determine the scenarios run via the Grid Planning Tool.
Speed to power
The goal in Rio, Crahan said, is to build a blueprint that other cities and countries can use to get their grids ready for AI load growth.
Things are moving quickly, thanks in large part to the fact that the city came to the table with buy-in from so many stakeholder groups, Crahan said: “We hope that Rio is one of several cities that Tapestry can be part of demonstrating what being AI ready looks like.”
“It must be the case that we can both get ready for AI load growth and deliver reliable, affordable power to our communities,” she added. Deploying Tapestry’s two tools together — working at both a distribution and transmission level simultaneously, as opposed to taking on a strict data center modeling project — is key to making that happen.
According to Crahan, two key metrics for success in Rio are whether the tools improve the quality of power for both the data center customer and the other customers of the local distribution utility (via the GridAware tool) and whether major loads like data centers have confidence in the grid. The third metric is speed to power.
In addition to giving grid planners higher confidence in their modeling, she added, Tapestry and Rio will be looking to see whether that process was faster and more efficient with the use of AI tools; in other words, can Tapestry help a data center get online faster?
Tapestry had actually been considering Brazil as a project location for several years before Mayor Paes reached out, Crahan said, because of the country’s vast energy generation resources. Brazil’s electric grid is one of the cleanest globally: In 2024, IEA reported that hydropower accounted for more than 56% of the country’s electricity generation mix, followed by wind and solar. Brazil is also a key producer of oil and gas, and became a net oil exporter in 2017.
But exact replication of Rio’s pathway to data center buildout will be challenging in many parts of the world. In the U.S., for example, where recent reports indicate that grid congestion costs are now consistently above $10 billion annually, thanks to transmission capacity constraints, access to data has long been a challenge.
The closest U.S. comparison to the effort underway in Rio, Crahan said, is Tapestry’s work on interconnection queue management with PJM. Transmission congestion costs there are among the highest in the country, at $1.75 billion a year in 2024, up from $529 million a year just five years ago, according to Grid Strategies.
While Tapestry’s work with PJM isn’t data center specific, it’s actually a good example of the level of coordination required for that type of effort, thanks to the fact that there are 13 states and Washington D.C. operating in the market, Crahan explained.
“The Venn diagram of what [the stakeholders] are solving for is not 100% overlapped,” she added. States have different priorities, attributes, and levels of urgency. Tapestry’s broad goal is “making sure that we’re delivering solutions that uplift all decision making for grid and industry and capital infrastructure investment,” Crahan said.


