Computing power stops being a tool and becomes basic infrastructure
En un momento en que las economías buscan nuevos cimientos, la computación de alto rendimiento ha dejado de ser un privilegio para convertirse en infraestructura esencial. Marcio Aguiar, responsable de la división empresarial de NVIDIA en América Latina, describe una transición que ya no es especulativa sino estructural: el mundo está reorganizando su arquitectura económica en torno a la capacidad de procesar información a escala. NVIDIA, con una valoración de cinco billones de dólares y un crecimiento del 92% en ingresos de centros de datos, no es solo una empresa tecnológica exitosa, sino el proveedor silencioso que sostiene la nueva electricidad del siglo XXI.
- La inteligencia artificial ha cruzado un umbral decisivo: ya no se debate si importa, sino a qué velocidad el mundo físico puede construir la infraestructura que demanda.
- NVIDIA alcanzó una valoración de cinco billones de dólares en 2025 y proyecta triplicar su mercado para 2030, señales de que los mercados financieros ven esto como una reconfiguración duradera, no una burbuja.
- El monopolio de los gigantes tecnológicos sobre la computación se está fracturando: los proveedores especializados en 'nubes de IA' ya capturan el 40% de los ingresos de NVIDIA, abriendo el acceso a empresas que antes quedaban excluidas.
- Anthropic avanza hacia su salida a bolsa y OpenAI se prepara para seguirla, señales de que Wall Street apuesta por una expansión sostenida de la infraestructura de datos, no por un ciclo pasajero.
La computación dejó de ser un lujo para convertirse en una necesidad. Esa es la tesis central que Marcio Aguiar, al frente de la división empresarial de NVIDIA en América Latina, desarrolló en una conversación reciente sobre el futuro de la inteligencia artificial. El interés financiero en la IA ya superó la fase especulativa: la economía se está reorganizando en torno a la capacidad de procesar información a escala, y quienes controlan esa infraestructura están ganando.
Los números de NVIDIA lo confirman. Los ingresos de centros de datos crecieron un 92% en el último trimestre, y la compañía cruzó los cinco billones de dólares de valoración en 2025, convirtiéndose en la primera en alcanzar ese umbral. Para el tercer trimestre de este año, NVIDIA comenzará a comercializar Vera Rubin, su nueva generación de procesadores combinados CPU-GPU diseñados específicamente para las exigencias computacionales de la IA. El mercado ya descuenta una expansión triple del negocio para 2030.
El lenguaje del sector también ha cambiado. Donde antes se hablaba de 'fábricas de software', ahora se habla de 'fábricas de IA': espacios donde la potencia computacional se trata como infraestructura básica, comparable a la electricidad o las telecomunicaciones. Aguiar también aclaró un malentendido frecuente: NVIDIA no posee ni opera centros de datos en ningún lugar del mundo, incluido México. La compañía desarrolla la tecnología que los hace funcionar, y está presente en todos los grandes centros de datos del planeta sin ser propietaria de ninguno.
El perfil de clientes también se está diversificando. Hace apenas dos años, los grandes proveedores de nube como Microsoft, Amazon y Google concentraban el 80% de los ingresos de NVIDIA. Hoy, los proveedores especializados en 'nubes de IA' capturan cerca del 40%, lo que significa que empresas industriales y corporaciones medianas tienen acceso a recursos computacionales que antes estaban reservados para un puñado de gigantes. Mientras tanto, Anthropic avanza hacia una oferta pública y OpenAI se prepara para hacer lo mismo, señales de que Wall Street confía en que esta expansión de infraestructura no es un pico temporal, sino un reordenamiento genuino de la economía global.
Computing power has stopped being a luxury and started being a necessity. That's the argument Marcio Aguiar, who runs NVIDIA's enterprise division across Latin America, made in a recent conversation about where artificial intelligence is headed. The financial world's obsession with AI has moved past the stage of speculation—the kind where investors bet on what might happen. Now it's structural. The economy is actually reorganizing itself around the ability to process information at scale, and the companies that can build and control that infrastructure are the ones winning.
NVIDIA's own numbers tell the story. In the most recent quarter, data center revenue jumped 92 percent. The company crossed the five trillion dollar valuation mark in 2025, becoming the first to reach that threshold. Aguiar pointed out that by the third quarter of this year, NVIDIA will begin selling a new generation of processors called Vera Rubin—the first CPU-GPU combination designed specifically to handle the computational demands of artificial intelligence. The market is already pricing in a threefold expansion of NVIDIA's business by 2030. That's not gradual growth. That's a reshaping.
What's happening now is a shift in how people think about computing itself. For years, the industry talked about "software factories" and then "smart factories." Now executives are talking about "AI factories"—places where computational power stops being treated as a tool and becomes treated as basic infrastructure, the way we treat electricity or telecommunications. More than a dozen companies have been selling large language models since late 2022, and every corporation with serious ambitions has had to rebuild its entire IT architecture to handle this new reality. The physical plants matter now. The data centers matter. The chips matter.
One misconception Aguiar felt compelled to correct involved claims that NVIDIA was building its own data centers in the Mexican state of Nuevo León. That's not accurate, he said. NVIDIA doesn't own or operate data centers anywhere in the world. What the company does is develop the technology that goes inside them—the processors and systems that make them work. NVIDIA is present in every major data center on the planet, but it owns none of them. The distinction matters because it clarifies what NVIDIA actually is: not an infrastructure operator, but the essential supplier to everyone who is.
The revenue picture has also shifted in revealing ways. Three years ago, NVIDIA's business was heavily concentrated among the hyperscalers—the massive cloud companies like Microsoft, Amazon, Google, and Oracle. Those giants still matter enormously. But the market is diversifying. A new category of competitors has emerged, companies specializing in what the industry calls "AI Clouds" or "Neo Clouds." These are firms built specifically to serve the artificial intelligence market rather than general cloud computing. In NVIDIA's most recent quarter, the split had changed dramatically. Hyperscalers now account for slightly more than half of revenue. The specialized AI cloud providers capture roughly 40 percent. That's a shift from an 80-20 split just a couple of years ago. What it means is that access to computing power is spreading beyond the handful of giants who used to monopolize it. Larger corporations and industrial customers who couldn't compete with Amazon or Microsoft for resources now have other options.
The momentum is visible elsewhere too. Anthropic, one of the leading AI companies, took steps this week toward a public offering. OpenAI, its main competitor, is expected to file for one soon. Wall Street is betting that the expansion of data centers will continue, that the infrastructure buildout will sustain itself, that this isn't a temporary spike but a genuine reordering of how the economy works. The question now isn't whether AI matters. It's how fast the physical world can catch up to the demand.
Notable Quotes
We've seen exponential growth. By Q3 we'll be selling the new Vera Rubin generation, the first CPU-GPU designed for AI computational demand. The market is estimating three times more growth than we have today by 2030.— Marcio Aguiar, NVIDIA Enterprise Division director for Latin America
We're not operating any data centers anywhere in the world. We develop the technology that goes inside them. We're present in every major data center on the planet, but we own none of them.— Marcio Aguiar
The Hearth Conversation Another angle on the story
When you say computing power is becoming like electricity, what does that actually mean for someone running a business?
It means they can't operate without it anymore. Just like you can't run a factory without power lines, you can't run a modern company without access to AI computing. It's not optional. It's foundational.
But NVIDIA doesn't own the data centers. So what exactly are they selling?
The brains of the operation. The chips and processors that make the data centers work. NVIDIA is like the company that makes the transformers and generators for the electrical grid—essential, everywhere, but invisible to most people.
The shift from hyperscalers to these AI Cloud companies—does that mean the big tech giants are losing power?
Not losing it, but sharing it. Microsoft and Google still get more than half the revenue. But now there are specialized competitors who can serve customers those giants don't want or can't reach. It's democratization, but it's happening because the demand is so enormous that there's room for everyone.
Why are Anthropic and OpenAI going public now?
Because Wall Street believes the infrastructure expansion will keep going. If investors thought this was a bubble, they wouldn't be filing. They're betting that data centers will keep multiplying, that computing power will keep being essential, and that the companies building the AI models will be worth real money.
What happens if the growth doesn't materialize?
Then you have massive overcapacity. But that's not what the numbers suggest. The market is tripling by 2030 according to current estimates. That's not a prediction—that's what's already priced in.