Nvidia just unveiled the future of the technology that made it the world's most valuable company

Nvidia has provided a closer look at its new AI data center computing platform, Vera Rubin, a release that could have major consequences for the future of AI as the industry relies heavily on the company's technology, CNN writes.

Nvidia had previously announced few details about Vera Rubin, but on Monday at the CES conference in Las Vegas, home to the world's largest international gadget and technology show, it explained in detail how the system works and the release schedule.

The platform is already in production, and the first Vera Rubin-based products will hit the market in the second half of 2026, according to the company.

Nvidia has become the epitome of the artificial intelligence boom, the ubiquity of its chips and platforms propelling the company briefly last year to the status of the world's first $5 trillion company.

At the same time, however, Nvidia faces growing fears of a possible AI bubble, fueled by increased competition and efforts by big tech companies to produce their own chips to reduce reliance on Nvidia.

Where does the money come from?

CEO Jensen Huang, dressed in his signature leather jacket, tackled head-on one of the key questions in the AI ​​bubble debate: where the money for this expansion is coming from.

In his keynote address from the theater at the Fontainebleau Las Vegas complex, Huang said that firms are redirecting budgets from classical computing research to artificial intelligence.

“People ask where the money comes from. That's where the money comes from,” he said.

By Vera Rubin Nvidia is trying to position itself as the answer to the growing challenges posed by advanced AI models, including the question of whether current infrastructure can handle increasingly complex demands.

The company claims in a statement that its upcoming AI server rack, dubbed the Vera Rubin NVL72, “offers more bandwidth than the entire Internet.”

Nvidia says it has developed, with Vera Rubin, a new type of storage system designed to help AI models process complex, context-rich requests faster and more efficiently.

Traditional storage and memory systems, including those currently used in data centers, will no longer be sufficient as companies like Google, OpenAI or Anthropic move from simple chatbots to complex AI assistants.

From chatbots to AI assistants

Huang illustrated this transition from chatbots to AI “agents” with a video demonstration: A user built a personal assistant by connecting a small office robot to several AI models running on the Nvidia DGX Spark computer.

The robot could, among other things, play back the user's to-do list or even tell the dog to get off the couch.

According to Huang, such an assistant would have been unimaginable a few years ago, but is now “commonplace” thanks to large-scale language models that are replacing classical programming in application development.

In other words, the old methods are no longer sufficient as AI becomes more sophisticated and “reasons” through tasks in multiple steps.

“The bottleneck is moving from computing power to managing context,” explained Dion Harris, Nvidia's senior director of high-performance computing and hyperscale AI solutions. “Storage can no longer be a secondary detail.”

Which large companies will be involved in the implementation of Vera Rubin

Before CES, Nvidia also announced a licensing deal with AI inference company Groq – a new signal that it is investing heavily in this area.

“Inference is no longer a one-shot answer, but a thought process,” Huang said, referring to how AI models “reason” to provide answers and perform tasks.

All major cloud providers, including Microsoft, Amazon Web Services, Google Cloud and CoreWeave, are among the first to implement Vera Rubin, according to Nvidia.

Companies like Dell and Cisco are expected to integrate the new chips into their data centers, and AI labs like OpenAI, Anthropic, Meta and xAI will likely adopt the technology to train models and provide more sophisticated answers.

Pressure on Nvidia

Nvidia also expanded its offensive in autonomous vehicles, with new models called Alpamayo, and in the area of ​​”physical AI,” the artificial intelligence that powers robots and other real-world systems, continuing the direction it laid out at the GTC conference in October.

But Nvidia's success and dominance come with huge pressures: The company must consistently beat Wall Street's lofty expectations and calm fears that investment in AI infrastructure is growing far faster than actual demand.

Meta, Microsoft and Amazon have already spent tens of billions of dollars in capital investment this year alone, and McKinsey estimates that by 2030 companies will invest nearly $7 trillion in global data center infrastructure.

“No one wants to depend on Nvidia”

Much of this money flows between a relatively small number of players in a system of “circular finance”, where technology and capital rotate between the same companies.

In parallel, Google and OpenAI are developing their own chips to better adapt the hardware to the needs of their models. Nvidia also faces increasing competition from AMD, and Qualcomm recently announced its entry into the data center market.

“No one wants to depend on Nvidia,” said Ben Barringer, global director of technology research at Quilter Cheviot, in a previous interview with CNN when asked about other companies like Google that could compete with Nvidia in the field of artificial intelligence chips. “Everyone is trying to diversify their chip portfolio.”