At CES, Nvidia's small $3,000 computer captures attention.

• Nvidia revealed a $3,000 mini computer for AI developers this week.
• The project, currently named Project Digits, utilizes an Nvidia chip known as GB10 as its foundation.
• In the future, the PC market, worth $50 billion annually, could be a potential target for Nvidia's chip.

This week at CES in Las Vegas, CEO Jensen Huang was treated like a celebrity due to the recent surge in demand for artificial intelligence, which has propelled the chipmaker to become the world's second most valuable company.

Huang's keynote speech at the annual conference on Monday, which lasted nearly two hours, filled a 12,000-seat arena and drew comparisons to Steve Jobs' product reveals at events.

Huang ended his presentation with a shocking twist: a concealed product unveiling. He displayed one of Nvidia's server racks and, with a theatrical flourish, revealed a miniature version, resembling a compact cube computer.

"Huang stated, "This is an AI supercomputer, clad in an alligator skin leather jacket, which runs the entire Nvidia AI stack and hosts all of Nvidia's software.""

Huang stated that the computer, known as Project Digits, utilizes a GPU that is a relative of the Grace Blackwell GPUs currently powering the most advanced AI server clusters. The GPU is paired with an Nvidia-based Grace central processing unit (CPU). Nvidia collaborated with MediaTek to develop the system-on-a-chip called GB10.

CES, previously known as the Consumer Electronics Show, is typically the platform for launching flashy and futuristic consumer gadgets. This year's show, which began on Tuesday and ends on Friday, saw several companies integrate AI into appliances, laptops, and even grills. Additionally, major announcements were made, such as a rollable laptop screen from Lenovo and a new robot, including a Roomba competitor with a robotic arm.

Unlike Nvidia's traditional GPUs for gaming, Project Digits is not intended for consumers. Instead, it is aimed at machine learning researchers, smaller companies, and universities that want to develop advanced AI but lack the billions of dollars to construct massive data centers or purchase sufficient cloud credits.

"Huang stated that there is a significant demand for data scientists and ML researchers who are actively working and building something. He suggested that it may not be necessary to have a large cluster for early model development and constant iteration. While cloud computing could be an option, it comes with a higher cost."

The supercomputer, which will be available in May, will cost approximately $3,000, according to Nvidia. It will be sold by the company as well as some of its manufacturing partners. Nvidia CEO Jensen Huang stated that Project Digits is a placeholder name and may be changed before the computer is released for sale.

Huang said, "Reach out to us if you have a good name for it."

Diversifying its business

Nvidia's data center GPUs are in high demand among AI model creators and cloud providers due to their ability to handle the most demanding models and workloads.

In the most recent quarter, Nvidia's $35 billion in revenue was primarily driven by data center sales, which accounted for 88% of the total revenue.

Wall Street is concerned about Nvidia's dependence on a few large customers purchasing large AI systems, and is evaluating the company's potential to expand its business beyond this narrow focus.

Nvidia's recent announcement of Project Digits and other products at a low price point has left Melius Research analyst Ben Reitzes feeling a bit uneasy, as he believes Nvidia may have "stolen the show" with its impressive offerings.

The Institute for Data Science at New Jersey Institute of Technology's Project Digits, which utilizes Linux and Nvidia software, offers a significant boost in capabilities for researchers and universities, according to David Bader, its director.

Bader, who has collaborated with Nvidia on research projects in the past, stated that the computer appears capable of handling enough data and information to train the most advanced models. He disclosed to CNBC, Anthropic, and others that they would be willing to pay $100 million to construct a supercomputer for training purposes to obtain a system with these capabilities.

Academics will be thrilled to know that for $3,000, they can acquire a device that can be plugged into a standard electrical outlet at home or in their office, said Bader. This product is particularly appealing to academics who have often had to leave for private industry to access more powerful computers.

Bader stated that any student who can afford a system that costs roughly the same as a high-end or gaming laptop will be able to conduct the same research and construct the same models.

Nvidia may enter the $50 billion market for PC and laptop chips with its first move, as stated by Reitzes.

"Reitzes wrote, "Imagining it being easy to do it all themselves and let the system run Windows someday is not too difficult. However, they don't want to step on too many toes.""

On Tuesday, when questioned by Wall Street analysts, Huang did not dismiss the possibility.

He suggested that MediaTek might be able to sell the GB10 chip to other computer manufacturers in the market. He left some ambiguity in his statement.

"Obviously, we have plans," Huang said.

WATCH: Nvidia pullback due to CES expectations