In the Race to Build the Best AI, There’s Already One Clear Winner

Attendees to the Game Developers Conference 2014 look at screens in the Nvidia booth in San Francisco.

Attendees to the Game Developers Conference 2014 look at screens in the Nvidia booth in San Francisco. Jeff Chiu/AP

Almost every major tech company is partnered up with Nvidia and uses its hardware.

As Google, Facebook, Microsoft and Baidu take turns leapfrogging each other in artificial intelligence innovation, one company stands to profit from any outcome: Nvidia.

Graphics processor units, the company’s biggest moneymaker, have become the industry standard for deep learning, a flavor of artificial intelligence widely used by tech companies to build personal virtual assistants, image recognition for tagging photos, and even the software behind self-driving cars.

Despite talks from Microsoft and Google about developing their own proprietary chips, almost every major tech company is partnered up with Nvidia and uses its hardware. Last month, Microsoft announced a partnership to work with Nvidia’s AI-tailored DGX-1 supercomputer, and Google’s recently revamped cloud services will offer the option to run on Nvidia GPUs in 2017. Facebook’s open-source Big Sur design for its server racks also rely on Nvidia hardware.

Nvidia’s autonomous car hardware, the PX2, powers the autopilot feature in all of Tesla’s cars. The chipmaker has also partnered with Chinese search giant Baidu to work on road mapping and self-driving systems.

The strength of the “data center” line item on Nvidia’s latest financial filing in November, which covers the GPUs sold for enterprise AI, reinforces the company’s strong industry ties. Data center revenue grew 193 percent year over year, from $82 million in Q3 in fiscal year 2016 to $240 million in Q3 fiscal year 2017.

The company still makes a bulk of its revenue ($1.2 billion) from gaming graphics cards, although Miles Brundage, an AI policy research fellow at the University of Oxford, points out independent and academic researchers do use gaming GPUs for their research. So Nvidia’s AI success might be even bigger than its financial records suggest.

Investors have also liked Nvidia’s 2016, reflected in more than 220 percent growth in the company’s stock value. In addition to its AI pursuits, Nvidia also confirmed in 2016 its hardware will power the Nintendo Switch (due out March 2017). For desktop and laptops, the company released a new line of low-cost, high-powered gaming GPUs this past year.

At Nvidia’s European developer conference in September, CEO and co-founder Jen-Hsun Huang led his keynote presentation talking about the company’s work in artificial intelligence.

“The AI revolution, the beginning of the fourth industrial revolution; however you describe it,” Huang said. “We think something really really big is around the corner.”