Armored Cars and Trillion Dollar Price Tags: How Some Tech Leaders Want to Solve the Chip Shortage

Some corporate technology leaders expect the shortage of advanced artificial-intelligence chips to improve next year, but until then, are not taking chances securing the hardware that will allow them to leverage new generative AI applications.

Graphics processing units, or GPUs, which are primarily produced by chip giant Nvidia, have been the engine behind the AI boom because they enable parallel processing, or the ability to run lots of computations at the same time.

Access to those chips was top of mind for chief information officers and tech leaders who convened this week at The Wall Street Journal’s annual CIO Network Summit in Menlo Park, Calif., on Tuesday.

“Those GPUs arrive by armored car,” said Cisco CIO Fletcher Previn, speaking at the summit. Cisco and Nvidia last week announced a joint effort to provide AI infrastructure products, including networking, software and servers to businesses.

The availability of GPUs and other so-called AI accelerator chips could be one of the key limiting factors in exploiting generative AI’s true potential. Manufacturing chips is enormously capital intensive. It is also one of the most intricately complex industries in the world, with a history of sharp cyclical swings that have made companies wary of radical expansion.

The Wall Street Journal earlier reported that OpenAI Chief Executive Sam Altman was pursuing investors including the United Arab Emirates for a project requiring up to $7 trillion to reshape the global semiconductor industry.

“He’s a smart guy,” said Andrew Ng, managing general partner of AI Fund, referring to Altman’s plans at the summit. “I don’t know where we find $7 trillion…That’s an interesting figure to try to raise.”

But Ng also said he is optimistic that chip availability could improve over the next year. Other offerings in the market beyond chips made by Nvidia, including those provided by Intel and Advanced Micro Devices, are rapidly improving.

So far Nvidia’s moat has been its programming language that works in tandem with its hardware, but that is an area where AMD’s open-source offering is increasingly working to compete, he said.

Hasmukh Ranjan, CIO of AMD, said at the summit that the chip giant is using AI to improve its own computing capabilities, and has seen a 5% improvement in efficiency. AMD in December began rolling out its newest AI chips—presenting perhaps the toughest challenge yet to Nvidia’s dominance in the space.

“It’s certainly a constant,” said Eli Collins, vice president of product management at Google DeepMind, remarking on the chip shortage. But it doesn’t keep him up at night, in part because Google has been working for years to develop its own chips.

“If you look at what we can train given the number of chips we have, that’s an area where we’ve been doing a ton of work in terms of efficiency,” Collins said at the summit.

“We were in such a regime where humans were expensive, and machines were cheap, in the PC era. And so it’s interesting to go back to this world now where machines are quite expensive,” Collins said. “The sheer amount of compute that goes into these is pretty wild.”

Write to Isabelle Bousquette at and Belle Lin at

Source link