Nvidia’s AI mania provides new urgency to Biden’s goal of boosting US chipmakers
A White House push on semiconductors confronts AI disruptions where 'nobody knows what the industry will look like in a couple of years.'
Nvidia’s blockbuster results on Wednesday demonstrated the insatiable demand for semiconductors to power new AI systems and the fear that supply may not be able to keep up.
That could make Washington’s goal of revitalizing US chip manufacturing in the years ahead even more pressing.
The Biden administration is currently in the early stages of a series of announcements about how it will spend $50 billion in taxpayer dollars in the coming years to help ensure that America is not left behind.
One big beneficiary is likely to be Intel (INTC), which is building a new factory in Ohio and is reportedly in talks for as much as $10 billion of that government money.
Commerce Secretary Gina Raimondo made a virtual appearance at Intel on Wednesday. She declined to offer new details on what type of government grant the company might receive but underlined her focus on AI.
"For you, frankly, it's a source of great upside and opportunity and customers," she told Intel CEO Pat Gelsinger at the event that also featured Microsoft (MSFT) CEO Satya Nadella and OpenAI CEO Sam Altman.
Other major players in AI chip manufacturing, from Taiwan Semiconductor Manufacturing Company (TSM) to Micron (MU), are also expected to receive billions and are currently in talks with the government.
"These are intensive negotiations; we are in the thick of it now with our major applicants," a senior administration official said this week, promising further announcements would be coming within weeks.
Thus far, three smaller manufacturing awards have been promised to BAE Systems (BAESY), Microchip Technology, and GlobalFoundries (GFS).
The administration also recently announced a likely award of $5 billion for a public-private consortium called the National Semiconductor Technology Center to quarterback the administration's research efforts.
An effort to 'try to spread their bets'
Raimondo oversees a team that has ballooned to more than 200 people charged with implementing the CHIPS and Science Act, legislation that was signed into law 18 months ago and designed to confront America's falling share of semiconductor manufacturing.
Raimondo’s CHIPS Strategy Office is made up of scientists, supply chain and workforce specialists, and others. The team is heavily focused, according to her department, on both immediate-term goals but also planning for the potential impact of AI in the years ahead.
The administration is "going to try to spread their bets a bit widely, which I think makes sense given that nobody knows what the industry will look like in a couple of years," says Christopher Miller, the author of a recent history of the semiconductor industry.
The question is how much the US government can impact things in the crowded landscape. It will have to compete with China, which is also offering more generous government incentives, and perhaps burgeoning private efforts like Altman's goal of raising up to $7 trillion for semiconductor manufacturing, according to a Wall Street Journal report.
The plan has garnered a skeptical media and industry response to whether such a project is possible, as well as the tone Altman has taken. Altman recently posted an apparent reference to the project — and suggested perhaps it could seek an even more astronomical sum — asking, "why not 8" as well as a shorthand for an expletive.
Either way, a wide range of officials agree that demand for AI chips will be high. On Wednesday, Raimondo referenced recent talks with Altman and others in the AI space, noting "the volume of chips that they project they need is mind-boggling, even if you take their projections and cut them in half."
'Manufacturing is definitely the bottleneck'
The US is running behind. It produced nearly 40% of the world's chips in 1990, but less than 10% are made in the US today, according to the White House.
The situation is even worse with the world's most advanced semiconductors: 100% of those are currently manufactured overseas, mostly in Taiwan.
TECHnalysis Research president and chief analyst Bob O'Donnell said AI demand will "organically" lead to a new wave of chip research and design, but restarting the manufacturing side may be aided by the government's help.
"Chip manufacturing is definitely the bottleneck, and we're seeing it now," he says.
That point was echoed by Nvidia (NVDA) CEO Jensen Huang Wednesday when he discussed one of the few constraints on his quickly growing business.
"We expect that demand will continue to be stronger than our supply provides through the year," he told investors on his quarterly earnings call.
Nvidia partners with other companies when it comes to actually manufacturing its chip designs, with TSMC being the main manufacturer of Nvidia's most-advanced chips.
Can AI be used to design more AI chips?
Industry figures and administration officials are also wrestling with the question of where the next generation of AI chips will be designed. This is an arena where the US is comfortably positioned at the moment but one that could be disrupted quickly.
For one thing, AI is already bringing a flood of new participants into chip designing business. Giants from Google (GOOG) to Amazon (AMZN) are already showing more interest as they seek designs to fuel their AI initiatives.
Arati Prabhakar, the director of the White House Office of Science and Technology Policy, recently addressed a roomful of industry representatives gathered at the White House and described the administration's overall approach as "creating the transformative pathways that drive the technology forward but can also reshape the industry."
"We want to make sure that that happens in a way that creates advantage for our industry, but also advantage for America," she added.
O'Donnell added that AI itself could be used to design the next generation of chips, noting that smaller companies like Cadence Design Systems, Inc. (CDNS) and Synopsys Inc. (SNPS) are hoping to utilize generative AI into their chip designs.
"It's going to help with perhaps coming up with more efficient chip designs," he said, adding, "right now, it's a market dominated by Nvidia, but it is not guaranteed to be that way in five years."
Ben Werschkul is Washington correspondent for Yahoo Finance.
Click here for politics news related to business and money
Read the latest financial and business news from Yahoo Finance
