Intel’s Troubles Complicate U.S. Chip Independence

On 1 August 2024, Intel announced financial results for the second quarter of 2024. They weren’t pretty; the company’s stock dropped more than 25 percent as it announced an aggressive plan to cut costs, including layoffs that will impact 15 percent of its entire workforce.

As the drastic hit to Intel’s stock makes plain, the cost reduction plan caught many off-guard. The company has suffered no shortage of bad news in recent years, but expected investments and incentives from the CHIPS and Science Act, which aims to boost the country’s domestic chip manufacturing, provided an edge of hope to cling to. Intel’s continued struggle begs the question: will the U.S. government need to do more?

“I don’t think we can lose Intel. That would be a bridge too far,” says Rob Atkinson, president at the Information Technology and Innovation Foundation, a technology think tank. “So, then the question becomes, what if Intel says ‘we need a cash infusion?’ I think the U.S. government would have to take that seriously.”

$8.5 billion is a lot. Is it enough?

The US CHIPS and Science Act is a law passed on 9 August 2022, with the goal of invigorating domestic production of semiconductors. The law did not name the recipients and instead approved funds for later allocation. That began to happen earlier this year and on 20 March 2024, the U.S. Department of Commerce and Intel reached a “preliminary memorandum of terms” that includes US $8.5 billion in direct funding and $11 billion in loans. Intel also plans to claim a 25 percent tax credit on investments made into semiconductor manufacturing facilities that become operational between 2023 and 2026.

That sounds like a lot of funding. Yet Intel’s stock has been cut in half since the announcement. One reason? Building new leading-edge semiconductor fabs is extremely expensive.

“Intel is the only U.S. company that’s in any position to develop technology to compete.” —Mike Demler, semiconductor analyst

“I don’t know that Intel is going to be making a lot of money,” says Atkinson. “What this is, really, is a subsidy to put a fab in a high cost country.” Building any semiconductor fab is expensive, he says, and building in the United States is even more so. Because of that, he believes funding from the CHIPS Act is unlikely to boost Intel’s bottom line.

He’s not alone in his thinking. Several reports have reiterated the high cost of leading-edge fabs. A December 2023 report from International Business Strategies estimated that TSMC’s global investment into its upcoming N2 (2-nanometer) process, slated for volume production in 2025, could approach $28 billion. A separate report on a pair of Samsung fabs to be built in Texas estimated costs of up to $44 billion.

As these investments show, companies held outside the United States are happy to leap on incentives to build fabs in the country. Yet Intel remains distinct from these peers in one key way: it’s the only U.S.-based company with leading-edge semiconductor fabs. Samsung and TSMC were awarded less funding under the CHIPS Act at $6.4 billion and $6.6 billion, respectively, and granted smaller loans.

Without Intel “the alternative would be that the fabs TSMC or Samsung are building would be under more control of the [United States],” says Mike Demler, a semiconductor analyst. Such a step might take the form of having foreign companies divest majority ownership of domestic fabs, he says. “That’s not going to happen.”

Atkinson says that while the CHIPS Act is a start, the United States is likely to need additional funding, support, or incentives if the goal is to see competitive leading-edge semiconductor fabs owned by U.S. corporations. “We’ve hit a double in the first inning, and that’s pretty good,” he says. “But the problem is everybody in Washington thinks we’re done.”

His think tank, ITIF, recently released a report on Chinese semiconductor innovation that makes several recommendations for American lawmakers, including an extension on the 25 percent tax credit for investments in semiconductor production through at least 2030. (They currently expire on 1 January 2027.)

Intel’s Foundry Future

Rough 2nd quarter financial results and pending layoffs paint a dismal picture of Intel’s future. Some have even speculated that a competitor, such as Broadcom, might try an acquisition in the coming years. But despite the recent bad news, Demler expressed optimism about the technology behind Intel’s foundry business.

“Intel is the only U.S. company that’s in any position to develop technology to compete. And in fact… in process technology, Intel has been an innovator,” says Demler. He pointed to process technology advancements such as the FinFET, a fin-shaped transistor put into production first by Intel in 2011, and more recently Foveros, an advanced chip packaging technology that allows vertical stacking of chips.

Much of Intel’s foundry future is bet on Intel 18A, the company’s next leading-edge semiconductor production process. This “1.8-nanometer” production process will combine multiple Intel innovations including 3D hybrid bonding, nanosheet transistors, and back-side power delivery. Demler says that, if all goes to plan, Intel’s 18A should compete directly with, or even exceed, TSMC’s upcoming 2N process technology.

But successful production technology, although a positive step, might not solve all Intel’s problems. In addition to the high costs of investing in new fabs, Intel must navigate the tricky prospect of attracting customers to its foundry business while it continues to design CPUs and other chips. “I don’t know if the foundry model works for Intel, because they’re competing with their customers,” says Atkinson.

Intel CEO Pat Gelsinger, anticipating the problem, announced a reorganization in February of 2024 to split the company into Intel Foundry Services and Intel Product. It’s too early to know if that split will calm customer’s fears.

At the moment, the fate of Intel’s foundry—and, by extension, U.S. domestic chip manufacturing—remains to be seen, with all eyes fixed on Intel’s 18A, which is expected to enter production in 2025. If it’s successful, 18A will put Intel back on the leading edge. If it falters, the effort to bolster U.S. chip manufacturing will be in serious trouble.

“I think next year,” says Demler, “the proof will be in the pudding.”