Inside Intel’s big bet to save the US chip industry — and itself

At Intel’s new 700-acre factory in the Arizona desert, the company has begun large-scale production of the most advanced chips it has ever made in the United States.

Intel The company claims it has broken through long-standing technical barriers in the new manufacturing process, after years of effort, to produce faster, more efficient chips that will begin appearing in laptops and data centers next year.

Its two $32 billion state-of-the-art factories in Arizona represent a crucial gamble to show skeptical Big Tech customers that Intel’s latest process can compete with dominant rival TSMC, based in Taiwan – and to prove that advanced chipmaking in the United States is still possible.

“It is the most advanced Semiconductors “The technology used in production today exists on planet Earth,” said Kevin O’Buckley, senior vice president of Intel’s Foundry Business. “But we know we have a long way to go to bring confidence to our customers.”

These big claims will be carefully tested by potential customers like Nvidia, Apple, and Qualcomm before they decide to entrust Intel with making future chips.

“I think the timeline is between 6 and 8 months,” said Ben Bajarin, CEO and principal analyst at consulting firm Creative Strategies.

The 18A manufacturing process demonstrated in Arizona should convince customers to place advanced orders for Intel’s next generation 14A chip manufacturing technology.

If it fails to impress, it could deal the death blow to Intel’s multi-billion-dollar bet on the leading US chip industry, potentially pushing the company into crisis again.

“There comes a time when they have to decide whether they can do it or not,” Bajarin said.

The pressure on Intel is intense. The Trump administration, recently taken 10 percent take into the company, It is determined to reduce America’s dependence on weak technology supply chains abroad.

It’s also difficult to overstate the size of Intel’s financial bet. Completion of the Chandler, Arizona, facility near Phoenix will cost the company more than half of its 2024 revenue.

Intel’s losses at its foundry have run into the billions of dollars every quarter for about two years as its revenues have declined, showing little so far.

“It is difficult to attribute much value to a foundry company losing more than $10 billion annually” with an uncertain path to profitability and burdened by Intel’s roughly $20 billion in net debt, Morgan Stanley analysts wrote.

The University of Arizona campus is a dense network of massive factory and office buildings, solar panels, towers, and dusty vacant lots. The new factory, called Fab 52, uses twice the amount of concrete as the Burj Khalifa in Dubai.

The shell of a second facility, Fab 62, has also been completed. But it remains empty, a testament to the American chipmaker’s turbulent recent history.

Former CEO Pat Gelsinger began construction in Arizona in September 2021, as part of his push to regain leadership in manufacturing.

The goal was to start manufacturing chips for other companies, and capture a larger share of the high-margin global foundry business. This market reached $138 billion in 2024 and continues to grow as demand for AI chips rises, according to research group Gartner.

TSMC produces more than 90 percent of the world’s most advanced chips for the likes of Nvidia, Apple, Google, Qualcomm and Broadcom.

Under Gelsinger, Intel has largely failed to attract these customers. The company also missed the opportunity to develop a product to rival Nvidia’s AI chips, which pushed it past a $4 trillion valuation.

The revenues that Gelsinger promised would support massive investments in manufacturing did not materialize, and they did He was ousted by Intel’s board of directors last year.

Some analysts expected his successor, Lip Bo Tan, to sell the entire struggling manufacturing division.

But that option has been constrained by pressure from the Trump administration — which is concerned that much of Taiwan’s vital chipmaking capacity remains within walking distance of China.

In August, the US government converted billions of dollars in planned manufacturing subsidies into shares, taking a 10 percent stake in Intel.

Investments by Nvidia and SoftBank followed, boosting Intel shares, which rose about 52 percent in the past month.

The US stake means Intel “has gone from being too big to save to too big to fail,” said Dan Hutchison, vice president of consultancy TechInsights.

Trump’s influence could now help the Arizona facility secure customers.

The president has used threats of tariffs and political pressure to persuade major technology groups like Apple to fall in line with his policy of boosting American manufacturing.

“All of these (large U.S. chip customers) are now being intimidated by the Trump administration,” Hutchison said. “I’ll be surprised if someone doesn’t come.”

Developments at Fab 52 are under the microscope of these potential customers. “Anyone who needs a high-end foundry is undoubtedly constantly evaluating Intel,” Bernstein analysts said.

Tan walks the line between reining in capex and showing that Intel has the potential to be a serious competitor to TSMC.

He has promised to curb excessive spending, trim Intel’s workforce, freeze manufacturing projects in Germany and Poland, and slow construction of another 1,000-acre campus outside Columbus, Ohio.

But Intel is still spending to prove that the 18A system at its Fab 52 can reverse a trend that began in late 2010, when TSMC outpaced Intel in introducing superior “contract” manufacturing — processes that involve producing smaller, more complex and efficient chip designs.

TSMC gained technological leadership to the point that Intel began using TSMC to produce its most advanced chips.

Intel will now produce two new Intel products, a personal computer chip called Panther Lake and a server chip called Clearwater Forest.

Hutchison said returning Intel’s own chips to its in-house factories was a “huge help.” He added that the commitment showed that he was “betting that Intel Foundry will still be here in two to three years.”

Jim Johnson, Intel’s senior vice president of customer computing, said initial problems with the process’s “yield” — the percentage of working chips that come out the other side — have been resolved. He expected the official launch of these new chips early next year to “change the narrative” around Intel.

A process known as EUV lithography is central to Intel’s new manufacturing technology. This technology allows patterns to be drawn on silicon wafers that are nanometers thick.

The Fab 52’s massive white EUV machines, made by Dutch company ASML, are about the size of a Winnebago. Each one costs Intel “hundreds of millions of dollars,” said Jason Smith, director of engineering.

The investment Intel has made in offering 18A by making its own chips in Arizona now needs to pay off by attracting customers to the 14A technology, which will go into production by 2028. Intel has warned that it will abandon 14A if it does not win customers.

Fab 52 only has a small window to prove its viability. “If the chips are good, that’s a positive sign to start taking Intel’s foundry seriously,” Bajarin said.