Intel aims to find customers and capture TSMC with new chip factory in Arizona

Intel was once the world’s largest semiconductor company, but its market capitalization has plummeted in recent years as it fell behind Taiwan Semiconductor Manufacturing and spent billions of dollars trying to catch up.

Now, Intel has started mass production of 18A, a new chip node that it says will turn things around.

The biggest problem? Convince major chip makers to trust Intel to manufacture at the new node. For now, Intel’s only major customer is itself. The company’s long-awaited Core Ultra Series 3 PC processors (codenamed Panther Lake) will be the first major products built at 18A to arrive in PCs in January.

“Right now, this is an internal node,” said Futurum Group CEO Daniel Newman. “So many companies are not going to make the switch yet because they have invested heavily in TSMC to ensure yield and ensure wafer capacity.”

Intel has hopes of attracting customers to its new chip manufacturing facility, Fab52, in Chandler, Arizona, and CNBC got an exclusive on-camera tour there in November. About 80 miles north, TSMC also has a new factory in Phoenix, making chips using 4-nanometer technology. Its cutting-edge 2nm technology is currently only manufactured in Taiwan.

Intel’s 18A is roughly on par with TSMC’s 2nm node in some metrics such as transistor density. But as Intel resolves issues with previous nodes after years of delays, some 18A wafers are defective, reducing the number of usable chips per wafer, commonly referred to as yield.

“Yield is always an issue at advanced nodes. It’s not an uncommon issue,” said David Yoffe, a Harvard Business School professor who served on Intel’s board from 1989 to 2018, pointing to early yield issues. NvidiaTSMC’s Blackwell GPUs were quickly fixed.

Intel’s renewed focus on foundries (manufacturing chips for external customers) began when Pat Gelsinger became CEO in 2021. Gelsinger was forced out last December and replaced in March by Lip Vu Tan.

“Over the past few years, the company has invested too much too quickly and without sufficient demand,” Tan said in a July note.

As Intel waits for a big outside customer, the U.S. government went even further in August, acquiring a 10% stake in the company for $8.9 billion, largely funded by subsidies promised under the CHIPS Act signed into law by President Joe Biden in 2022.

A few days ago Softbank invested $2 billion in Intel, followed by a $5 billion investment in September from Nvidia, which agreed to use some of Intel’s technology but did not commit to using Intel’s foundry.

Here’s a peek behind the curtain at Intel’s new chip factory. That’s where we find customers at major foundries and work with them to redeem them.

fall of the giant

Founded in 1968 by Silicon Valley chip pioneers Robert Noyce and Gordon Moore and legendary investor Arthur Locke, Intel brought the world’s first commercial microprocessor to market just three years later.

From the late 1970s to the early 2000s, Intel rapidly introduced advanced process nodes, giving rise to the term “Moore’s Law.” This means that the components on a chip double every few years.

“The 1990s were a time of surprise and excitement for Intel,” Yoffe said. “We were the world’s largest semiconductor company and the world’s most profitable.”

However, Intel was largely late to the mobile revolution, famously turning down a contract to manufacture Apple’s processors for the original iPhone. Then, the AI ​​failed.

In 2024, Intel lost about 60% of its value, making it the worst year in history. This precipitous decline came after the launches of two previous chip nodes, 10nm and 7nm, were delayed by several years. Analysts say ASML’s earlier choice to refrain from using expensive extreme ultraviolet lithography equipment may have triggered the delay.

“I think we’ve lost cycle time discipline,” said Jim Johnson, head of client computing, who joined Intel more than 30 years ago. “To achieve cycle time, you have to commit and execute. We started saying to ourselves, maybe we can achieve more and do more with longer cycle times.”

As it struggles to get back on track, Intel told CNBC there are at least 15 EUV machines in Fab52.

By 2021, TSMC became the node leader and Intel began outsourcing some cutting-edge chip production to the Taiwanese giant. Around the same time, apple began replacing Intel chips in Mac computers with its own M-series chips, also primarily manufactured by TSMC.

During his time at Intel, more than a decade before he rejoined the company as CEO, Gelsinger “was given the responsibility of building GPUs to compete with NVIDIA,” Yoffe said. “Unfortunately, that project failed and ultimately meant we were unable to play a significant role in the AI ​​revolution.”

Intel may be currently considering a deal to acquire custom AI chip design startup SambaNova for $1.6 billion, but the company declined to comment on the matter.

“Changing our culture”

A trademark of Mr. Gelsinger’s tenure as CEO was Intel’s focus on chip manufacturing. With his ambitious roadmap, Intel caught up with TSMC by releasing five nodes in four years.

Currently, Tan is the CEO and Naga Chandrasekaran is the foundry manager.

In an interview with CNBC in November, Chandrasekaran said, “We are improving our yield month-over-month, improving our defect density, and we are achieving our goals.” “So I believe we have turned the corner.”

Chandrasekaran joined Intel last year after more than 20 years at a major memory manufacturer. micron. He said his biggest goal is to find customers for the foundry.

“I have to be part of their team and convince them that they can trust Intel Foundry to execute,” Chandrasekaran said. “That’s number one. And to do that, we’re changing the culture. We’re putting a big focus on execution within the Intel Foundry.”

Chandrasekaran told CNBC that Fab52 is capable of starting more than 10,000 18A wafers per week. Arizona has more than 1 million square feet of manufacturing cleanroom space and all five fabs are connected by 30 miles of overhead track to move wafers between fabs. The sixth fab, Fab62, is expected to be completed around 2028.

The 18A also uses RibbonFet, Intel’s gate-all-round architecture that improves power control by completely surrounding the transistor, unlike previous designs that only touch the top and sides. Chandrasakaran said 18A offers “over 15% more performance per watt” than Intel 3.

Perhaps the biggest way Intel stands out is through its advanced packaging, the assembly and connection of chips into the final systems used in real-world applications.

CNBC visited Intel’s advanced packaging lab in Chandler to see several steps in the process, including protecting the chips with polymer-based seals and exposing them to liquids that detect defects. Yoffe said Intel’s advanced packaging “helps alleviate some of the power consumption issues.”

“For everyone who makes chips for data centers, one of the biggest issues today is the power the chips consume,” Yoffe said.

Chandrasekaran said the Arizona plant is powered by nearly 100% renewable energy. When it comes to water, Intel’s Arizona facility used more than 3 billion gallons in 2024 and returned 2.4 billion gallons to the local supply through an on-site water recycling plant.

“There are no blank checks.”

When it comes to spending on future foundry nodes, Tan’s message to employees is clear. “No more blank checks.” A company needs customers.

Construction of Intel’s massive new chip factory in Ohio has been delayed until at least 2030, and Tan made significant cost cuts in July by cutting 15% of the workforce and canceling projects in Germany and Poland.

“That’s what the company needed,” said Futurum’s Newman. “I needed to be faster. I needed to be leaner. I needed to be more focused. I needed someone a little more astute.”

Tan is waiting to see how demand evolves before revealing firm details about Intel’s next node, 14A. Chandrasekaran told CNBC that the device will initially be developed in Oregon, with a goal of mass production in 2028.

Finding customers for 18A is not easy. Unlike TSMC, which only makes chips for external customers, Intel also makes devices with its chips, positioning it as a competitor to some of the customers it hopes to win.

“If I were Nvidia, or AMD or Qualcomm or broadcomdo you really want to put a secret sauce into a manufacturing operation that gives Intel access to that secret sauce?” Yoffie said.

He suggests splitting the foundry into a separate company.

“If we can actually separate the two, I think it gives Intel a much better chance of success,” Yoffe said. “It will also further strengthen the United States’ position as the home of major semiconductor manufacturing organizations.”

For now, Intel is hoping that Panther Lake will be a big proof point when it comes to PCs from major companies such as: samsungDell, HP, lenovoASUS and Acer in January. Intel’s next data center chip, the Xeon 6+, is also manufactured at 18A.

“If you’re a major company that wants to bet on the process node, you’ll be much more comfortable seeing Intel increasing the core of its client product line to high capacity at that process node,” Johnson said.

Microsoft and Amazon signed an initial deal last year committing to use Intel’s foundries for some of their custom chips.

“That’s a good sign, but of course the volume is very small compared to Nvidia and other big chip companies,” Yoffe said.

Recent reports have suggested that AMD is considering manufacturing with Intel, and one analyst predicts that Apple could return to making some Mac chips with Intel by 2027.

Meanwhile, Intel received a lifeline: a 10% stake in the U.S. government.

“This shows the confidence the U.S. government has in Intel and the belief that cutting-edge research, development and manufacturing must be done in the United States,” Chandrasekaran said.

The government’s investment comes days after President Donald Trump called for Tan’s resignation and later reversed course.

“I sometimes worry about the scope here and how the United States could decide to fund all sorts of things,” Newman said. “But there are industries that we’ve pulled out of the United States that put us at risk we can’t protect against, and we need to get them back.”

The share of U.S.-made chips has been declining for decades, and about 92% of the world’s most advanced chips now come from Taiwan.

“The stakes are incredibly high for Intel, for the United States, and for the world,” Yoffe said. “The very idea that the world’s most advanced products depend on a single site on an island miles off the coast of China is a terrible situation for the entire world to deal with.”

Chandrasekaran is working to transform Intel into a maker of advanced chips.

“As a semiconductor community, we must enable this solution for the world to move forward with AI,” he said. “We have no choice but to succeed.”

WATCH: What’s inside Arizona’s chip fab could be key to Intel’s recovery