Intel Bets Its Comeback on $20 Billion Arizona
Chip Plant
The centerpiece
of the company’s comeback hopes is a multibillion-dollar facility in the Arizona
desert. Will it draw new customers?
Intel is pinning
its turnaround hopes on a new multibillion-dollar semiconductor factory in Chandler,
Arizona, as it races to reclaim lost ground in chip manufacturing. The $20 billion
“Fab 52” plant, outfitted with advanced lithography tools from ASML, marks the company’s
first use of the technology in U.S. production in nearly a decade.
Under new
CEO Lip-Bu Tan, Intel is rolling out a next-generation chip process called 18a
and a new chip architecture, Panther Lake, both seen as vital to restoring
its technological edge. The 18a process promises denser, more power-efficient chips
by stacking transistors and rerouting power from the chip’s backside.
The stakes
are high: Intel lags behind Taiwan’s TSMC, which leads in next-generation chip yields.
The company aims to prove it can make its own chips efficiently and eventually attract
major clients like Apple or Nvidia to outsource production to Intel’s foundries.
The U.S. government
recently injected $8.9 billion into Intel, taking a 10% stake under the CHIPS
and Science Act, which supports domestic semiconductor production. Analysts
say Fab 52’s success—or failure—could determine whether Intel can once again “fire
on all cylinders” or remain in TSMC’s shadow.
On
the desert floor of the Phoenix Valley, Intel has poured more than $20 billion into
a four-story manufacturing plant that is the centerpiece
of the ailing chipmaker’s comeback bid.
Inside
the building, known as Fab 52, the company is rolling out a new manufacturing process
to make more powerful and efficient computer chips. The process uses the tools from
ASML, the Dutch manufacturer of lithography machines, to make Intel’s cutting-edge
semiconductors in the United States for the first time in nearly a decade.
During
a factory visit last month, two of those $250 million machines sat largely idle
while ASML engineers in sterilized white bunny suits worked on one. Two trailer-size
footprints for additional machines sat empty nearby, a nod to Intel’s hopes it can
eventually expand.
Intel
has recently brought a string of potential customers through the plant in Chandler,
Ariz., and pitched them on manufacturing their chips at Fab 52. But analysts say
most chip companies want to see whether Intel can succeed in making its own computer
chips before asking it to produce their chips for smartphones, artificial intelligence
systems and other technologies.
The
focus on the new facility underscores how critical this moment is to Intel’s future.
Once an exemplar of Silicon Valley success, the company has fallen behind competitors
as a chip manufacturer and designer. Intel was leapfrogged by Taiwan Semiconductor
Manufacturing Company, or TSMC, after it failed to implement ASML technology. And
it was dropped from Apple’s laptops after its chips struggled with battery life
and performance.
As
sales declined, Intel churned through leaders. In March, it named Lip-Bu Tan its
third chief executive in five years. From his predecessor, he inherited a troubled
balance sheet and an audacious strategy to introduce five new production processes
in four years while remaking Intel as a manufacturer for other chip designers.
Mr.
Tan shored up the company’s finances in August when the Trump administration pumped
$8.9 billion into the company for 10 percent of its business. The deal made Intel
the recipient of one of the largest U.S. government investments in a company since
the 2008 financial crisis.
The
Trump administration made the investment by releasing money promised to Intel under
the CHIPS and Science Act, a federal program signed into law in 2022 to provide
companies with grants to revive U.S. semiconductor manufacturing.
In
2021, as Washington lawmakers were weighing the CHIPS Act, Intel committed to investing
$32 billion in two new factories, Fab 52 and Fab 62, on its Ocotillo campus in Arizona.
It later received federal funding to support the project.
Now,
Mr. Tan is focused on executing on the second part of Intel’s strategy: its technical
turnaround.
At
an event last month for a few hundred technology analysts and members of the news
media, Sachin Katti, Intel’s chief technology officer,
acknowledged the stakes. He said the new manufacturing process, which Intel calls
18a, and the new computer chip, which it calls Panther Lake, are “foundational to
our future.”
“We
are of course making two really big bets here,” he said. He added that the manufacturing
efforts were “mission critical, not just for Intel but for our country.”
Intel
became the world’s dominant chipmaker by staggering the way it adopted new technology,
said Ben Bajarin, the principal analyst at Creative Strategies,
a tech research firm. The company typically developed a new manufacturing process
to etch silicon with billions of small transistors that are slimmer than a human
hair. It then perfected that process on an older generation of chips before using
it to produce chips designed with new features.
But
this time, Intel is trying to make manufacturing and design leaps simultaneously.
The 18a process uses the ASML tools to produce chips that stack transistors, so
they take up less space. It also routes power through the backside of a chip rather
than the top, where data and power collided in the past. The result is a denser,
more energy-efficient chip, the company said, capable of piling 10,000 sheets of
silicon on top of one another in a stack thinner than a sheet of paper. Intel said
its Panther Lake series of chips would allow laptops to run A.I. systems and do
all-day computing. It will start shipping the chips early next year.
The
company has enough faith in the process that it is moving production of its newest
chips to its own factories rather than relying on TSMC, which it turned to after
its previous processes faltered. But the changes create more opportunities for something
to go wrong, Mr. Bajarin said, which is especially risky
in a process where a particle of dust can destroy a chip and lead to financial losses.
“The product has to be good, and the process has
to be good,” Mr. Bajarin said. If all those things happen,
he said, the industry will “know that Intel can again fire on all cylinders.”
Intel’s
success is far from guaranteed. Late last year, it informed some customers that
its 18a process was lagging behind rivals. At the time, TSMC was producing 30 percent
of its leading-edge chips, known as 2 nanometer chips,
without any flaws, while Intel’s process was producing fewer than 10 percent of
its 18a chips without flaws.
At
the two-day event last month, Intel executives declined to comment on the percentage
of chips it was making without any flaws at Fab 52. The number of chips that don’t
work won’t become clear until the company makes financial reports next year, analysts
say.
By
bringing some production back in-house from TSMC, the company could save money and
improve its profits, said Patrick Moorhead, founder of Moor Insights & Strategy,
a tech research firm. It’s also a proving ground to show big producers of chips
like Nvidia or Apple, which exclusively rely on TSMC, that Intel could produce their
semiconductors.
Intel
doesn’t make enough of its own chips to fill the Fab 52 plant, Mr. Moorhead said.
But if it can persuade others in the tech industry that its process works, it could
buy more ASML machines and put more of its factory space to work.
“It’s
like a shopping center that needs an anchor tenant,” Mr.
Moorhead said. “If it can find one, it would be game on.”