Intel is making a bold play to reclaim its former glory with the launch of its 18A chip manufacturing process. Once the undisputed leader in semiconductor technology, Intel has seen its market share erode, largely due to its struggles to keep pace with rivals like Taiwan Semiconductor Manufacturing Company (TSMC). The company’s latest endeavor, the 18A node, is positioned as its technological comeback, promising to deliver cutting-edge performance and efficiency.
However, the path to recovery is fraught with challenges. A key hurdle is convincing major chip manufacturers to entrust Intel with their most critical production. Currently, Intel’s primary customer for the 18A process is itself, with its upcoming Core Ultra series 3 PC processor, codenamed Panther Lake, set to be the first major product manufactured on this advanced node.
“Right now, it’s an internal node,” explained Daniel Newman, CEO of Futurum Group. “Many companies have invested heavily in TSMC for assurance of yield and capacity. They’re not ready to switch just yet.” This reliance on internal production highlights the immediate need for Intel to secure external foundry clients to validate and scale its new technology.
Intel’s strategy hinges on its state-of-the-art Fab52 facility in Chandler, Arizona. CNBC was granted an exclusive look inside this cutting-edge plant, which represents a significant investment in Intel’s foundry ambitions. This facility is positioned to compete directly with TSMC’s new fab in Phoenix, which is producing chips using 4-nanometer technology, with its more advanced 2nm process still exclusive to Taiwan.
The 18A node is technically on par with TSMC’s 2nm offering in terms of transistor density. However, Intel is still working through the complexities of mass-producing at this advanced scale, a process that has historically involved challenges with yield – the number of usable chips produced per wafer. This is not an uncommon issue for new advanced nodes, as Professor David Yoffie of Harvard Business School noted, drawing parallels to early yield issues experienced by Nvidia’s Blackwell GPUs at TSMC, which were subsequently resolved.
Intel’s renewed focus on its foundry business, manufacturing chips for external clients, was a strategic pivot initiated by CEO Pat Gelsinger in 2021. Following a period of significant investment and ambitious targets, the company’s leadership has seen a transition, with Lip-Bu Tan taking the helm as CEO and Naga Chandrasekaran now leading the foundry division.
“Over the past few years, the company invested too much, too soon – without adequate demand,” Tan acknowledged in a company memo, signaling a shift towards a more disciplined approach to capital allocation.
The U.S. government has stepped in to support Intel’s foundry ambitions, acquiring a 10% stake in the company through an $8.9 billion investment, largely funded by the CHIPS Act. This government backing underscores the strategic importance of bolstering domestic semiconductor manufacturing capabilities. Additionally, SoftBank invested $2 billion, and Nvidia made a $5 billion investment, though Nvidia’s commitment is for the use of certain Intel technologies rather than a direct foundry engagement.
**The Fall of a Giant and the Road to Redemption**
Founded in 1968 by Silicon Valley pioneers, Intel was instrumental in the development of the microprocessor. For decades, the company dominated the industry, embodying the rapid advancements described by “Moore’s Law.” However, Intel largely missed the mobile computing revolution and struggled to adapt to the evolving landscape of artificial intelligence.
The year 2024 marked Intel’s worst performance, with its market value plummeting by approximately 60%. This decline was exacerbated by multi-year delays in its 10nm and 7nm process nodes, partly attributed to a cautious approach toward adopting ASML’s costly Extreme Ultraviolet (EUV) lithography technology.
“We lost the discipline of cycle time,” admitted Jim Johnson, head of client computing at Intel. “Cycle time requires you to commit and deliver, and we started talking ourselves into, hey, we can have longer cycle times and try and lift more or do more.” To address this, Intel plans to deploy at least 15 EUV machines in its new Fab52 facility.
By 2021, TSMC had established itself as the clear leader in advanced manufacturing, prompting Intel to begin outsourcing some of its own leading-edge chip production. Concurrently, Apple transitioned away from Intel processors for its Mac lineup, opting for its in-house M-series chips, also largely manufactured by TSMC.
“Intel may now be considering a deal to buy custom AI chip design startup SambaNova for $1.6 billion, though the company declined to comment on the matter,” according to recent reports. This potential acquisition signals Intel’s intent to deepen its engagement in the AI chip market.
**A Culture Shift and a Focus on Execution**
Under Pat Gelsinger’s leadership, Intel embarked on an ambitious roadmap to reclaim technological parity within four years. Now, under Lip-Bu Tan, the focus remains on manufacturing prowess, with Naga Chandrasekaran spearheading the foundry efforts.
“We are making yield improvements, defect density improvements, month-over-month and hitting our goals,” Chandrasekaran stated. “So I believe we have turned the corner.” Chandrasekaran, who joined Intel last year after more than two decades at Micron, emphasizes his primary objective: attracting foundry customers.
“I have to become part of their team and convince them that they can trust Intel Foundry to execute,” he explained. “That’s number one. And to do that, we are changing our culture. We are bringing a huge execution focus internally into Intel Foundry.”
Chandrasekaran revealed that Fab52 has the capacity for over 10,000 18A wafer starts per week. The Arizona campus boasts over a million square feet of cleanroom space, with five interconnected fabs. A sixth fab, Fab62, is anticipated to be operational by 2028.
The 18A process incorporates RibbonFet technology, Intel’s gate-all-around transistor architecture designed to enhance power efficiency. Chandrasekaran claims 18A offers a “more than 15% performance per watt improvement” over Intel 3.
Intel is also differentiating itself through advanced packaging solutions, which involve the sophisticated assembly and interconnection of chips. These techniques are crucial for optimizing performance and mitigating power consumption, a growing concern for data center chips.
Furthermore, Intel’s Arizona facilities are committed to sustainability, operating on nearly 100% renewable energy. The company has also implemented robust water recycling systems, returning a significant portion of its water usage to the local supply.
**Disciplined Investment and Strategic Partnerships**
CEO Lip-Bu Tan has signaled a more fiscally conservative approach, stating, “No more blank checks.” This sentiment reflects the need for tangible customer commitments to justify further investment in future process nodes. The delayed opening of Intel’s Ohio fab until at least 2030 and significant workforce reductions underscore this new era of cost discipline.
“That’s what the company needed,” commented Newman. “It needed to be faster. It needed to be leaner. It needed to be more focused. It needed someone that would be a little bit more shrewd.”
While Tan is cautiously assessing demand before detailing plans for the 14A node, Chandrasekaran confirmed its development in Oregon, targeting volume production in 2028.
Securing external customers for the 18A node presents a unique challenge for Intel. Unlike TSMC, which operates solely as a foundry, Intel also manufactures its own branded chips, positioning itself as a direct competitor to potential clients. This inherent conflict of interest raises concerns for companies like Nvidia, AMD, Qualcomm, and Broadcom, as expressed by Professor Yoffie.
“If I’m an Nvidia or AMD or Qualcomm or Broadcom, do you really want to put your secret sauce into a manufacturing operation where you’re giving Intel access to that secret sauce?” Yoffie questioned, suggesting that separating the foundry business into a distinct entity could enhance its appeal and strengthen the U.S. position in semiconductor manufacturing.
Intel is hoping that the upcoming Panther Lake processor, featured in PCs from major brands like Samsung, Dell, HP, Lenovo, Asus, and Acer launching in January, will serve as a significant proof point for the 18A node. The new Xeon 6+ data center chip is also manufactured on 18A.
“If you’re a major company that wants to bet on a process node, you’re going to feel a lot more comfortable if you see Intel ramping the heart of their client product line to high volume on that process node,” stated Jim Johnson.
Early foundry agreements with Microsoft and Amazon for custom chips offer initial validation, though Yoffie notes their volumes are currently modest compared to major chip companies. However, emerging reports suggest AMD may consider Intel for manufacturing, and some analysts predict Apple could return to Intel for some Mac chip production by 2027.
The U.S. government’s substantial investment signifies a strategic commitment to reshoring critical technology manufacturing. This move is particularly significant given that approximately 92% of the world’s most advanced chips are produced in Taiwan, a consequence of decades of declining U.S. semiconductor manufacturing capacity.
“The stakes are incredibly high for Intel, for the U.S. and for the world,” Yoffie emphasized. “The whole idea that the world’s most advanced products are dependent on a single location in an island a few miles off the Chinese coast is a terrible situation for the whole world to have to deal with.”
Chandrasekaran remains resolute in his mission: “As a semiconductor community, we have to enable this solution for the world to move forward with AI. There’s no other option than to be successful.”
Original article, Author: Tobias. If you wish to reprint this article, please indicate the source:https://aicnbc.com/14760.html
