Chip giant Intel has thrown down the gauntlet in the semiconductor industry. Chief Executive Lip-Bu Tan declared that the global foundry market must break free from Taiwan Semiconductor Manufacturing's dominant grip, urging that production orders be distributed across multiple vendors to keep pace with the explosive demand for AI chips.

In a bold keynote address at the Splunk.conf26 conference in Denver, Colorado on April 14, Tan made it clear that Intel's future hinges on revitalizing its foundry business. He pointed out that the current CPU market is witnessing a surge in demand so intense that supply constraints will persist for the foreseeable future.

Where the industry is headed

The semiconductor landscape is shifting toward integrated system solutions and advanced packaging technologies, Tan argued during his conversation with Cisco's President and Chief Product Officer Jeetu Patel. Tan stressed that possessing the trifecta of design, manufacturing, and advanced packaging capabilities is critical. "That's precisely why we are diving into this arena," he said, emphasizing Intel's iconic standing both domestically and globally.

Tan didn't mince words about the sector's difficulties. Foundry operations demand cutting-edge process technology, pristine yields, rigorous defect control, predictable production cycles, and minimal process variation. Each client requires custom intellectual property and access to electronic design automation tools. Advanced packaging, particularly integrating CPU, memory, I/O, and silicon dies into a single package, remains a monumental engineering challenge that demands deep technical expertise.

Why just 10 ASX 200 shares?

The CEO's message was unambiguous: the stranglehold that Taiwan Semiconductor Manufacturing holds over AI chip production must be shattered. "The industry is moving toward system-level solutions and packaging technology - that's the future," Tan stated. "Relying on a single company for 95% of capacity carries extreme risk." His words were a pointed reference to major US players like Nvidia, AMD, and Apple, which currently hand nearly all their AI chip manufacturing to Taiwan Semiconductor Manufacturing. The remarks also signaled Intel's ambition to capture more foundry orders, aligning with the Trump administration's policy priorities.

Tan acknowledged the formidable capital expenditures and meticulous planning required to succeed in foundry work. "The good news is that things have improved dramatically over the past 18 months," he noted.

Intel's progress on its 14A (1.4-nanometer class) process initially encountered yield challenges. However, the company struck a deal in April to supply its 14A technology to Terafab, a venture led by Tesla and SpaceX. The first Terafab product is expected to debut by mid-2028.

In another strategic move this past June, Intel recruited former SK Hynix CEO Lee Seok-hee as a senior vice president to oversee advanced packaging and back-end technology development within its foundry division. The high-profile hire brings a former Intel executive back into the fold to strengthen back-end manufacturing and tackle yield issues. Industry watchers speculate that SK Hynix may soon collaborate with Intel on producing fundamental logic chiplets needed for high-bandwidth memory.

Tan also voiced strong confidence in Intel's EMIB advanced packaging technology, which uses silicon bridges to connect multiple chiplets so they operate as a single cohesive chip.

CPU demand outpaces supply

Shifting focus to Intel's core product line, Tan projected that current shortages will only deepen as the number of AI agents deployed for inference tasks climbs into the trillions. While GPUs dominate model training, CPUs serve as the orchestrators, managing the coordination of numerous parallel GPUs and running critical applications simultaneously. This positions the CPU as the linchpin of the AI inference era.

"CPU demand is exceptionally strong - we can only fulfill about half of what customers are asking for," Tan admitted. "I receive calls from CEOs regularly, and I can only apologize that our capacity is insufficient."

He elaborated: "We're witnessing the rise of massive numbers of AI agents. In the future, we're looking at millions, even trillions of these agents, all requiring compute resources. We need abundant processing power and security capabilities to support this reality."