"You Don't Have This"…ARM CEO's Remark to Intel CEO [Tech Talk]

Prerequisite for Foundry Success: 'Order Volume'
Both TSMC and ARM Achieved Success by Securing Volume
Intel Lacks Significant Third-Party Volume

Ren?e Haas, CEO of ARM, recently recalled a conversation with former Intel CEO Pat Gelsinger during a podcast interview with the American tech media outlet 'The Verge.'

At the time, former CEO Gelsinger was considering various strategies to transform Intel into a semiconductor foundry company. Unlike now, investors had high expectations that Gelsinger would turn around Intel's sluggish performance. Haas said he advised Gelsinger, "You need large volume orders."

According to Haas, Gelsinger did not take ARM's advice seriously at the time. While Haas expressed regret over Intel's crisis, he argued that the situation might have been different if Intel had collaborated with ARM.

Intel's Crossroads: Vertical Integration vs. Collaboration

Pat Gelsinger, former CEO of Intel. Photo by AFP Yonhap News

Why did Haas emphasize the need for 'large volume orders' to Intel? And what is the connection between large volume orders and foundry business? Currently, Intel is one of the few true integrated semiconductor companies worldwide. This means it can perform both semiconductor design and manufacturing processes simultaneously.

However, Intel's semiconductor manufacturing division, 'Intel Foundry Services' (IFS), is hardly an independent company. This is because the orders Intel Foundry receives are practically only for chips designed by Intel itself. In the past, when Intel almost monopolized markets from PCs to data center central processing units (CPUs), Intel's own demand was sufficient to sustain the foundry. But the situation is different now.

Haas pointed out that Intel's true ambition was 'vertical integration of advanced semiconductors.' This strategy aimed to regain control over all processes from semiconductor design to manufacturing. If realized, Intel could have once again stood tall as a giant in the semiconductor industry, but "vertical integration requires enormous costs and time," which was the problem. Therefore, ARM's diagnosis was that Intel first needed to secure large volume orders from third parties to mature its foundry business.

TSMC and ARM Also Grew Thanks to 'Order Volume'

Ren?e Haas, CEO of ARM Holdings. Photo by Hyunmin Kim

Haas's argument aligns with the success formula of TSMC, currently the world's largest semiconductor foundry. TSMC stabilized its semiconductor manufacturing processes through large volume orders and was able to outpace competitors like Samsung Foundry in advanced process competition. In fact, Samsung was recognized for its technology, competing neck and neck with TSMC in the 10-nanometer (nm) process around 2016.

TSMC emerged as a dominant company in both market share and technology after breaking through the 10nm barrier and securing orders for Apple's proprietary application processors (AP) in the A series and M series.

With a large volume of orders, naturally, more chips can be produced. For manufacturers at the time, it was an opportunity to accumulate know-how in operating the then-novel EUV lithography equipment, optimize the latest wafer production processes, and, most importantly, collaborate more closely with customers entrusting their foundry production. In semiconductor manufacturing, where increasing yield is a top priority, securing as many orders as possible is essential to stay ahead.

ARM also understands the power of 'order volume.' Unlike typical fabless companies, ARM provides various intellectual properties (IPs) that form the 'foundation' of semiconductor design. Countless companies manufacture semiconductors using ARM's design IP, and over 30 billion ARM chips are produced worldwide annually.

ARM architecture-based chips have secured numerous clients, and the volume of orders itself becomes a competitive advantage. Photo shows Alibaba's ARM computer chip for servers. Photo by Yonhap News

Not all chips are advanced logic processors. Among ARM's design assets are simple microcontrollers and many 'legacy process' chips developed with nodes larger than 20nm. However, all these contribute to enriching ARM's ecosystem. New design techniques developed in collaboration with ARM's customers remain as libraries, attracting more clients.

Haas appears to have proposed that ARM could share the benefits of 'large volume orders' with Intel Foundry in this way. Of course, even if Intel Foundry had collaborated with ARM to produce semiconductors, immediate noticeable profit increases might not have occurred. Unlike vertical integration businesses, collaborative systems require sharing a portion of total profits. Instead, Intel could have secured the most important 'customers' when growing its foundry.

"Intel Foundry Has No Volume"

As former CEO Gelsinger retired, taking responsibility for Intel's sluggishness, dark clouds seem to be gathering over Intel's future. If Intel's downfall or spin-off truly materializes, it will cause a massive tectonic shift in the semiconductor industry worldwide, not just in the U.S. Above all, the possibility that the instruction set architecture (ISA) 'x86,' which has dominated the CPU industry thanks to Intel, could be replaced by 'ARM' cannot be overlooked.

As the 'only' domestic advanced semiconductor manufacturer in the U.S., some voices say that Intel's downfall cannot be watched passively. Some suggest that at least Intel's semiconductor manufacturing division should be maintained separately. However, once again, 'large volume orders' hinder these opinions.

Chris Miller, author of 'Chip War,' evaluated the possibility of Intel Foundry's independence in an interview with 'Bloomberg' on the 11th: "(Gelsinger's strategy of pursuing vertical integration) is close to a decision reflecting Intel's reality. At least in the short term, Intel Foundry existed solely because of orders for Intel's own products. In fact, Intel Foundry has no (third-party) order volume at all."