[Apple Shockwave](39) NVIDIA, Did They Buy ARM for This?

NVIDIA Begins ARM-Based CPU Design Amid Lunar New Year Spread
Synergy Expected from Combining GPUs and CPUs Powered by Generative AI
Card to Block Rise of ARM-Based Apple Computers
Market Cap Gap with Apple Expected to Narrow if 3Q Earnings Stron

Editor's Note[Apple Shockwave] is a content series that explores the upheaval triggered when Apple entered the semiconductor market. You might wonder: why would Apple get involved in semiconductors? Today, Apple is no longer just a company making smartphones and computers. After a long period of effort that began with the late founder Steve Jobs, Apple has succeeded in designing the world’s leading semiconductors for mobile devices. Just as Intel dominated the PC era, Apple has emerged as the top predator in the mobile semiconductor ecosystem. Amid a global semiconductor supply chain crisis and massive investments in semiconductor production lines, Apple Silicon has brought about dramatic change, and we will closely examine its impact and future prospects for our readers. Apple Shockwave will greet you every weekend, and after serializing more than 40 episodes, it will be published as a book.

NVIDIA, now enjoying its greatest heyday with its generative artificial intelligence (AI) chips, has revealed a hidden “claw.” The company is preparing to enter the central processing unit (CPU) field in earnest—a sector many thought NVIDIA had abandoned. Multiple reports that NVIDIA is working on a PC CPU based on ARM architecture, not Intel’s x86, are notable. This could lead to a scenario akin to when Intel, which once dominated the CPU market, bundled its own graphics chips (GPUs) with CPUs. In fact, even Apple is now within NVIDIA’s scope. A showdown between the so-called “Green Team” representing NVIDIA in the AI world and Apple is now a possibility.

NVIDIA’s recent moves appear to be a countermove in a landscape where Apple, Microsoft, and others are bolstering AI chips to power “on-device” AI on terminals. While Apple is also preparing generative AI, its fundamental direction is on-device AI. For example, the neural processing unit (NPU) responsible for AI computations in the A17 Pro chip of the iPhone 15 Pro delivers roughly double the performance compared to the A16. Intel, on the Windows side, has also jumped into the on-device AI game, with its latest Meteor Lake CPUs featuring NPUs. Enhancing AI capabilities relies not on boosting CPUs, but on improving GPU or NPU performance since on-device AI computations happen directly on users’ devices. While generative AI is mainly the concept behind cloud services built on NVIDIA chips, it stands in marked contrast to on-device AI.

Although generative AI, led by OpenAI, has risen rapidly, on-device AI currently offers users more direct benefits. It can boost a wide variety of everyday functions on smartphones. In addition to Apple, Qualcomm and Samsung Electronics are also strengthening their smartphone application processors (AP) with on-device AI, intensifying competition with generative AI solutions.

To defend its dominance in AI against this push from AP players, NVIDIA has introduced a new strategy: an offensive targeting CPUs, the traditional stronghold of its rivals. This appears to exploit a notable vulnerability, as progress in the CPU sector has slowed. For example, even with the world’s first 3-nanometer process applied to Apple’s A17 Pro, improvements in performance were not dramatic—an indication that CPU innovation may have plateaued, giving NVIDIA more room to catch up.

The CPU has long been something of a sore spot for NVIDIA. After acquiring PortalPlayer—the key chip designer first used in Apple’s iPod MP3 player—NVIDIA launched its Tegra chip. As previously discussed in episode 3 of Apple Shockwave, PortalPlayer, Apple, Samsung, and NVIDIA formed a complex “quadrangle.” While NVIDIA, renowned for its graphics processing prowess, set its sights on entering the AP market through Tegra, this remained more hope than reality. Though used in products like the Tesla Model S and Nintendo Switch, Tegra’s market impact was negligible compared to NVIDIA’s soaring GPU business amid the cryptocurrency boom and explosive growth of generative AI. In this context, Tegra’s reputation seemed even more dwarfed.

However, recent developments have reconfirmed that NVIDIA has not given up on CPUs. Some international media have reported that NVIDIA is now designing ARM-based CPUs for Windows PCs. The fact that this is for PCs, not mobile devices, is significant.

In the current PC market—especially for laptops—Apple Silicon-powered Mac computers have witnessed remarkable growth. The fusion of Apple Silicon and Mac has fueled rapid expansion, propelled by a surge in PC replacements following COVID-19. As Apple’s chips enabled longer battery life and outstanding performance compared to Intel’s chips, consumers flocked to Apple computers. For Microsoft (MS), the developer of Windows OS, this surge in ARM-based PCs, a category in which MS had little presence, represents a considerable threat. For MS, which once drove Apple to the brink of bankruptcy in the dawn of the personal PC era with Apple II, this shift is far from welcome.

For MS to support ARM-based PCs, it needs suitable chips. While Qualcomm recently unveiled the “Snapdragon X Elite,” this alone will not make competition with Apple easy. NVIDIA’s plans to launch an ARM-based CPU compatible with Windows PC align closely with MS’s strategy to expand the ARM Windows PC ecosystem. Industry forecasts suggest that an NVIDIA CPU could debut as early as the second half of 2024 or in 2025. Notably, NVIDIA already possesses the Grace CPU: a high-performance ARM-based processor for servers.

This dynamic creates mutual benefit for both NVIDIA and MS. Reports have even emerged that AMD, the prominent player in the x86 camp, is also developing ARM-based CPUs.

NVIDIA Grace CPU

NVIDIA’s previous attempt to acquire ARM also highlights its ambitions in ARM CPU development. The company tried to pull off one of the most dramatic mergers and acquisitions in semiconductor industry history when it moved to purchase ARM—put up for sale by Japan’s SoftBank in 2020. At about 40 billion dollars, it would have been by far the largest M&A deal ever in semiconductors. Had it succeeded, it would have been the deal of the century. For NVIDIA, its failure remains a painful memory. With its market capitalization soaring past 1 trillion dollars on the back of the generative AI boom, the company might have achieved an even higher valuation had it secured a new growth engine in CPUs.

NVIDIA’s current market value towers over other semiconductor companies, surpassing Intel, AMD, and TSMC as the unrivaled number one. Last quarter as well, its stock price surged beyond market expectations—and again, investment banks are predicting another “earnings surprise.” BOA Merrill Lynch forecast that NVIDIA’s earnings per share will increase by 475 percent year-on-year to 3.34 dollars, issuing a “buy” recommendation and setting a target price of 650 dollars. The 22 percent surge in NVIDIA’s stock price this November demonstrates significant investor optimism about its results.

NVIDIA’s third-quarter results will be released on November 21, 2023. With this earnings announcement, the market is closely watching how much the valuation gap between Apple and NVIDIA will narrow.

This content was produced with the assistance of AI translation services.