Apple Says “Pat G. Done Good”

My first two reactions to the news that Apple (NASDAQ: AAPL) will likely shift some of the manufacturing for its M-series chips away from Taiwan Semiconductor Manufacturing Company Ltd. (NYSE: TSM) and to Intel (NASDAQ: INTC) were (1) skepticism and (2) despair.

The skepticism arises because for so many years, manufacturing was Intel’s weak point. As our colleague Dr. Robert Castellano has put it, the problem with Intel Foundry Services was that it kept promising to make for others chips it could not even make for Intel.

The despair comes in because if the U.S. government, by subtle and not so subtle political pressure plus massive subsidies, can actually steer major strategic decisions by Apple, one of the most valuable companies on earth, that only confirms U.S. industrial policy under Trump is all too real and all too powerful.

The appropriate amount of influence the U.S. government should exert over where U.S. companies get their manufacturing done is zero, with the sole exception being procuring actually secret and sensitive military systems. Historically, this problem has come up far less often than most politicians seem to believe.

The great to dual-use myth seems to imply that we can and should exclude from e.g., China, any civilian technology that could be put to military use. The truth is closer to the opposite. A perpetual problem for the U.S. Department of Defense (DoD) is that our defense procurement process is so slow that DoD’s deep tech military capabilities often lag behind the civilian sector by several years. “Dual-use” in the United States should be renamed “delayed use.”

The United States is not Israel, where the military routinely pioneers tech progress ahead of the civilian sector. In Israel, most new military tech comes from the military or recent military veterans whose start-up light bulbs went off back while they were still in service. The Israel deep tech sector is led primarily by alums of Israel’s elite intelligence units.

I continue to believe that U.S. government influence over the decisions of U.S. businesses will build towards disaster. Nevertheless, in Intel’s case, there are some other things going on. One is that Pat Gelsinger, Intel’s CEO for an all-too-brief four years, accomplished a lot more then he’s usually given credit for.

For more than a decade before Gelsinger, Intel was lost in the mists of vaporware, repeatedly announcing new products that it ultimately failed to deliver. Intel stayed on “14nm” for far too long, then massively over-reached on 10nm, packing in too many (admirable) design changes too fast. It got stuck in a loop of low yields and repeated delays. Then Intel’s “7nm” (later rebranded as Intel 4) also slipped. Advanced Micro Devices (NASDAQ: AMD) (fabbed at TSMC) used that window to catch up and then pass Intel in several client and server segments.

Pat Gelsinger’s “five nodes in four years” plan (Intel 7 → 4 → 3 → 20A → 18A) was the attempt to dig out of that hole. We now have some real datapoints. And it mostly worked.

• Intel 7 is in high-volume manufacturing (Alder Lake, Raptor Lake, Sapphire Rapids) etc.
• Intel 4 is in volume production for Meteor Lake.
• Intel 3 is used for Sierra Forest and Granite Rapids server chips, launched in 2024.

Those three nodes are “proof-of-life” for Intel’s fabs. They may be still behind TSMC at equivalent nodes, and Foundry is still losing money, but is executing new nodes again.

That matters because it supports the idea that Apple really could use Intel 18A, which clearly is more than vaporware. 18A exists in real silicon and real programs. Especially notable is its early adoption by the U.S. Department of Defense under the RAMP-C program. RAMP-C tries to solve the old problem of Department of Defense technology lagging the commercial sector by several years. Under RAMP-C, military contractors can use the most recent viable technology if approved by the DoD. And Intel 18A already has that approval.

Intel 18A is the first commercial node with both:

• RibbonFET GAA transistors, which suppress electron leakage, and
• PowerVia backside power delivery, which reduces power requirements.

This combination could provide Apple with better performance per watt and cooler devices. Since Apple focuses on mobility, these are crucial parameters.

Though he has been gone almost a full year, in retrospect, Gelsinger’s tenure looks brilliant:

• He prevented Intel from becoming permanently irrelevant.
• He re-established a credible technical path by actually (almost) delivering on “five nodes in four years.” (Intel 7 → 4 → 3 → 20A → 18A) True, 20A was never deployed but getting to 18A more than compensates.
• Intel Foundry services, still losing money, has actual paying customers and with increased prospects from Microsoft and Apple
• Without Gelsinger, 18A doesn’t exist.

And without Gelsinger, the CHIPS Act doesn’t prioritize Intel. We hate the CHIPS Act, but good for Pat scoring the money.

If Intel survives and history is written fairly, Pat Gelsinger will get the credit.