Over the past decade or so, investors have seen Taiwan Semiconductor Manufacturing (TSM 6.29%) surpass Intel (INTC 6.53%) in terms of process technology.
As a result, virtually all major chipmakers, from Apple to Nvidia and many others now fully depend on TSMC for their chipmaking needs. In fact, even Intel is using TSMC to produce its latest client PC chips.
However, Intel is on the cusp of completing its "five nodes in four years" turnaround, while management promises its technology will regain and even surpass TSMC's next year. And the following statistic shows just how big a bet Intel's management has made.
NASDAQ: INTC
Key Data Points
Intel has more PP&E than TSMC
The foundry business is extremely capital-intensive, with chipmakers having to build huge manufacturing plants and fill them with extremely expensive capital equipment to produce semiconductors. This is why so many semiconductor designers have abandoned producing their chips altogether.
As TSMC overtook Intel in leading process technology, allowing Intel rivals and TSMC customers like Advanced Micro Devices and Nvidia to leap ahead, it's not surprising that TSMC's investments in property, plant, and equipment overtook Intel's in late 2020. But what's surprising today is that after Intel began ramping its PP&E investments in late 2021, it has now surpassed TSMC once again in terms of PP&E on its balance sheet (of note: the chart below has the wrong figure for TSMC in the most recent Q3 2024 quarter, in which TSMC recorded $97.1 billion in PP&E).
INTC Net PP&E (Quarterly) data by YCharts
Intel retaking the PP&E crown is all the more shocking since Intel's revenues have gone in the opposite direction since 2022.
INTC Net PP&E (Quarterly) data by YCharts
Not only does Intel's PP&E surge stand in contrast to its revenue declines, but company's PP&E is even understated by these figures. That's because Intel is actually currently using TSMC to produce its most advanced PC chips, including the new Lunar Lake and Arrow Lake CPUs, as well as its Gaudi accelerators.
So with Intel's revenue declining and with its most advanced PC chips today now produced by TSMC, what's the deal with Intel's surging PP&E?
Build it and they will come
The answer lies is Intel CEO Pat Gelsinger's ambitious plan to catch up to TSMC by 2025. Gelsinger was appointed CEO in February 2021, and soon came up with a plan to hit, "five nodes in four years," a really fast pace of innovation, while also becoming a foundry for other chip designers to become a true competitor to TSMC.
The ambitious plan required a massive amount of investment years before revenues would theoretically start flowing from those investments. Even though the plan began in 2021, the first node in the plan, Intel 7, entered volume manufacturing in 2022. The next node, Intel 4, which was the first Intel node to use EUV lithography, entered high-volume manufacturing in September of 2023. Intel 3, the third node, entered high volume manufacturing in June 2024.
Intel then actually decided to skip its 20A node, the fourth node in the plan, and outsourced production of the 20A Arrow Lake desktop processor products to TSMC. This was done to accelerate the development of 18A, the fifth node, while also saving money on capital equipment.
Thus, a lot of Intel's investments will potentially pay off next year. That's when Intel 3 data center chips will have ramped to high volumes, Intel's new advanced client chips built at TSMC will be ramping as well, and 18A will start cranking out its first products.
Of note, Intel's current revenues mostly reflect products built on just the first two of the five nodes in the plan. On the second quarter conference call with analysts, CFO David Zinsner noted that over 85% of Intel's wafer volume was still coming from pre-EUV nodes. Thus, current revenues and margins don't reflect the fruits of the company's massive investments.
All eyes on 18A and 2025
Thus, mid-to-late 2025 will be when the payoff should present itself. That of course assumes a leap of faith that Intel has designed good products, and that it will produce them at acceptable yields.
That's a skillset that was lost under prior Intel management, but appears to be back on track under Gelsinger. Of note, Gelsinger said back in September that 18A had achieved a "defect density" of 0.4. That would be a solid figure that, while not high-volume ready, does indicate 18A is on schedule at this stage of the process, and should be ready for high volumes by mid-2025, hitting the target laid out four years ago. Intel has also said it expects to produce eight "tape-ins," or chip designs, on 18A next year, including both internal and external foundry customer products.
So will this massive bet pay off? Intel investors should pray it does, while TSMC and Samsung investors should pray it doesn't. Only time will tell.
