Splitting Intel’s Fabs From Its Design Business Now Doesn’t Make Sense
Intel has been having a rough go of it, and lately it’s been “knives out” among some of its competitors who would love nothing better than to see it fail. For starters, the company’s huge capital outlay to expand its manufacturing and supply chain—especially CEO Pat Gelsinger’s ambitious “five nodes in four years” plan—hasn’t yet been matched by increased revenue. Worse, the forecast for the coming quarters isn’t as robust as you’d like to see, either. This, of course, is from an investor point of view.
Beyond that, Intel is facing more competitive pressure from more rivals—starting with Nvidia, AMD and Qualcomm—than it has since it pivoted from memory chips to microprocessors in the 1980s. The fact is, Intel shot itself in the foot under previous management years ago with some of its transistor technology and process nodes, and it’s hard to come back from that. When you have these issues, it impacts your designs. For example, if architects had originally planned for 16-core chips, they would need to go to 12 cores to accommodate power limits or wafer size.
The best bet is that a full turnaround is going to require a couple more years at least. Things have been bad enough to force major adjustments, most obviously in Intel’s announcement a month ago that it would cut 15,000 jobs. That’s part of $10 billion in cost savings promised for 2025. All job cuts aren’t bad, but they do disrupt everything unless you sell a business unit wholesale or shut down a science project without revenues or profit.
And then Bloomberg put the icing on the cake when it reported that Intel is considering a host of other options, most notably the possibility of splitting off its manufacturing and foundry operations (Intel Foundry) as a separate business. But while drastic change may be necessary to ensure Intel’s recovery, splitting off manufacturing from design wouldn’t do the company much good. Let’s break down the pros and cons.
Full disclosure: While this article reflects my independent view as an analyst, Intel is a client of my firm, Moor Insights & Strategy, as are AMD, Qualcomm, Arm and Nvidia.
The Potential Upside Of A Freestanding Intel Foundry Business
The only thing Intel stands to gain from this split is the chance that an independent foundry business could bring in new customers that compete with Intel products—and that don’t trust Intel’s design side not to snoop on their own chip designs.
AMD and Qualcomm are at the head of this list. I believe that if Intel’s new 18A process node is as strong as Intel says—and the defect density numbers do look good so far—AMD might need to adopt it. At least, that could happen if AMD can’t get the second-sourcing foundry capacity it needs from Samsung as we see TSMC continuing to raise prices. Mind you that AMD really will need a second source for production capacity to match its ambitions, which I recently discussed in the context of AMD’s planned acquisition of ZT Systems.
Meanwhile, Qualcomm has a strong tendency to go with the best available production technology wherever it’s available, so Intel’s more advanced 14A process node might be a fit—though it’s also likely more than a year away from production. It’s worth remembering that Qualcomm contracted with Samsung in an arrangement like this for a few production cycles. Qualcomm also multi-sources its RF across many vendors. But Qualcomm doesn’t have any reason to trust Intel if a fab spinoff is essentially a tracking stock; it would have to be a genuine split into a fully independent foundry company.
I will add that Intel 14A is likely a more appropriate node for mobile SoCs. 18A does have a low-power flow, but I’m not hearing anything positive for smartphone or IoT mobility. If Intel announces a smartphone SoC or IoT vendor as a customer, my opinion would change. There is promising news out of Intel about 18A. Intel’s AI PC client processor (Panther Lake) and its server processor (Clearwater Forest), are out of the fab and have powered on and booted operating systems. Additionally, Intel announced that the first external customer tapeouts are coming in the first half of 2025. Microsoft is one of those, and I am guessing it is for Maia, its homegrown datacenter AI accelerator.
Keep in mind that any scenarios with AMD, Qualcomm and other potential foundry customers beyond what’s already been announced are hypotheticals. There could be a future in which Intel Foundry picks up more new and sizable customers like these, but that future would look very different from the status quo. Today, Intel Foundry mostly produces Intel’s own chips; there aren’t many other customers, at least so far.
Why A Split Like This Doesn’t Make Sense Now
Sometimes there is a compelling reason—or at least a good argument—for splitting a big company into independent parts. The history of Hewlett Packard provides multiple examples from different periods, including the spinoff of Agilent in 1999 and the split of HP and HPE in 2015.
But this isn’t that. Based on Wall Street’s math, the current value of Intel Foundry is less than zero, which means that a spinoff would likely need to be private. How much value that would create is dubious. And there would still be all the challenges (and expense) of building up a foundry customer base separate from the Intel mothership.
Could we come up with a scenario for taking it private? Sure, if we want to get even more hypothetical about the specific mix of debt, capital allocation and investments from customers (Apple, Nvidia and Broadcom?) that could make it happen. You still wouldn’t be talking about a profitable freestanding foundry company before 2027 or 2028. And that’s before we get to the incredible upheaval for Intel’s business model and internal culture if you made the design side of the business into a fabless company.
Long story short: if Intel spins out Intel Foundry before the design business and the foundry business are both healthy, it will fail.
What’s Really At Stake?
Though there are some bright spots, Intel is certainly in a serious crunch. From all accounts, it is taking what steps it can to reduce costs, including slowing its factory expansion plans and nixing projects that don’t have a clear path to profitability sooner than later. It’s also moving resources from the 20A production node to 18A, which makes sense to me as a short- to medium-term move because 18A seems to be on the right track. Putting more engineers on it ought to reduce design risks and make it even better.
If we want to take the optimistic view, it could turn out that the 18A and 14A production nodes turn into real winners, and I’m on the record about my belief that Intel could start taking some datacenter AI market share from Nvidia in 2025 and 2026. It should also get a boost when its Lunar Lake processors start making their way into large numbers of AI PCs in the coming quarters, and somewhere down the line—2026 or so—Intel should have a proper datacenter GPU (not the current Gaudi ASIC AI accelerator) to directly compete with Nvidia’s.
With all of those wheels in motion, why consider an Intel Foundry spinoff that doesn’t seem to offer a concrete payoff that would make the move worthwhile? My bet is that Intel’s leadership is doing it to ward off a potential coup, possibly from recently resigned board member—and semiconductor industry icon—Lip-Bu Tan. By going through detailed scenarios with its bankers from Goldman Sachs and Morgan Stanley, Gelsinger and the rest of the board can say to investors, “We did the analysis, and here’s what it says. It’s not a smart idea to split off Foundry right now. Here are the conditions where a split would make sense.”
Call it preventative due diligence. Intel has enough problems to face without further loss of investor confidence, or a potential C-suite shakeup driven by Tan or others. So it’s in Intel’s interest to let investors know that it’s looking at all the options—even if the idea is a non-starter. Maybe later when the foundry business and design business are humming, but not now. If you want to research a similar split, look no further than my alma mater, AMD, as it split off its design and manufacturing business as Globalfoundries. AMD had to do this to survive, and while Globalfoundries is very strong today, at the time it was hemorrhaging money and was privately funded by Mubadala Technology Investment Company. And so far there’s no scenario like that on the horizon for Intel.
link