The Real Story – A Column -Analysis by “Bagger” – Guest Commentary From A Learned Sage
Intel’s $90 Billion Question
This is the fifth story we have written about what is really happening with Intel, both worldwide and here in Hillsboro. Advising me has been a 45-year veteran of the Silicon Forest and its biggest players, whom I have come to call Bagger. He sees what others do not. This weekend, we did a deep dive and compared our notes; this is our report.
We feel Intel’s swings in our jobs, traffic, and housing. From the highly advanced Ronler Acres complex, where the parking lot is not as full as it once was, to Fab 4 in Aloha (Or), which opened in 1976, the company and our community share our past and our future. At this 2026 time marker, it was time once again to check in with my seer on all matters. Intel. Bagger Vance, as I have nicknamed him, sent me this latest take; it cut through a lot of Wall Street noise, so I’m sharing it here. I thought it might be opportune to take another look at Intel, considering its recent earnings and all the volatility surrounding the new year.
I asked my friend Bagger for his thoughts, and he started with a good question:
“How does Intel go about spending nearly $90B in capex (capital expenditures) over the past four years, and not have adequate capacity to meet the current market demand — all while producing only 10% of their 2025 wafers on that recent capex?”
Tech-fluent readers would say it takes a long time to bring new capex online, but when you walk through each investment, that excuse doesn’t completely hold up.
The Market’s Anger – and Intel’s Silence
The devil is in the details of what went wrong and who is to blame for Intel’s most recent miscues. As Bagger was quick to point out, the problems are likely to be resolved early this year, but that doesn’t excuse the way Intel has managed to confuse the market and the analysts who follow the company.
The anger was palpable on the most recent earnings call, with Stacy Rasgon of Bernstein asking some of the most direct and obvious questions about how Intel mismanaged this so badly. The market responded the next day with brutal selling, sending Intel shares down more than 17%.
All of this happened after Intel laid off tens of thousands of employees, under the banner that “better and fewer” people were needed. We all felt that, and we’re feeling it here in Oregon, with Hillsboro being Ground Zero. Yet it was with these “better/fewer” employees that they completely missed a major market shift in their core businesses — and failed to have the production ready to support that demand.
So with all that at the forefront, what went wrong, who’s to blame, and are they really fixing it? Here is one set of opinions worth considering.
Where Did All That Capex Go?
The best place to start is with their capex. Where did they spend? What does it build? When does it build it? How did they pay for it?
Chandler: Fab 42, 52, 62
In total, Intel will spend about $30B across the three Chandler fabs when done.
- Fab 42 was retrofitted for HNA (high numerical aperture) EUV (extreme ultraviolet) tools in 2022 and was previously home to many of Intel’s DUV (deep ultraviolet) client products, Raptor Lake et al.
- Fab 52 was finished in 2024, and the first tools were added in February of 2025. It too is HNA-ready.
Fab 42 and Fab 52 form the foundation for Intel 18A (Intel’s 18-angstrom process node), the first node to use BSP (Back Side Power) and GAA (Gate-All-Around) transistor technology. Intel will lead the industry in BSP deployment, delivering Panther Lake in Q4 2025 and ramping further in 2026.
Reading the Form 10-K (annual SEC filing) closely, it appears that Fab 52 capex has moved from “construction in progress” to a fully deployed asset, as noted by a change of roughly $15.9B in the 2025 10-K. That shift likely occurred earlier last year, as hinted by the increase in the depreciation expense run rate in Q2. That increase was painful: Intel didn’t have offsetting 18A revenue on Panther Lake until December.
None of that has been clearly explained to the investing public.
Key point: Intel turned on a very expensive Fab 52 before meaningful 18A revenue showed up, depressing margins and confusing investors.
The Problems That Fade… Eventually
There are more details that are important to Intel’s current predicament, but suffice it to say that each eventually leads to a very positive outcome for 18A, building Panther Lake out of Fab 42/52 in Arizona.
You got a bit of that on the earnings call, but nowhere near enough to draw reasonable conclusions on Intel’s client business going forward. It takes deep research to understand that some of these problems should fade in Q2 to Q3. Instead of explaining that timeline now, Intel offered an analyst meeting in the second half of the year.
Key point: Much of what looks broken is actually timing, but Intel hasn’t walked investors through it in plain language.
The Bigger Miss: Xeon Demand
Each of Intel’s other fabs has a similarly complex story that requires real homework to understand. But the most important question today is this:
How did Intel miscall demand on its Xeon products so badly that it failed to meet market needs for six months?
The answer hides in investments made by former CEO Brian Krzanich (“BK”) over a decade ago. BK’s decision to forgo EUV and double down on DUV tools ultimately proved to be his undoing. Those DUV tools sat in all of Intel’s main fabs:
- Fab 42 – Chandler, Arizona
- Fab 24 – Leixlip, Ireland
- Fab 28 – Kiryat Gat, Israel
These were Intel’s DUV workhorses that took Intel to near all-time highs under BK, but they also set the stage for his demise.
Pat Gelsinger comes along, recognizes the mistake, and starts moving the company to EUV tools — initially Low-NA EUV focused in Ireland, but eventually High-NA EUV starting in Oregon and propagating to Chandler.
What matters today about Chandler is that 18A is working and yielding. Client share should return to Intel as Panther Lake and Nova Lake on the 18A ramp, and the immediate problems below are resolved. On that front:
Congrats to the team. Congrats to Pat for putting in place all of these outcomes.
Key point: The long tail of BK’s “no EUV yet” strategy is still haunting Intel’s capacity mix, even as Pat’s 18A/High-NA bet finally starts to pay off.
The “Good Problem”: Granite Rapids Demand
Now to what Bagger calls “the kind of problem you want to have.”
Intel (under Pat) developed a Xeon product called Granite Rapids that has AI (artificial intelligence) features combined with classic Xeon CPU (central processing unit) management capabilities. That combination turned out to be perfect for solving “agentic AI” problems with customers like AWS (Amazon Web Services) and many others.
Once that solution was understood, those customers wanted all the Granite Rapids silicon Intel could provide. We can only speculate on how those conversations went, but the outcome has been clear in Intel’s own comments:
Demand for Granite Rapids and its predecessor Sapphire Rapids is much greater than supply, and that condition, according to customers and the company, will last through 2027.
Granite Rapids is mostly built on Intel 3 out of Ireland. Sapphire Rapids is built on Intel 7 out of Israel. Granite Rapids also uses an I/O (input/output) tile built on Intel 7.
Intel 3 in Ireland is a relatively underused facility (a very long story). Ireland’s output is data that can be understood because Intel has to pay Apollo (its co-investment partner) for every wafer shipped out of Ireland. Those payments are visible in SEC filings. So it doesn’t appear that the problem is Intel 3 capacity.
Intel 7 is another story.
The Intel 7 Chokepoint
That “ancient” DUV technology that BK took so far and which helped push him out is in very high demand today. Intel 7 also ended up filling much of Intel’s CPU demand for diverse reasons, not the least of which was the advantage it had in using memory types that are more readily available today.
Regardless of the reasons, Intel now has one of the hottest data center product stacks in the market in Granite Rapids and Sapphire Rapids — and without more Intel 7 capacity, that demand simply can’t be met.
Key point: The bottleneck isn’t fancy 18A EUV — it’s Intel 7 on DUV, the very class of tools Intel had been de-emphasizing and even selling.
To meet Xeon demand, Intel needs to kill wafer starts on its Intel 7 client CPUs and reallocate those wafers to build Sapphire Rapids and the I/O tiles for Granite Rapids. Again, that’s a good problem to have, except that the transition creates a hole in Intel’s production capacity and client revenue.
That might be acceptable as good business, except for one nagging problem.
Selling the Tools They Now Need Back
This surge in demand and the capacity shortage occurred at the same time that Intel was writing down and selling off the DUV tools required to build its way out of this problem.
Now, they need to buy more DUV tools to meet demand — tools they’ve been selling frequently over the past few years.
Yes, that part is on Pat, too.
Key point: Intel spent years cleaning up its DUV fleet — only to discover those “old” tools are exactly what they need to feed Granite/Sapphire demand.
From Cornfields to Capex: Hillsboro’s Stake in Intel’s Next Move
From the old pheasant fields to the cleanrooms at Ronler Acres and Hillsboro, our people have lived Intel’s ups and downs for decades. When Intel blows a quarter or misses a market shift, it doesn’t just show up on CNBC — it shows up in our layoffs, home sales, school enrollments, and traffic on Brookwood.
What Bagger’s pointing out here is important: the failure wasn’t that Hillsboro’s engineers couldn’t get 18A or High-NA working. They did. The failure was in how Intel sized and timed its older Intel 7 and DUV capacity, just as demand for Granite and Sapphire Rapids took off.
If Intel fixes that — and starts listening more to the people who see the whole field, not just the next quarter — Ronler Acres only becomes more important. Every new node, every clever mix of old and new tools, every custom AI chip deal starts its life here before it ever ships from Chandler, Ireland, or Ohio. All roads to an Intel-winning future begin here on our home field. We have the players and the equipment take the company all the way and get the W. All we need now is for a GM and a Head Coach to take us all to the promised land.
If Team Blue doesn’t right its ship, we won’t lose Intel overnight. But we may lose the version of Intel that made Hillsboro matter to the rest of the world.
Bagger’s Corner: Hard Questions, No Easy Villain
At this point, one has to ask:
- Is part of this problem that Intel fired the people who had a deep understanding of the internal mechanisms needed to avoid it?
- Who is to blame for not understanding market demand?
- Who wrong-sized the capacity on a decade-old technology that suddenly matters again?
- Who cut capex in Q3 2025 by more than half at the exact moment they needed it?
- And why does the market have to wait until the second half of this year to have a real conversation with management to understand what happened?
All good questions from my old friend Bagger.
A good set of concluding questions. If you think on those, its clear realization that the top leadership remains incompetent to navigate Intel. Translating to poor execution and inability to deliver even to a seasonally expected demand. Whereas industry peers and competitors are growing their business. More cuts are needed to save Intel, but the right ones aren’t in the ranks – LBT needs to clean up the attic, and focus the priorities.
You are one of many we have heard from – this same opinion is prevalent. What happens when the pirates are running the company? The company ship sinks- this is the not the Intel I have ever known or seen. Seems like the Titanic that is just taking 2 deacades to go under. On behalf of all the greta employees of Team Blue I sure hope someone wakes up and that the company returns to greatness. History is against that happening but I am still cheering for that outcome.
Dirk – I’m curious about your cheering for the outcome of Intel’s return to greatness. I’m all for it, but I’ve been following your positions on data centers as well as the battery facility proposed on Cornelius Pass and some of the positions here seem incongruous. Your opposition to the battery project in particular seems to be rooted in arguments around land use, industrial noise, and fire/environmental risk.
Intel’s fabs in Hillsboro present all of these factors to a much greater degree than a giant battery pack. Fabs have incredibly potent hazardous chemicals, major ongoing traffic, massive electrical power use, water consumption, and have plenty of emissions. Those emissions are controlled and monitored, the water use is blunted by on site recycling, the risks of the hazardous chemicals are controlled by a multitude of well-engineered systems, and in spite of a couple of DEQ issue it’s my belief that the fabs are generally good neighbors. But so long as you’re making arguments about risks of catastrophic battery fires, the emissions from a major fab disaster have the potential to be far, far worse than a big pile of burning lithium batteries. Additionally, while the decision is now well behind us, all of the Intel facilities are sited on former farmland – one needs only to look at the campus names to see that.
Is the opposition to the battery site and data centers really due to environmental risk, land use, etc or is it rooted in lower job creation rates and some ideological opposition to the idea that capital allocators stand to benefit? On the point of capital – undoubtedly every Intel project is aimed at returning profits to investors, even if they’ve had a negative run in recent years (I’ll note that this is the type of risk every capital project takes on). Look at the 2024 Apollo capital deal giving APO a 49% equity interest in F34, for example.
I’m hoping these come across as good faith questions – it just seems to me that there are some inconsistencies in your positions and I hope you can clarify.
My own views are that the economic growth from all of these projects – from fabs to data centers to industrial batteries – stand to benefit the residents of the area, as the fabs certainly have for decades. I’m constantly alarmed at the narrative of opposition to any and all business growth from Oregonians. I love living here and I hate to see business chased out of the state by NIMBYism and anti-technology FUD. Holding tech projects and corporations accountable to environmental standards, proper labor practices, and other matters of compliance is certainly necessary and correct, but not to the extent that we put so much friction into the system that progress and growth becomes completely impossible.
Check out PDX-TIE.ORG for more information. There is no coincidence in the story of Intel’s fall from grace. The reality is that the fish always begin rotting from the head, down. Back in 2006 I saw how bad management brought Intel to to bury their $300,000,000 cell phone business and the board of directors seem to be too stoned on cannabis, which led to missing the “smart phone” market. Back in 2014 my technical leader colleagues and I already saw how (then) CEO Krzanich used the board of directors ignorance and incompetence to lead Intel into the abyss, by using PR stunts to cover for inability to produce 14nm chips at scale. Any attempts of the technical staff to change the working procedures to match reality were quashed immediately by the top executives, particularly those who were in charge of the FABs (TMG). Then, by laying off the most experienced (and therefore older) employees by the thousands Intel executives demolished what remained of the productive work culture that was established by the late CEO Andy Grove. This essentially left the company with lower-cost inexperienced (younger) employees who were operating without mentors. Ever since the 2015/2016 massive layoffs, Intel continued to survive through mostly PR campaigns and taxpayers dollars, corporate socialism (as in the CHIPS ACT), denial-ism in the investment community and as of late, tens of billion dollars direct subsidy from the US Gov. This, yet another attempt to rescue a a dying dinosaur like intel can best be described as “Jurassic Pork”. Instead of diversifying its economy, the State of Oregon is marching back to the Mesozoic era and no one dares calling “The King is Naked”, in the case of Intel.
Intel needed “better and fewer” employees, but they fired all the better ones 10 years ago, so all they have left are the fewer. It’s the brain drain that has done Intel in.
The reason they cut capex spending in that quarter in 2025 was so that they could figure out a plan and stop just spending money without real goals. We literally took out tools that they had put in and hadn’t ran. The idea was to stop figure out a plan moving forward and then go ahead and execute on it.