Several years after the EU Chips Act set out goals of strengthening supply chains and reducing strategic dependence, Intel's new-fab plans in Germany and Poland—once held up as symbols of that effort—were shelved one after another in 2025. Intel itself summed it up by saying it had "invested too much ahead of demand," a statement that seemed to confirm the company's retreat from Europe. Yet on July 13, 2026, that same Intel announced an additional €5 billion ($5.7 billion) investment at its Leixlip campus in Ireland. What's more, it was only three months earlier that Intel bought back its equity stake in that very same Fab 34 for $14.2 billion. Behind the simultaneous unfolding of a European retreat and a massive investment—by the same company, on the same continent—lies a logic of selection and concentration: capital lost in Germany and Poland is being redirected to a single site in Ireland.
$20 Billion Flowing Into Fab 34 in Just Three Months
The investment Intel announced on July 13, 2026, will go toward expanding overall production capacity across the entire Leixlip campus in Ireland. At Fab 34, the campus's core facility, the investment covers manufacturing of Intel Xeon 6 and next-generation Intel Xeon processors, both using the Intel 3 node. The bulk of the spending will go toward upgrading existing equipment, installing the latest manufacturing tools, and expanding the automated material handling systems that connect multiple modules across the campus—the company says no new fab construction is involved. The program has already been underway since the first half of 2026, with more than half of the total investment planned to be spent by the end of 2027.
On April 8, 2026, Intel bought back the 49% stake in Fab 34 from Apollo Global Management for $14.2 billion, restoring full 100% ownership. That stake had been sold to Apollo for $11.2 billion in June 2024, meaning the buyback price came in roughly $3 billion higher than the sale price. The $14.2 billion buyback was a single lump-sum capital transaction completed on April 8, whereas the new $5.7 billion investment is a multi-year plan with more than half to be spent by the end of 2027—the two are different in nature. Even so, the fact remains that two massive capital commitments to the single Leixlip site were decided just 96 days apart. Combined, the two amount to roughly $19.9 billion—about ¥3.22 trillion—directed at a single site within a remarkably short span.
The stake in Fab 34 thus changed hands between Intel and Apollo twice within just two years. Intel sold 49% to Apollo in June 2024, bought it back for $14.2 billion in April 2026 to restore full ownership, and then added another $5.7 billion just three months later. Intel CFO David Zinsner has reportedly explained the buyback as a move to improve the company's financial position. The shift from temporarily relying on outside capital to bearing the full cost alone has played out within just a few years.
Over that same period, the capital that was supposed to go into Germany and Poland dropped to zero. Both the Magdeburg, Germany plan (ultimately sized at €30 billion, with €9.9 billion in subsidies) and the Wrocław, Poland plan (up to $4.6 billion according to Intel's own announcement) were completely withdrawn in 2025. On the same continent, the same company is simultaneously piling on massive investment in one place while erasing entire plans in others.
Why Leixlip Alone Was Chosen: The Shortest Path Is an Already-Operating Fab
The reason Leixlip was chosen is simple: it's a fab that's already up and running. Both Magdeburg and Wrocław were greenfield projects built from scratch, involving years of time and risk—from cleanroom startup through to improving mass-production yields—before reaching full operation. Leixlip, by contrast, is a site where Intel has invested more than €30 billion since first establishing operations there in 1989. It already employs 4,900 people and has a proven track record of mass production on the Intel 3 node. Within Europe, it was the only place where Intel could add supply capacity without taking on the risk of a new build.
Building a new semiconductor fab requires not just constructing the building itself, but also bringing in and calibrating equipment, running trial production, and validating yields before mass production can begin. This ramp-up period often spans two to three years, and if demand forecasts miss the mark, the investment itself can become a stranded asset. This is exactly what Intel faced in Germany and Poland, and CEO Tan summed it up: "We invested too much ahead of demand, and as a result our factory network became needlessly fragmented, and utilization rates fell." The fact that this new investment is limited to upgrading equipment at an existing fab can be read as a way of avoiding that same ramp-up risk.
Naga Chandrasekaran, EVP of Intel Foundry, said in the official announcement: "This €5 billion investment represents our firm commitment to maximizing production capacity at the Leixlip campus and increasing the volume of supply to Intel Foundry customers." Irish Taoiseach Martin also commented that "this multi-billion-euro investment by Intel is a strong vote of confidence in Ireland, in our skills base, and in our position at the center of Europe's advanced manufacturing ecosystem." IDA Ireland CEO Michael Lohan likewise welcomed the news, calling Intel "one of Ireland's longest-standing and most strategically important investors." New employment will be limited to "hundreds of full-time roles" and "thousands of construction-related jobs," with Intel itself not disclosing specific figures.
An Announcement That Speaks of "Contributing to European Technological Sovereignty" but Says Nothing About the Germany and Poland Withdrawals
Intel's official announcement goes so far as to describe this investment as "a contribution to the European Union's technological sovereignty." But the European reality behind those words is nowhere to be found in the announcement itself. There is no mention whatsoever of the withdrawals from Magdeburg and Poland, nor of the equity buyback from Apollo. Behind an investment touted as a contribution to technological sovereignty, the fact that two new-fab plans on the same continent have simply vanished goes unmentioned.
The scale of the plans that were actually scrapped is not small. The Magdeburg project was first announced in 2022 at €17 billion, expanded to €30 billion in 2023, with the German government prepared to provide €9.9 billion in subsidies. The Wrocław, Poland project was also expected to create around 2,000 jobs, but was first delayed by two years in 2024 before being formally canceled in July 2025. Around the same time, Intel also announced company-wide layoffs of 24,000 employees.
In August 2025, the U.S. government converted CHIPS Act-related subsidies and Secure Enclave funding into equity, acquiring a 9.9% stake in Intel worth $8.9 billion at the time; following the 2026 stock price rally, that stake's value reached roughly $36 billion. Meanwhile, the new fab in Ohio (Ohio One) has slipped by several years from its original schedule: Mod 1 construction is now expected to complete in 2030 with operations starting in 2030–2031, and Mod 2 construction is expected to complete in 2031 with operations starting in 2032. None of this—the capital concentration at Leixlip, the U.S. government's equity stake, or the delays in Ohio—appears anywhere in this press release.
The winners from this whole reallocation are the Irish government, IDA Ireland, and Apollo, which pocketed a $3 billion premium on the buyback. The losers are the German state of Saxony-Anhalt, which lost roughly 3,000 jobs and €9.9 billion in subsidy allocation; Poland, which lost its planned 2,000 jobs; and the EU's chip policy itself, built on the goals of "strengthening supply chains and reducing strategic dependence." The European Commission is reported to have adopted its Chips Act 2.0 proposal in June 2026, even as Europe's largest chip investor is, at the very same moment, concentrating its capital into a single country and a single fab.
Compared with TSMC's Dresden Fab and Japan's Rapidus, the Difference in Strategy Becomes Clear
There is another major fab investment currently underway in Europe: ESMC, which TSMC is building in Dresden, Germany. The investment exceeds €10 billion, with the German government subsidizing roughly €5 billion of it. The project is expected to move from structural construction to equipment installation in the second half of 2026, aiming for mass production to begin in the second half of 2027 (though some customer companies cite 2028, so the timeline varies somewhat). However, the target nodes are the mature 28/22nm and 16/12nm processes, aimed at automotive, industrial equipment, and IoT applications—not advanced AI logic.
In other words, the two major investments currently underway in Europe are not the kind of ventures that break new ground in leading-edge logic. TSMC chose mature nodes for automotive and industrial applications where demand is easier to forecast, while Intel chose to expand a proven, already-operating fab rather than take on a new build. The EU's vision of "securing advanced logic production capacity within the region" is not being realized by either investment.
This contrast in capital-allocation thinking can also be applied to Japan's Rapidus. Rapidus is investing roughly ¥4 trillion (about $24.7 billion at current exchange rates) on the 2nm generation alone, aiming to begin mass production in the second half of fiscal 2027. This is a bet on launching an entirely new, not-yet-operational production line all at once on the unproven, leading-edge 2nm node. The roughly $19.9 billion committed to Leixlip within a 96-day span is smaller in scale, but the character of the bet is entirely opposite.
What Intel chose was a steady strategy: expanding production of Intel 3—an advanced node, though one generation behind the cutting edge—at an existing site with a proven mass-production track record. What Rapidus chose was a strategy of tackling an unproven node at a site with no track record. Even though both are described as "massive investments," the risk profiles are opposite. At current exchange rates (roughly ¥185 to the euro and ¥162 to the dollar, as of July 2026), the €5 billion comes to about ¥925 billion and the $5.7 billion comes to about ¥923 billion—converting these figures into yen makes the scale easier to grasp intuitively.
The Focus Shifts to Winning External Customers for Intel Foundry; Execution by End of 2027 Will Be the Test
How Intel itself frames this investment can also be read from the scale of the capex involved. The Irish Times reports that the $5.7 billion represents roughly 30% of Intel's annual capital expenditure budget. The fact that a single fab—and one that involves no new construction, only an upgrade—accounts for nearly 30% of the annual investment budget underscores just how concentrated this capital commitment to Leixlip really is.
In remarks to Irish media, Chandrasekaran reportedly acknowledged the risk of an AI bubble as well, saying that "the AI narrative is strongly established, but there's also the possibility of a correction." This expansion investment, centered on Xeon 6, is premised on continued growth in demand for AI servers. It's worth noting that the very people behind the investment are themselves acknowledging the risk inherent in that premise.
The focus going forward will be on whether the expanded capacity at Leixlip ends up filled entirely by Intel's own Xeon demand, or whether it becomes available supply capacity for external foundry customers. Whether Chandrasekaran's stated goal in the announcement—"increasing the volume of supply to Intel Foundry customers"—turns out to have real substance will become clear by the end of 2027, when more than half of the investment will have been spent, based on whether the Intel 3 line has actually secured orders from external customers by then. Having given up both new-fab plans in Germany and Poland, Intel has no remaining means of demonstrating its foundry business presence in Europe other than winning external customers at Leixlip.