A look at Intel's chip-making process at its Hillsboro, Oregon and Chandler, Arizona campuses and the infrastructure behind its two planned $10B Arizona fabs
As the global chip shortage continues, we take an inside look at how semiconductors are fabricated. Tweets: @sub8u , @jimpethokoukis , @dylan522p , @carlmalamud , and @puiwingtam Tweets: Subrahmanyam Kvj / @sub8u : Fab photo feature on semiconductor fabs. Highlighting why it is so complex, and expensive, to quickly fix the global chip shortage. https://www.nytimes.com/... https://twitter.com/... James Pethokoukis / @jimpethokoukis : Don't take our modern world for granted. A microchip is a marvel. Great piece here. https://www.nytimes.com/... Dylan Patel / @dylan522p : Intel Ireland now has their first EUV tool What's odd is that it's not from the Netherlands, but instead it was shipped from Oregon to Netherlands to Ireland. Why not straight from Netherlands? Did they transfer an earlier tool to get a massive upgrade? https://www.youtube.com/... Carl Malamud / @carlmalamud : Fascinating article on the new semiconductor fabs being built by Intel. https://www.nytimes.com/... Pui-Wing Tam / @puiwingtam : An up-close look at why the world can't suddenly recover from a shortage of microchips. Stunning video, photo and text from @alanacelii @donal888 @philip_cheung https://www.nytimes.com/...
Context & Ripple Effects
This Times feature walks through Intel's Hillsboro and Chandler campuses while the company is planning two $10B fabs in Arizona, and it lands mid-arc: Protocol had already gone inside the Ronler complex where Intel does its most advanced research, and the Washington Post had toured a GlobalFoundries plant filled with $10B of machinery running a three-month, 700-step process. Together these pieces make the same point from different angles — a fab is one of the most expensive, slowest-to-build objects in industry.
That framing matters because the coverage around it brackets the cycle: months after this piece ran, analysts were warning of an inventory correction that could become the worst downturn in a decade, and Nikkei was dissecting how efforts to replicate chipmaking within single countries keep exposing bottlenecks rather than removing them.
First-order effects
- Intel's two planned $10B Chandler fabs commit the company to multi-year construction spending even as the shortage narrative begins to crack — capacity arrives on a fixed schedule while demand does not.
- Readers get a rare ground-level view of why 'just build more fabs' fails: the machinery, cleanrooms, and process steps described here are exactly what makes rapid capacity response impossible.
Second-order effects
- Rivals like GlobalFoundries face the same capital wall — their own $10B-scale plants show that no competitor can out-build the shortage quickly either, keeping pricing power with whoever has existing capacity.
- If the downturn hits before the Arizona fabs are productive, Intel carries idle or underutilized new capacity into a weak market, pressuring returns on precisely the investment meant to fix the shortage.
Third-order effects
- The pattern holds over time: by 2025 Intel's Arizona build-out had grown into the $32B Fab 52 starting large-scale 18A production — far beyond the $10B-per-fab plan described here — suggesting per-site costs only escalate as process nodes advance.
- Combined with the resilience critique, this points toward regionalization producing fewer, larger, more concentrated mega-fabs rather than distributed redundancy — the opposite of what supply-chain hawks intend.
The trend: Chip manufacturing is consolidating into ever-larger, ever-costlier fab sites, where each generation of process technology raises the capital floor faster than policy-driven capacity pushes can spread it.