Sanjay Mehrotra marked America's 250th anniversary by attaching the same number to Micron Technology's industrial ambition. On 9 July, the company increased planned US fab and technology investment to more than $250 billion through 2035 and poured the first concrete at its Clay, New York site. The pour came more than a quarter ahead of the original schedule. The larger goal is more demanding: Micron wants eventually to produce 40 per cent of its DRAM in the United States.
That promise changes the scale of Mehrotra's job. He is no longer allocating capital among a familiar set of Asian plants and a research base in Idaho. He is trying to create a new manufacturing geography while several Asian sites expand at the same time. New York, Idaho and Virginia must become credible high-volume production centres without disrupting technology transfers from Japan and Taiwan or packaging ramps in Singapore and India.
The programme is often described as reshoring. Operationally, it is closer to synchronisation. A memory process developed in one country must reach competitive yield in another; wafers must meet packaging capacity at the right moment; suppliers must build near new fabs; and thousands of employees must be trained before equipment begins producing saleable output. Government support can lower the bill. It cannot make those hand-offs work.
The US timetable has become measurable
New York supplies the most visible milestone. Less than six months after breaking ground, Micron had completed early site work and moved into vertical construction. Bechtel is the engineering, procurement and construction partner for the first fab, Jacobs is handling architectural and engineering design, and Gilbane has led preconstruction and site infrastructure. The planned campus can contain as many as four fabs and is expected to support 9,000 direct Micron jobs and 50,000 jobs overall.
Local execution is already part of the political bargain. Micron said about $675 million had been directed through the early phase, with more than half of awarded value going to New York contractors and suppliers. More than 80 per cent of workers on site had been state residents. Those figures create accountability before a single wafer is produced. Delays, cost inflation or a weak local supplier base will be assessed not only by shareholders but by governments and communities that organised housing, transport and training around the project.
Idaho is closer to production. Micron expects first wafer output from its first Boise fab in mid-2027 and from the second in late 2028. Co-locating manufacturing with research and development should shorten the route from a process breakthrough to volume output. It also raises the execution standard: if a node is late in the company's home technology centre, there is no geographic excuse.
Virginia plays a different role. The Manassas plant began manufacturing 1-alpha DRAM in May for long-lifecycle applications including vehicles, medical devices, industrial equipment, networking, defence and aerospace. Micron expects qualified production by the end of 2026 and says the expansion will quadruple the site's DDR4 wafer supply. The decision recognises that resilience is not confined to the newest AI product. Older memory generations can remain critical for years after leading-edge factories move on.
A fab without suppliers is an expensive shell
Mehrotra has begun financing the ecosystem around the new plants. Alongside the July construction announcement, Micron committed up to $3 billion for US semiconductor-supply investments. The first disclosed transaction is $500 million of financing support for GlobalWafers' 300-millimetre raw-silicon facility in Sherman, Texas, paired with a ten-year supply agreement. The proposed deal remains subject to definitive documentation and approvals.
This is a notable extension of capital allocation. Micron is using its balance sheet not only for cleanrooms and tools but to make sure a strategic input exists domestically at useful scale. The arrangement can improve planning and supply assurance. It can also concentrate exposure if a subsidised supplier is slower or costlier than an established Asian alternative. Long duration helps a wafer producer justify investment, but it leaves Micron responsible for choosing the right partner and technology many years ahead.
The company has more capacity to carry that burden than at any earlier point. Its fiscal third quarter produced $41.46 billion of revenue, $28.24 billion of GAAP net income and $18.3 billion of adjusted free cash flow. Net capital expenditure for fiscal 2026 is expected near $27 billion, with roughly $10 billion in the August quarter alone. These cash flows can accelerate construction, yet the useful financial measure is not the amount spent. It is the amount of qualified output each dollar ultimately creates.
Asia remains the launch engine
The most important counterweight to the American narrative arrived in Hiroshima five days before the New York pour. Micron broke ground on a new cleanroom at its Japanese site, with a first phase of about 300,000 square feet and equipment installation planned from late 2028. The project is supported by up to 536 billion yen from Japan's Ministry of Economy, Trade and Industry. Hiroshima has helped move Micron from 1-beta DRAM into the 1-gamma generation using extreme-ultraviolet lithography and will contribute to future advanced DRAM and high-bandwidth memory.
Taiwan offers the speed of an established semiconductor cluster. In March, Micron completed the purchase of Powerchip Semiconductor Manufacturing Corporation's Tongluo P5 site, about 15 miles from its Taichung campus. The acquired facility contains roughly 300,000 square feet of existing 300-millimetre cleanroom space. Retrofit work began immediately, meaningful shipments are targeted for fiscal 2028, and construction of a second facility with about 270,000 square feet is due to start by the end of fiscal 2026.
Singapore is carrying two different transitions. A new high-bandwidth-memory packaging facility is expected to contribute materially to supply during 2027. In January, Micron also broke ground on a double-storey wafer fab within its NAND complex, a planned investment of about $24 billion over ten years that can provide 700,000 square feet of cleanroom space. Initial wafer output is scheduled for the second half of 2028. Co-locating NAND production and HBM packaging can share infrastructure and talent, but the two operations require different process disciplines.
India completes the chain rather than duplicating a wafer fab. Micron opened its assembly and test facility in Sanand, Gujarat, in February and began commercial shipments during 2026. The project represents about $2.75 billion of combined investment by Micron and government partners. It gives India a position in a semiconductor activity that must meet tight quality, traceability and delivery standards before a wider local ecosystem has matured.
Together, these projects show why producing 40 per cent of DRAM in America will not make Micron an exclusively American production system. Japan and Taiwan remain central to advanced process development and rapid capacity conversion. Singapore links wafer fabrication with sophisticated packaging. India expands assembly and testing. The United States is adding scale, research proximity and strategic redundancy. Mehrotra needs the sites to specialise without becoming silos.
Technology transfer is the hidden balance sheet
Construction photographs make progress easy to see. Yield transfer is less visible and more valuable. A cleanroom becomes economically productive only after equipment is installed, recipes are tuned, defects fall and customers qualify the output. Leading-edge DRAM requires exact control across hundreds of steps. HBM adds stacking, thermal and packaging constraints. A process that performs well in Hiroshima or Taichung must be reproduced by a newly assembled team in Boise or New York.
The timing magnifies the risk. Idaho output is due in 2027, while new capacity in Taiwan, Singapore and later Japan clusters around 2028. Equipment, specialist contractors and experienced engineers will be demanded by several sites simultaneously. Micron must decide where scarce EUV systems, packaging tools and process leaders produce the highest return. Moving too many experienced people can weaken the plants expected to train the new ones.
Workforce promises also extend beyond recruitment. New York needs craft labour during construction and thousands of technicians and engineers for operations. Idaho must expand around an already competitive technology market. Virginia needs expertise in long-lifecycle qualification. Asian sites compete for many of the same engineering skills. Micron has committed more than $325 million to workforce and community development across the US projects, but training must arrive before production, not after vacancies become constraints.
Government money creates several scorecards
Micron's network is being built through public-private partnerships in every major region. US CHIPS Act awards support Idaho, New York and Virginia. New York has a $500 million community-investment framework. Japan is subsidising advanced memory in Hiroshima. India's central and Gujarat governments contribute heavily to Sanand. Singapore provides an established industrial platform and talent pipeline, while Taiwan's authorities helped accelerate the Tongluo acquisition and ramp.
These partnerships reduce private capital requirements and reflect the strategic status of memory. They also multiply obligations. Governments care about local employment, domestic sourcing, technology leadership and schedule. Customers care about qualification and continuity. Investors care about returns. A project can satisfy one group while disappointing another. Spending locally may increase resilience but raise unit cost; slowing a fab may protect returns but break a public commitment.
Mehrotra therefore needs a manufacturing dashboard more exacting than capital expenditure. The relevant measures are construction against schedule, tool-installation readiness, yield at equivalent process stages, time to customer qualification, output per cleanroom area and supplier concentration. Local hiring matters, but so does whether trained teams can sustain quality without permanent transfers from Asia.
The July concrete pour proved that Micron can move a landmark American project faster than its first timetable. It did not prove that the output will reach market with competitive cost and yield. That evidence will arrive in sequence: qualified Virginia products by the end of 2026, Idaho wafers in 2027, Singapore packaging and then multiple 2028 ramps across Idaho, Taiwan and Singapore.
Mehrotra has put Micron at the centre of industrial policy on both sides of the Pacific. His achievement will not be measured by which country appears to win a factory announcement. It will be measured by whether an American expansion and an Asian technology engine operate as one production system. More than $250 billion buys concrete, equipment and political attention. Only coordinated execution turns them into memory.
