Announcing a fab and building one are different activities, and this collection is organised around the distance between them. Commitments are recorded as commitments, milestones only when a source states they actually happened, and no date is invented to make a timeline look balanced.
The clearest current illustration is European. The EU's Digital Decade target is a 20% share of the global semiconductor market value chain by revenue by 2030. In special report 12/2025 the European Court of Auditors — the union's own supreme audit institution — concluded that the Chips Act is very unlikely to be sufficient to reach it, cited the Commission's own July 2024 forecast of just 11.7% by that date, and questioned whether the target was ever realistic given the Commission's limited mandate and resources. The Act announced a minimum of 43 billion euros in policy-driven investment expected to be matched by private money for at least 86 billion in total; the Commission controls roughly a tenth of the public share. An audit body putting a policy's likely failure in its own report title is unusual, and it is better evidence than any commentary about it.
The most recent development here is not a fab at all but the tool that will fill one. On 7 September 2026, ahead of the SPIE BACUS conference, ASML and TSMC formed an industry initiative to move from today's 6-inch EUV photomasks to 12-inch, targeting a pilot line by 2031 and full lithography readiness for advanced node production by 2033. High-NA EUV arrives first on the existing 6-inch masks. The stated payoff is higher fab productivity, lower cost, and the removal of stitching constraints — stitching being what a designer must do when a die outgrows a single exposure, which is the same physical wall the accelerator collections on this site run into from the other side.
TSMC states it intends to use High-NA in high-volume manufacturing for advanced nodes from 2030, and expects the number of layers needing it to rise, driven primarily by the increasingly complex transistor architectures required for AI applications. That is a manufacturer describing its own roadmap, which it is the right authority for, and also a causal claim about AI made by a company selling the capacity — carried here as its stated expectation, not as a finding.
The dated milestones in this collection are concentrated in the United States and Japan, because those are the jurisdictions with retrievable dated primary sources. Europe and China are covered through concepts instead. What is missing is named: EUR-Lex's Chips Act summary still returns an empty response, and TSMC's additional Arizona commitment — widely reported at around $100 billion for further 2nm fabs — was made verbally at an earnings conference and has no corresponding press release on TSMC's own newsroom, so it is not asserted here at all.