A datacentre or chip fab in the US, China, or Taiwan is hit by a strike

Where things stand (14 Oct 2026)

Nothing qualifies yet. Since the question opened on 16 Sep 2026, there has been no armed strike on the US, mainland China or Taiwan. Recent activity is pressure short of war. The PLA is running unusually wide live-fire drills off its own coast (12–20 Oct), and its coast guard and militia keep probing around Pratas and east of Taiwan. Taiwan's defence ministry is only reporting what it observes (HKFP/AFP, 12 Oct; AEI/ISW, 6 Oct). US–China diplomacy is active: Xi visited Washington on 23–25 Sep, the trade truce was extended, and the two leaders are due to meet again at APEC in Shenzhen in November and the G20 in Miami in December (Atlantic Council). Reuters reported on 1 Oct that US officials think an invasion before 2028 is "increasingly unlikely". They cite Xi's purges of the PLA and the January 2028 Taiwan election, and one official said 2028 could become "pretty sporty" if the DPP wins again (summary via Ground News). The top commander Zhang Youxia and the chief of the Joint Staff, Liu Zhenli, were expelled on 21 Sep. That leaves China's top military body, the Central Military Commission, with two active members (Modern Diplomacy, 30 Sep).

Reference class and outside views

In practice, YES needs a war: a Chinese strike campaign on Taiwan, possibly widening into US–China strikes on Guam or the Chinese mainland. China and Taiwan have not exchanged fire on land since the Kinmen shelling ended in 1979. Outside views:

• Metaculus (question 11480, community forecast scraped 12–13 Oct 2026): full-scale invasion before 2028 5%, before 2030 15%, before 2035 26%. A September 2025 archive snapshot showed 15%, 25% and 38%, so the community has revised down sharply. Interpolating gives about 19% by the end of 2031.
• Good Judgment superforecasters (27 Jul 2026): 88% that neither side accuses the other of using a weapon before July 2028 (source).
• Forecasting Research Institute (2024): median expert 25%, median superforecaster 19% for an invasion before 2030 (source). This predates the purges and the détente.
• Polymarket (public API, 14 Oct): invasion by the end of 2026 about 2%, by mid-2027 7.5%, by the end of 2027 10.5%; China–Taiwan military clash before 2027 about 4.7%; US–China clash 2.6%. Polymarket's own history shows these prices run high: the 2025 invasion market opened at 12.5% (price history from the exchange's data feed) and resolved No.

I put somewhat less weight on the long-dated Metaculus numbers, which have been revised down sharply. My blend is about 11–12% for a full-scale invasion by the end of 2031, about 3% for a blockade with strikes on land, and about 2.5% for limited strikes.

Chance of an official confirmation, given a war

Data centres are now explicit wartime targets. In March 2026, Iranian drones hit AWS sites in the UAE and Bahrain (Reuters). Russia and Ukraine are now striking each other's data centres, and Ukraine's foreign minister publicly confirmed that "Russia struck major Ukrainian data centers" (UNN, Sep 2026). That is the kind of government statement this question needs.

A full Chinese campaign would aim to paralyse Taiwan's command and communications networks, and Taiwan's fabs sit near its air bases. Taipei would have strong reasons to publicise any damage. My estimates of an official confirmation are:

• about 0.75–0.8 given an invasion;
• about 0.4 given a blockade with strikes;
• about 0.12 given limited strikes.

US or Chinese confirmation (Guam, the mainland) adds a little.

Other routes add about 0.5–0.7% in total:

• an attack on US soil by Iran's Revolutionary Guard. Iran threatens data centres "in the region" (Fortune/AP, 11 Oct), and Britain says the Guard was "almost certainly" behind the RAF Fairford drone plot;
• a US strike on Chinese AI data centres, an idea floated by the think tank CNAS (Global Times);
• a US–China clash in the South China Sea;
• North Korea.

Overall: 0.115×0.78 + 0.03×0.4 + 0.025×0.12 + 0.006 ≈ 11%.

How the risk is spread over time

I computed quarterly hazards in Python:

• Late 2026: 0.18% (summit calendar, no signs of mobilisation).
• 2027: 0.40–0.45% per quarter (purges, US assessment).
• 2028–2029: 0.5–0.6% per quarter (Taiwan election, inauguration and the US election). Kalshi has Lai at roughly 65–70% to win in 2028.
• 2030–2031: 0.58% per quarter.

Cumulative: 1.9% by the end of 2027, 6.4% by the end of 2029, 10.7% by the end of 2031.

Strongest case against

Higher: taken at face value, Metaculus implies roughly 16–17% for this question. A DPP win in 2028, US interceptor stocks drained by the Iran war, and a transactional US stance could push the risk higher. Lower: with superforecaster-style war odds (about 7–9% by 2031), and allowing that wartime secrecy might keep Taipei from naming a specific "data centre" or "fab", the answer falls to about 6–7%. I sit between the two.

Question. By the last day of each quarter, has the government of the United States, mainland China, or Taiwan confirmed that a datacentre or chip fabrication plant on its territory was physically damaged by a military strike?

Definitions.

• Territory: for the United States, the 50 states, District of Columbia, and US territories, including Guam and Puerto Rico, but not overseas bases on foreign territory; for mainland China, PRC-administered territory excluding Hong Kong, Macao, and Taiwan-administered territory; for Taiwan, the main island and outlying islands administered by Taipei. Geographic coverage is fixed to administration on the opening date; subsequent occupation does not change the applicable confirming government.
• Datacentre: a purpose-built or dedicated facility housing computer servers for computing, storage, or network services, including enterprise, cloud, colocation, government, and military facilities. Ordinary office computers and telecommunications equipment without a server facility do not qualify.
• Chip fabrication plant: a facility that fabricates semiconductor devices on wafers. Packaging-only, assembly-only, testing-only, equipment-manufacturing, and chip-design facilities do not qualify. No restriction applies to chip generation, operator, or AI use.
• Facility status: operating facilities, temporarily idle facilities, and facilities under physical construction count; proposed sites and permanently decommissioned facilities do not.
• Military strike: a deliberate kinetic attack conducted by a state's armed forces using missiles, bombs, drones, artillery, gunfire, or other physical weapons. A declaration of war, foreign attacker, deliberate targeting of the facility, or AI-related motive is not required. The confirming government must identify the incident as an armed-forces attack, but need not identify the attacking state.
• Damaged: actual physical harm to a qualifying facility's buildings, installed equipment, or on-site supporting infrastructure. Any confirmed amount counts. Direct hits, blast effects, debris, and fires caused by the strike count, including collateral damage from an attack on another target.

Resolves YES if an eligible government source affirmatively confirms the facility's type and location, physical damage, and causal connection to a military strike. These facts may be established across multiple official publications; the qualifying publication date is the date the last necessary fact is confirmed. A named operator or exact street address is unnecessary if the government unambiguously identifies at least one qualifying facility in the covered territory.

Exclusions (resolve NO).

• Threats, targeting announcements, attempted attacks, or statements that damage is merely possible, suspected, or under investigation.
• Service interruption, loss of data, production stoppage, or off-site power or communications damage without physical damage at the facility itself.
• Cyberattacks, covert sabotage, terrorism or other non-state attacks not confirmed as an operation of state armed forces, accidents, exercises, and weapons tests.
• Damage to foreign facilities owned by US, Chinese, or Taiwanese companies, unless the facility is physically within the covered territory.
• Confirmation only by the attacker, facility operator, media, or another government, without confirmation by the applicable government below.

Resolving source. Official statements, reports, or published briefing transcripts of the applicable national government: the US federal government for US territory; the PRC central government for mainland China; and the Republic of China government in Taipei for Taiwan-administered territory. Eligible publishers include their executive offices, defense ministries or departments, armed forces, foreign ministries or departments, and other national ministries or agencies. Examples include Taiwan's Ministry of National Defense and China's Ministry of National Defense. A government-hosted academic commentary or state-media report is not an official confirmation unless it reproduces an attributable official statement. This is a narrower government-only source rule than Public confirmation (A2); developer or evaluator confirmation cannot substitute. Apply Resolution date (A3) and the Resolution window. The incident may predate the opening date, but the qualifying official publication must fall within the resolution window.