Taiwan: the air activity is down and the risk is not

What is observable

On AEI's tracking, the PLA conducted 190 aerial incursions into Taiwan's ADIZ in July 2026 — less than half the total for July 2025 and July 2024, with the monthly average back at its pre-2024 baseline since January. The Taiwan Security Monitor recorded the same direction earlier in the year: 270 sorties in January, 190 in February, 173 in March, a sixty per cent fall against March 2025.

Naval presence has not followed the aircraft down. The same monitoring put PLA vessels at an average of 7.2 a day in March, inside the five-to-nine band held through 2025. Sustained ships with fewer aircraft is not a reduction in posture; it is a change in which instrument is forward.

The clearer development is on the law-enforcement side. Taiwan reported 55 sightings of Chinese coastguard and research vessels in June 2026 against 30 in May, and in June and July Beijing ran what it described as special maritime law-enforcement operations in waters east of Taiwan — outside the strait, which is new — with commercial vessels asked to give origin and destination. The UK, Germany and France objected jointly in late June.

Why the falling number is the less informative one

Aircraft sorties are a good measure of military signalling and a poor measure of commercial risk. They are visible, countable, widely reported, and they do not touch a ship. Radio challenges, document requests and inspections do, and they generate almost no headline volume.

There is also a structural asymmetry in escalation. A large air exercise is unambiguous, expensive and hard to sustain. A law-enforcement regime is deniable, cheap, incremental and — the point — normalising. Each step establishes the precedent for the next, and the ratchet only runs one way. Nothing in the observable record says a quarantine is planned. What the record does say is that the components of one are being practised separately, which is a different and more useful statement than a probability.

For a company, the operational distinction is that a conflict branch is a discontinuity you cannot trade through, while an interference branch is a slow tax on lead times, insurance and contractual reliability that most groups have no line item for.

The concentration behind the strait

The exposure that makes this a board matter rather than a regional one is concentration. On CSIS figures for 2024, over a fifth of world maritime commerce passed through the Taiwan Strait, including some $1.3 trillion of Chinese trade, around $413 billion of Japan's, roughly $300 billion of South Korea's and about $102 billion of Australia's exports. Taiwan produces over ninety per cent of the world's most advanced chips.

Concentration of that kind means the usual mitigations do not behave as expected. Dual-sourcing a component whose only advanced-node capacity sits on one island is not dual-sourcing. Rerouting shipping is possible; rerouting a foundry is a multi-year capital programme. And unlike a chokepoint that can be sailed around at a cost in days, the substitution problem here is measured in years, which is why it belongs in capex sequencing rather than in a logistics contingency.

The practical consequence is that the right question is not how likely a conflict is. It is which of your revenue, cost base and committed investment sits behind a single strait, how long the alternatives would take to stand up, and what you would want to have already done by the time the answer matters.

What we do not claim

We are not forecasting a blockade, a quarantine or an invasion, and no probability in this note attaches to one. The assessment is narrower and, we think, more useful: the composition of pressure has changed, the most-reported indicator is now the least informative one, and a register keyed to sortie counts will read the current period as improving.

Confidence is moderate. The observable record is good on activity counts and thin on intent, and we would rather say that than average the two into a score.