That extended as hyperscalers began vertically integrating into their supply chains through partnerships and multi‑quarter commitments, I mean, and management now describes a good view of base demand up to three quarters ahead, you know, expecting data‑center momentum through calendar 2026. It draws a hard line past that— as a practice we will not forecast beyond the end of the calendar year and define the ‘foreseeable future’ as the next three quarters, right?
We set a record of $71 million in high‑end IC mask revenue in Q1, up 19% YoY, then saw it dip to about $57 million in Q2 – down roughly 5% YoY and about 20% sequentially – on delayed design releases. The long‑term drivers remain, alongside temporary headwinds, as demand is inherently variable and our visibility stays limited with a typical backlog of one to three weeks. Because high‑end mask sets carry much higher prices, even a small number of orders can materially move revenue and earnings.
Our IC mask revenue fell about 5% even as leading-edge AI chip demand stayed strong — we follow design releases, not wafer starts. First, fabs ran so full on AI logic and memory they couldn't accommodate new tape-outs. Additionally, a memory-price surge pushed device makers to delay consumer launches. The final factor: geopolitical uncertainty, including the U.S.-Iran conflict, added caution. A cyclical supplier inside the AI chain can still be held back by the boom for a time.