China's biggest chip packaging and test house just went back to investors for more money, again. JCET Group, the mainland's largest outsourced semiconductor assembly and test (OSAT) operator, is seeking roughly 968 million US dollars to expand advanced packaging and test capacity for AI chips 1 its third capital-spending escalation of 2026, after a July round that pushed its annual budget to about 1.4 billion dollars 2 and a first-half profit surge of 79 percent that management credited almost entirely to AI infrastructure demand 3. On paper this looks like China's chip-assembly base finally scaling to match its chip-design ambitions. Underneath, it exposes a narrower and more stubborn bottleneck: the machines that actually test those chips before they ship are still overwhelmingly built in the United States and Japan, by exactly the companies Washington's export-control regime is trying hardest to keep out of China.

The Capex Keeps Climbing, the Machines Inside Don't Change

JCET's expansion is real and it is fast. Revenue hit a record 19.5 billion yuan in the first half of 2026 on booming computing-electronics sales, with net profit up 79.4 percent year on year as AI, high-performance computing and chiplet demand pushed utilization at its plants toward capacity 3. But every stage of that expansion depends on automated test equipment, the CP (chip probe), FT (final test) and SLT (system-level test) tools that verify silicon before it leaves the factory. That equipment category has resisted China's broader localization push even as other tool categories advance: domestic suppliers now account for roughly 35 percent of the semiconductor equipment used in Chinese fabs, up from about a quarter a year earlier, led by etch and deposition tools 4. Chip testers have not kept the same pace.

A Domestic Champion, Still a Rounding Error

Hangzhou Changchuan Technology, China's leading homegrown maker of automated test equipment and handlers, is growing at a rate that would be remarkable anywhere else: 2025 revenue of about 5.3 billion yuan, up 45 percent, and a first-half 2026 profit forecast of 900 million to 1 billion yuan, up as much as 134 percent year on year 5. Scaled against the market leader, the gap is still stark. Teradyne's semiconductor-test division alone booked 1.11 billion dollars of revenue in the first quarter of 2026, roughly what Changchuan generates across four quarters, as AI-linked orders drove about 70 percent of Teradyne's total revenue that quarter 6. Advantest, the other half of the duopoly that has supplied the overwhelming majority of the world's high-end chip testers for two decades, is riding the same AI cycle. JCET's capex can multiply as many times as it wants; the tester bolted into a new packaging line is still, in practice, a Teradyne or an Advantest machine, sold under a US export license.

Licenses, Not Bans, Are the New Chokepoint

That licensing detail is the second half of the story, and it changed at the start of the year. The validated end-user exemptions that once let major foundries import US-controlled fab tools into China automatically expired at the end of 2025, replaced by annual, case-by-case export licenses that Washington now renews, or withholds, one year at a time 7. TSMC secured its 2026 authorization on the first day of January, but the shift means every chipmaker, and by extension every toolmaker selling into its fabs, is now negotiating access annually rather than assuming it indefinitely. A bipartisan group of US lawmakers pushed to go further still, writing to the Commerce and State departments in February to demand a blanket ban on selling chokepoint chipmaking tools anywhere in China, arguing that entity-by-entity restrictions had failed and that every chokepoint tool that enters China represents a permanent loss of American leverage 8.

What This Means for the Trade

None of this stops JCET from packaging chips this year. What it does is set the terms on which China's assembly-and-test boom can keep compounding. If Washington tightens ATE licensing the way the February letter urged, JCET and every other Chinese OSAT scaling for AI demand face a slower, costlier path to the tester capacity their new plants already assume they will get, since domestic substitutes at the high end barely exist in volume. If licensing instead stays merely annual and negotiable, foreign toolmakers keep a durable, high-margin foothold inside China's AI buildout precisely because no domestic alternative yet clears the bar. Either way, investors pricing Chinese OSAT and packaging names purely on capex announcements are missing the harder variable underneath: whether Changchuan and its domestic peers can close a gap that, on the current numbers, is not measured in single-digit percentage points but in multiples of annual revenue. Until that changes, the fastest-growing part of China's chip-assembly buildout remains dependent on equipment supplied, and licensed, by the country trying hardest to slow it down. Watch Changchuan's order backlog and JCET's next capex update together; a widening gap between the two would be the clearest signal yet that China's packaging boom is outrunning its test-equipment sovereignty, not closing the distance to it.