China’s next AI frontier is being forged in diamond and a quiet reshuffle of the country’s universities

On 16 June 2026, two dispatches from opposite ends of China’s technology stack pointed at the same wager. The South China Morning Post reported that researchers in China are racing to commercialise single-crystal diamond wafers as wide as a basketball — substrates that can dissipate the punishing heat generated by the most advanced AI accelerators far more efficiently than the silicon used in today’s chips. Hours earlier, an X thread compiled from official Chinese education data had circulated a more granular figure: between 2021 and 2025, Chinese universities dropped or paused roughly 12,200 degree programmes and added about 10,200 new ones, with artificial intelligence and adjacent fields absorbing much of the redirection. Read separately, each story is a curiosity. Read together, they describe an industrial-policy reorganisation with a horizon measured in decades, not quarters.

The thesis is unfashionable but worth stating plainly. The next phase of the AI race will not be decided by chatbot benchmarks or model-release calendars alone. It will hinge on who controls the physical layer — the wafers, the machines that pattern them, the cooling, the photonics, and the pipeline of engineers fluent in all of the above — and on who can retool a national university system quickly enough to feed that layer with talent. China is signalling, with state money and with administrative choices, that it intends to be that actor.

A wafer the size of a basketball

Diamond has long been a technology’s-technology material — exceptional thermal conductivity, a wide bandgap, and the ability to operate at voltages and temperatures that destroy silicon. What is new, according to South China Morning Post’s reporting, is the prospect of growing single-crystal diamond wafers at diameters approaching twelve inches. The existing global market for synthetic diamond is dominated by Chinese producers, but the high-end substrate business — the slice of the market that matters for AI accelerators, radio-frequency devices and quantum hardware — has so far been the preserve of a handful of Japanese and Western players. Closing that gap would give Chinese chip designers a domestic route to the kind of thermal headroom that next-generation AI training runs demand, and would do so without waiting for foreign equipment approvals.

The strategic logic is straightforward. The United States and its allies have spent three years tightening access to advanced lithography, extreme-ultraviolet tooling and the most capable accelerators. Each restriction has raised the premium on substrate-level innovation — on materials and packaging choices that can extract more performance from older or domestically produced nodes. A basketball-wide diamond wafer, if it can be manufactured at yield and at cost, would be exactly the kind of bet that pays off under those constraints. It would also be the kind of bet that is easier to fund inside a system where provincial governments, state laboratories and listed companies can be aligned behind a common target.

That is not to say the bet is won. Growing large-area single-crystal diamond at scale is a materials-science problem that has defeated well-funded programmes in the United States, Japan and Europe for two decades. The Chinese teams cited in the South China Morning Post report are credible, but the article does not claim a shipping product; it describes an emerging capability with a national-security read. Sceptics, including several Western analysts quoted in adjacent industry coverage, point out that diamond’s processing — cutting, polishing, doping — remains stubbornly expensive, and that thermal substrates are only one of several bottlenecks in advanced AI hardware. The honest answer is that the technology is plausible, the timetable is uncertain, and the geopolitical pressure to compress that timetable is real.

The other retooling: 12,200 programmes out, 10,200 in

The educational reorganisation, captured in the figures circulating on X, is in some ways the more consequential story. It is also the easier one to verify. Over a five-year window, China’s higher-education system removed or paused about 12,200 degree programmes and added around 10,200. The net contraction is modest; the composition is the news. Programmes in fields with weak labour-market demand or limited alignment with the country’s industrial priorities were trimmed. New programmes cluster in artificial intelligence, integrated circuits, advanced materials, robotics, biomedical engineering and the data sciences.

This is not a market response. Chinese higher-education admissions and programme approvals are steered by the Ministry of Education, by the provincial education departments, and by an explicit hierarchy of “double first-class” universities that receive disproportionate funding. When the central authorities decide that a field matters, the system can move at a pace that Western universities — bound by faculty hiring cycles, senate votes and donor preferences — cannot match. The trade-off is real and worth naming: institutional autonomy is thinner, and the system is more exposed to political fashion. The upside, by the standards of any industrial planner, is speed.

The Western wire has tended to read Chinese university reform through the lens of soft-power concerns or, more crudely, as evidence of ideological tightening. That framing captures something, but it misses the engineering of the change. Closing a tourism degree at a third-tier college and opening an AI programme at a first-tier university is, in policy terms, an act of capital allocation. Done at this scale, it is a redirection of the country’s most consequential long-cycle investment — its young people.

A counter-narrative worth taking seriously

The dominant Western line on Chinese industrial policy in semiconductors and AI is, broadly, that the United States’ export controls are working: that China can still assemble chips at mature nodes, that its most advanced foundries remain a generation or two behind, and that the gap is widening rather than narrowing. There is evidence for that view. There is also evidence for the opposing one. Diamond substrates, mature-node accelerator designs, chiplet packaging, photonics, and a deep bench of AI engineers are precisely the areas where a constrained actor can win without needing the most advanced lithography. Chinese foundries have, in the past three years, brought new capacity online at a pace that surprised most forecasters, and the country’s installed base of AI compute — measured in shipped units rather than top-end performance — is now the largest in the world.

The reasonable reading is that both things are true: the ceiling has been lowered in some technologies, and the floor of capability has been raised in others. The national-security establishment in Washington has, in fact, signalled exactly this concern in its own internal debates. The Chinese position — articulated in the pages of the South China Morning Post, in the English-language output of Xinhua and the Global Times, and in briefings by spokespeople of the Ministry of Foreign Affairs — is that technological decoupling is futile, that innovation under constraint is now a national habit, and that the United States is accelerating a process it cannot control. That claim deserves to be taken seriously even by those who reject its moral framing, because the data on Chinese research output, patent filings and STEM graduates supports parts of it.

What is at stake

If the diamond-wafer programme reaches even a fraction of its ambition, and if the university reorientation produces a generation of engineers fluent in AI hardware, materials science and photonics, the geopolitical consequences are not subtle. The United States’ lead in frontier model training depends on a relatively narrow stack: a small number of leading foundries, a smaller number of equipment makers, and a handful of cloud providers willing to write the capital expenditure. China’s strategy, as it has been articulated and as these two dispatches suggest, is to make that stack less narrow on the supply side and less decisive on the demand side — to build an alternative physical layer for AI that does not need Western permission.

The winners, if the trajectory continues, are the Chinese provincial governments and listed companies positioned in diamond growth, advanced packaging and AI education. The losers are the Western equipment vendors who currently price scarcity at a premium, and the Western policymakers who built a strategy around denying access to a handful of tools. The time horizon is not a single product cycle. It is the length of a doctoral cohort — five to ten years — and a materials-science programme that may take a decade to mature or to fail.

What remains uncertain

The sources do not settle the central question. South China Morning Post’s reporting describes a credible research effort; it does not announce a shipped product, a price point, or a yield curve. The X-circulated education figures summarise administrative changes that, in the absence of detailed ministry documentation, are best read as direction-of-travel rather than destination. It is also worth noting that the two stories come from outlets with different editorial positions and different relationships to the Chinese state — a regional paper of record in Hong Kong and a social-media thread aggregating official statistics — and the framing of national success is, in both cases, part of the work the sources are doing. What can be said with confidence is that the direction of travel is real, the money is being spent, and the people being trained will, in a decade, be the ones answering the question of whether the bet paid off.

Desk note: Monexus treats both dispatches as inputs to the same story rather than as separate beats. The framing is deliberately double-sided: the substrate story is read for what it suggests about constraints, and the university story is read for what it suggests about patience.

Wire provenance

This editorial synthesis draws on the following public wire/social posts:

• https://x.com/pirat_nation/status/China-university-AI-overhaul-2026-06-16