The Price of Abundance

Part one of a diptych on how China and America finance capacity nobody is paid to count, and who ends up holding it.

ChangXin Memory Technologies lost money for nine years. In July it asked the Shanghai Stock Exchange for 58 billion yuan, the largest sum a semiconductor company has ever raised on a mainland exchange. At 8.66 yuan a share, the company was worth something close to 580 billion before a single lot had traded. A constellation of related stocks hit their daily limits that morning as anticipation spread beyond the institutions managing the sale.

Every institution that touched the offering did its job. The result says less about negligence than about how those duties have been divided.

The national guidance fund’s own literature describes it as an early bearer of risks no private institution will touch. It operates according to a whole-portfolio logic under which the failure of an individual company need not invalidate the programme. Nine years of losses, including an inventory writedown of 11.5 billion yuan in 2023 alone, enter that account as expenditure on capability. Hefei works to a different horizon. Its investment vehicles hold at least a third of CXMT and control the main fabricating subsidiary through concerted-action agreements that turn 30.68 percent of the economics into 73.01 percent of the votes. For the municipality, the working fab already constitutes a return, whatever memory earns in 2029.

The underwriter’s horizon is measured in weeks. Its success consists of a fee on the proceeds and an orderly stabilisation period, after which the greenshoe expires and the bank’s mandate ends. The exchange and the regulator decide whether the issuer qualifies, and CXMT did. Utilisation had climbed from 87 percent to roughly 96, while first-quarter revenue reached 50.8 billion yuan after a loss-making quarter a year earlier. A filing reporting revenue growth of 719 percent was unlikely to be returned for want of momentum.

The danger appears only when the field of vision widens. A firm can be sound by one measure, a transaction successful by another, while thirty individually justified projects create a market in which nobody can earn an adequate return. Chinese policymakers have worried about that possibility for more than a decade and built an elaborate apparatus to prevent it.

In October 2013 the State Council issued Document 41, naming five industries in severe excess, from steel and cement to shipbuilding. The order barred government departments from approving new capacity projects in those sectors and instructed banks to withhold credit from projects lacking the necessary permissions. Unstarted construction was forbidden to begin, while work already under way was suspended pending central review. The same document prevented offending companies from issuing bonds or listing on a stock exchange. New capacity thereafter had to be matched by the retirement of old plant, with still tighter requirements around Beijing, Shanghai and Guangzhou.

Document 41 also ordered the relevant ministries to build a dynamic monitoring system and an early-warning mechanism for industrial excess. Both were among the objectives to be achieved within five years. Beijing therefore possessed a diagnosis, a set of prohibitions and a deadline well before the latest round of overcapacity began.

Document 51 followed two years later and raised the minimum capital ratios for fixed-asset investment. Steel and electrolytic aluminium were set at 40 percent, with cement at 35. A 30 percent floor applied to coal and several energy-intensive chemical industries, including polysilicon. Financial institutions were instructed to enforce the earlier restrictions wherever severe excess had already been identified.

Polysilicon consequently carried an elevated financing hurdle from 2015. Chinese manufacturers nevertheless built somewhere between three-quarters and nineteen-twentieths of world capacity, roughly twice global demand. Module prices fell by about half in 2023 and continued downward. Polysilicon itself dropped from more than 300,000 yuan a tonne to around 30,000, while thirty-one listed solar companies lost a combined 12.6 billion yuan in a single quarter of 2025.

The capital-ratio rule remained in force throughout. It had been written for banks considering identifiable projects, whereas much of the new capacity arrived through local-government support routed through investment vehicles and provincial credit. A national bank could be prevented from financing another plant without stopping a municipality from acquiring an industry.

Beijing’s answer arrived in 2025 through an anti-involution campaign covering ten sectors. Two-year plans held output growth below its 2024 level, while a national energy-consumption standard gave inefficient polysilicon producers a deadline to improve or close. The planning commission and industry ministry also endorsed a proposed 50 billion yuan fund to purchase obsolete capacity and retire it. These measures began once the investment had been made, although the early-warning mechanism commissioned in 2013 had been due in 2018.

Sectoral campaigns can move investment elsewhere without changing the incentive beneath it. Michael Pettis has observed that investment slowed in electric vehicles and solar while accelerating in petrochemicals. The utilisation rates that caused five industries to be named in 2013 ranged from 71.9 to 75 percent. Chinese manufacturing as a whole was operating at about 74 percent in 2025.

Memory remains outside that regime because Beijing still treats domestic chip capacity as a strategic deficiency to be remedied. In September 2023 Xi Jinping instructed cadres to develop new quality productive forces, and local governments responded with the enthusiasm that such an instruction invites. Two years later he criticised excessive investment in the industries he had encouraged. The fourth plenum communiqué of October 2025 continued to demand technological self-reliance while the anti-involution campaign was under way. Both policies remain operative, with each industry’s treatment determined by its political classification.

Semiconductors were absent from the 2013 restrictions and the 2015 capital schedule. The Semiconductor Industry Association told the US Trade Representative last year that the anti-involution campaign was not directed at China’s chip industry. Domestic-content mandates operating across several levels of government manufacture demand for Chinese chips while the state restrains supply elsewhere. Document 41 contains a real and enforceable prohibition on stock-market listings, but that prohibition points towards shipyards rather than memory fabs.

A household purchasing CXMT was also buying an assumption about the manners of an oligopoly. Three companies have controlled around 90 percent of the DRAM market for a decade and, since 2016, have generally resisted destroying prices during downturns. That restraint underlies the current re-rating. High-bandwidth memory now absorbs close to a quarter of their wafer input while returning only about 9 percent of the bits, since stacking and yield loss make an HBM bit consume roughly three times the wafer area of a DDR5 one. The resulting shortage travels down through the product range.

Such squeezes have occurred before. Each eventually ended when one manufacturer decided that market share mattered more than price, and Samsung’s chairman is reportedly pressing for that course now. The present shortage may last longer because AI demand is consuming wafer capacity so rapidly, but the discipline supporting current valuations remains a choice made by producers rather than a permanent property of the technology.

CXMT accounted for 7.67 percent of global DRAM bits and ranked fourth. It supplies Xiaomi, OPPO, vivo and Transsion, while PC manufacturers have reportedly booked its capacity into 2027. Those selling the flotation argue that the company remains too small to move the global price. Yet prices are set at the margin by the producer willing to sell for least, and previous collapses began in the lower end of the market where CXMT is strongest.

The company itself was assembled from the wreckage of the last participant a price war destroyed. Qimonda, Infineon’s memory arm, entered insolvency in January 2009 and ended Germany’s last serious position in DRAM. CXMT founder Zhu Yiming subsequently acquired roughly 2.8 terabytes of its technical files and spent something close to $2.5 billion turning the remains of German process engineering into a Chinese manufacturing platform. Knowledge that had become worthless on one balance sheet reappeared in Hefei as a production line.

That history also marks the limit of the comparison with solar. Once a solar-production process has been mastered, additional capacity can rapidly produce modules that are difficult to distinguish from one another. DRAM fabrication requires continuing advances in yield and qualification, so additional floor space does not by itself create a competitive producer.

The Korean industry offers the successful precedent for Hefei’s wager. Samsung entered memory as a state-supported latecomer and survived losses widely expected to finish it before becoming dominant. Creditor banks rescued Hynix after 2001, and the returns that followed made the intervention look like patience rather than charity. Hefei is entitled to see those cases as precedents, although CXMT confronts the consolidated firms produced by those earlier struggles. Its Korean rivals are profitable, technically ahead and no longer dividing an incumbency among themselves.

CXMT’s prospectus remains candid about the risk. Average selling prices moved 55 percent in 2024 and 34 percent in 2025, and the company warns that a reversal in the balance between supply and demand could make its first-half growth unsustainable. The valuation therefore depends most heavily on the figure the issuer itself identifies as unstable.

Zhongji Innolight went to Hong Kong three days later seeking as much as $7 billion. It was already the heaviest weighting in the CSI 300 and the world’s largest optical-interconnect supplier by revenue. Its first-quarter net profit exceeded its earnings for the whole of 2024, giving the offering a commercial record CXMT has yet to acquire.

Between November and January, Zhongji’s controlling shareholder sold 5.5 million shares through block trades worth close to 2.9 billion yuan. An insider named Wang Xiaodong disposed of around 300 million yuan as the price crossed 1,000 yuan. Such sales admit many explanations, but the people personally exposed to the downside were reducing their positions while outside investors prepared to increase theirs.

The public lock-up schedule provides one way of testing the account. Substantial selling by Hefei’s vehicles or the guidance fund into a strong market would suggest that somebody had been marking the cycle after all. An output ceiling or binding energy limit imposed on memory while the shortage continued would show that the state had learned to restrain aggregate capacity before the financial damage appeared. Neither development currently seems as probable as impaired loans remaining unrecognised somewhere in Anhui.

The generous interpretation deserves to be stated plainly. Accelerator demand may consume bits quickly enough for CXMT’s new capacity to arrive without producing another glut, allowing even the fourth firm to earn an adequate return. No computing cycle has sustained the required compound growth in bit demand for much longer than three years, however, and the valuation leaves little room for that growth to falter.

The timing also admits a more prosaic explanation. Utilisation reached 96 percent and the first full-year profit arrived, followed by quarterly revenue growth of 719 percent. Companies list when their figures permit them to do so, and regulators approve issuers that satisfy the rules. July required no ingenuity in Anhui because ordinary institutional competence was sufficient to produce the sequence. That is also why the outcome cannot be repaired by locating the person responsible.

Hefei will retain engineers trained through somebody else’s drawdown and an equipment base that no longer has to be imported. The cost of building solar was eventually distributed among public balance sheets capable of carrying it, and the same options exist here. Anhui and the guidance fund can hold part of the burden, with the state banking system standing behind both. Recognition can be deferred while the industrial residue accumulates.

Whether that residue ultimately covers the financial cost remains a separate question. The household cannot enter trained engineers or strategic autonomy against a falling share price, and will encounter the answer in whatever another buyer is eventually prepared to pay.