A Wall Street Journal investigation has pulled back the curtain on the political machinery behind China’s domestic AI chip drive, revealing a top-down mobilization campaign led by Vice Premier Ding Xuexiang, one of Xi Jinping’s most trusted lieutenants. According to WSJ reporters Josh Chin and Raffaele Huang, Ding has invoked the language of Mao-era mega-projects, framing the effort to replace Nvidia inside China as a national security imperative, and branding companies that resist adopting domestic chips as “traitors.”
The reporting confirms what industry insiders have long suspected: the pressure on Chinese cloud operators, model labs and internet giants to abandon Nvidia hardware is not simply market-driven or a byproduct of US export controls. It is a coordinated political campaign run at the very top of the Chinese state, with Huawei as its chosen national champion.
The “Two Bombs, One Satellite” playbook, applied to silicon
Ding Xuexiang, a Politburo Standing Committee member and Xi’s former chief of staff, has explicitly modeled the AI chip push on China’s 1960s “Two Bombs, One Satellite” program, the crash effort that produced Beijing’s atomic bomb, hydrogen bomb and first satellite in the space of a decade. According to the WSJ, Huawei briefed Ding directly on its AI chip roadmap and told him that China could achieve self-sufficiency in some AI hardware categories within three years.
That framing matters. In Chinese political vocabulary, invoking the “Two Bombs, One Satellite” precedent signals that failure is not an option and that market efficiency arguments will lose to strategic imperative. It also signals unlimited resource mobilization, state banks, local government financing vehicles, procurement quotas and academic labs all pulled into a single vertical stack.
The blunt “traitor” warning aimed at Chinese AI companies who preferred Nvidia hardware is consistent with the tightening EastFrontier has tracked over recent months. In May, Beijing banned Nvidia’s RTX 5090D V2 during Jensen Huang’s visit precisely to pressure tech giants toward domestic silicon. In April, Beijing had already barred foreign AI chips from state-funded data centers. Ding’s rhetoric is the political scaffolding underneath those individual moves.
Huawei’s 1.5 million-chip year
The most concrete number in the WSJ report is Huawei’s shipment target: roughly 1.5 million AI chips in 2026, approximately double its 2025 volume. That figure aligns with EastFrontier’s earlier reporting on Huawei’s Ascend 950 scaling to 750,000 units in 2026 and the company’s forecast that AI chip revenue would jump 60% to $12 billion this year.
The scale is significant but should be read in context. Nvidia still ships an order of magnitude more advanced AI accelerators globally each year, and Huawei’s Ascend line remains behind Nvidia’s Blackwell generation on per-chip performance. But Beijing’s strategy does not require chip-for-chip parity. It requires enough domestic supply to keep the country’s frontier labs — DeepSeek, Moonshot, Zhipu, MiniMax, Alibaba, ByteDance — training and serving models without depending on export-controlled American silicon. On that metric, 1.5 million Ascend chips represents a genuine inflection point.
The strategy is already producing results. Z.ai’s GLM-5.2 was trained entirely on Huawei chips and now tops open-weight rankings. DeepSeek V4 runs on Ascend hardware, and Huawei-led researchers have demonstrated full-parameter post-training of DeepSeek V4 Pro on the Ascend 910C. Even a year ago, none of that was operationally proven.
A campaign that also hits Chinese memory rivals
Ding’s “all-out” approach is not without friction inside China’s own supply chain. Reuters reported this week that memory chipmaker CXMT, set to complete an $8.6 billion Shanghai IPO on July 27, recently expelled engineers from Huawei-linked equipment maker SiCarrier from its R&D zone after a pricing standoff. CXMT is now charging more than Samsung for 64GB DDR5 server memory, an extraordinary reversal for a Chinese memory maker that only recently erased a decade of losses. That kind of leverage means even inside the “national team,” some players are strong enough to push back.
The tension illustrates the limits of Ding’s political mobilization. Beijing can order state cloud operators to buy Huawei accelerators, but it cannot easily override commercial disputes between two of the country’s most valuable chip companies. And it cannot instantly close the gap in EUV lithography, where CXMT remains roughly two generations behind Samsung and SK Hynix in high-bandwidth memory, according to Reuters.
Jensen Huang’s Beijing flight, in a new light
The WSJ reporting also recontextualizes Nvidia’s continued engagement with China. In May, Jensen Huang boarded Air Force One for Beijing after a personal call from President Trump, and Nvidia has since won partial approval to sell H20 and even limited H200 chips to Chinese customers. From Ding’s vantage point, however, every Nvidia sale into China risks undermining the “traitor” narrative and giving Chinese companies a commercial escape valve from the domestic mandate. That is presumably why Beijing has repeatedly slowed or blocked Nvidia shipments even after Washington approved them, a dynamic EastFrontier covered when the US cleared H200 sales to 10 Chinese firms but Beijing blocked delivery.
The picture that emerges is one of two governments simultaneously trying to wall off their AI chip ecosystems, with the private sector — Nvidia on one side, Chinese cloud giants on the other, caught between the demands of politics and the logic of the market.
What Washington should take from this
For US policymakers still debating whether China’s chip self-sufficiency drive is a serious strategic threat or a propaganda talking point, the WSJ reporting should settle the question. When a Politburo Standing Committee member invokes the “Two Bombs, One Satellite” template and personally warns companies that non-compliance amounts to treason, the effort has moved beyond industrial policy into national security mobilization.
That has direct implications for the Trump-Xi summit scheduled for September 24, where AI is expected to be on the agenda for the bilateral meeting, and where the stakes of the chip mobilization campaign will be on full display.