The Quiet War Over Silicon, Sovereignty, and Apple’s Plan to Buy Chinese Chips

Apple is running out of chips. Not the processors it designs itself, but the everyday, commodity memory used by the entire industry which its products simply can’t function without. DRAM, the memory chips found inside every iPhone, Mac, iPad and Apple TV. NAND flash, the storage in every Mac SSD. Both are in critically short supply, and prices are skyrocketing.

The reason for the shortage is simple — the AI boom, giant cloud providers, the hyperscalers who run the server farms and GPU clusters that power AI. Between them all they’ve managed to pre-buy virtually all available manufacturing capacity from the major Fabs and locked it up years in advance. Every wafer going into an AI data centre is a wafer that isn’t becoming an iPhone component.

Most companies would have been in crisis years ago. Apple dodged a bullet, largely because of Tim Cook. The outgoing CEO built an unrivalled reputation on supply chain mastery, spotting shortages early, pre-buying capacity, securing long-term deals before competitors realised there was a problem. For a while, the strategy worked brilliantly. But the AI spending wave was bigger and faster than anyone had predicted. At some point the hyperscalers simply outbid Apple, locking up virtually all the capacity Apple had assumed was theirs for the taking. Cook’s famous playbook finally hit a wall it couldn’t buy its way around once the deals had been signed.

That left Apple exposed to a market it hadn’t been vulnerable to in years: constrained, volatile, and outside its control. The company it has turned to for a lifeline is CXMT. Whether that turns out to be Apple’s last chance saloon, or its hill to die on in a battle with Washington, is exactly what this piece is about.

The Art of Chip Warfare

In the high-stakes theatre of the tech war, semiconductors are usually framed as weapons of digital espionage or critical nodes of artificial intelligence. Whilst a multi-billion-dollar lobbying battle is quietly playing out in Washington it exposes a deeper reality: the global tech market is no longer free — it is managed from afar, and its primary currency is no longer efficiency, but raw geopolitical leverage.

At the eye of this current storm is Apple, starved of consumer grade memory chips (DRAM – Dynamic Random Access Memory) by a historic, AI-driven global shortage of chips. The iPhone maker is actively lobbying the US government for permission to source chips from ChangXin Memory Technologies (CXMT) — a Chinese memory manufacturer that the Pentagon recently designated as a ‘Chinese military company’ in June this year, under Section 1260H of the National Defense Authorization Act. Apple has not confirmed the discussions publicly, but the Financial Times, citing six people familiar with the matter, reported that the company has been in contact with Commerce Department officials and other members of the Trump administration since at least May of this year seeking that guarantee.¹ For the casual observer, this sounds like a death knell for any business relationship. However, in the high-stakes world of semiconductor trade, there is a vital distinction between this list and the Department of Commerce’s Entity List.

The 1260H list is primarily a tool for restricting Department of Defense contracting and signalling national security concerns; it carries significant reputational weight but does not, in itself, prohibit private companies from conducting trade. The Entity List, by contrast, is the nuclear option. It would trigger an outright ban on exports and imports, effectively cutting off CXMT from the US market and technology ecosystem.

Apple’s current lobbying effort is not an attempt to circumvent a ban that already exists — it is a preventative strike. Apple is effectively asking Washington to provide a ‘safe harbor’ by guaranteeing that CXMT will not be moved from the Pentagon’s watch list to the Commerce Department’s restricted Entity List. By positioning the request this way, Apple is playing a sophisticated game of geopolitical insurance, ensuring that their supply chain isn’t suddenly severed by a future regulatory pivot.

This clash cuts to the core of modern global trade, as one government attempts to stop another from benefiting from market conditions that no one was prepared for. Can corporate capital run a borderless playbook while two superpower states are busy drawing and re-drawing hard digital borders?

Engineering vs. The Empire

American opponents of this arrangement argue that buying these chips has two major downsides. The first is straightforward: purchasing from a Chinese state-backed industry pours funds into Chinese state coffers. The second, that DRAM chips will somehow be used to spy on consumers is technically unfounded, though as we’ll see, the broader security picture is rather more complicated than either side admits.

From a purely technical standpoint, Apple’s anti-espionage argument is unimpeachable. A DRAM chip is a passive, silicon scratchpad. It has no processing power, no firmware layer, and no network access. It connects directly to Apple’s proprietary on-board memory controller. It cannot “spy” on a consumer any more than a block of aluminium can.

The global market, meanwhile, has hit a historic shortage. Because allied memory giants like Samsung, SK Hynix, and Micron are aggressively pivoting their factories to manufacture premium High-Bandwidth Memory (HBM) for AI servers, a more expensive extremely fast type of memory used in GPUs and high end servers, which have better margins and are higher in the value chain than commodity memory chips. The market for standard consumer RAM has been starved of supply and prices have increased rapidly as a result. For every single wafer of HBM produced, roughly three wafers of standard consumer DRAM production are lost due to finite manufacturing capacity.

The result? Consumer RAM prices have surged, forcing Apple to raise retail prices on MacBooks and iPads, precipitating a $263 billion single-day wipeout of its market capitalisation earlier this year. Apple’s classic corporate playbook dictates a simple solution: find the vendor with spare capacity, inject capital to fund their expansion, and lock down preferred, private pricing in exchange for multi-year exclusivity.

CXMT is the vendor in question. Founded as Innotron in 2016, the company delivered its first silicon just two years later in 2018, when it settled on its current name. The foundations of its technology were acquired, rather than invented. In 2019 it obtained the technical data and patent portfolios of bankrupt German chipmaker Qimonda via a deal with Polaris Innovations. It subsequently successfully leveraged the IP as an engineering baseline to scale up its own Buried Wordline Cell Architecture.²

Along the way CXMT expanded its operations, transitioned through successive process nodes to shrink chip sizes, and restructured its supply chain to navigate US chip equipment export restrictions. In 2023 it delivered the first domestically designed LPDDR5 memory chips (LPDDR5 is a low power variant of DDR5, particularly suitable for battery powered devices) for high-end smartphones, with production capacity crossing 100,000 wafers per month. In 2024 it entered the HBM sector and began volume shipments of DDR5 memory modules for the PC OEM market. It recorded its first profitable fiscal year in 2025, something that would have been impossible without patient government backing from China’s National Integrated Circuit Industry Investment Fund, known as the ‘Big Fund’.

Today, CXMT operates as a highly capable, independent alternative completely outside of Western supply lines. But for Apple, that independence is no longer a virtue — it’s an impediment. Given the market supply squeeze, Apple must receive a green light from Washington before committing its resources to any remedy.

The Illusion of Starvation

When the Pentagon placed CXMT on its 1260H list under the philosophy of asymmetric containment, the logic was simple: prevent Western capital from bankrolling a Chinese national champion whose manufacturing base ultimately serves military ends.

But fresh industry data proves this “capital starvation” strategy is largely an illusion. CXMT’s IPO prospectus, filed on the Shanghai STAR Market, reveals a company that has already achieved financial escape velocity. Fuelled by a massive, captive domestic market and the broader DRAM supercycle, CXMT posted revenue of $7.4 billion in Q1 2026 alone (a massive 719% year-on-year increase) with net profit attributable to shareholders of approximately $3.6 billion for the same period.³ SemiAnalysis projects full-year 2026 revenue could exceed $50 billion, driven primarily by soaring average selling prices rather than dramatic market share gains.⁴

Those margin figures need context. The headline consolidated net margin of roughly 65% looks extraordinary, but it significantly overstates what public shareholders will actually receive. The reason is structural: CXMT controls 73-75% of the voting rights in its key fabrication subsidiaries through long-term acting-in-concert arrangements, while holding only around 31% of the economic rights in them. In plain terms, it consolidates the profits of fabs it mostly doesn’t own. The attributable net margin, once you strip that out, is closer to 49%. Still a remarkable figure by any industry standard, but one driven by the DRAM supercycle rather than structural dominance.⁵

Domestic Chinese hyper-scalers such as Alibaba Cloud are already pre-booking CXMT’s capacity, and Asian mobile giants including Xiaomi and Vivo are buying up every bit of silicon the company can produce.⁶ The truth is that CXMT does not need Apple to survive; it only wants Apple to confer prestige and reaffirm its pedigree as a top-flight manufacturer. Delivering to Western clients would also provide substantial revenue in hard international currency, rather than relying solely on domestic RMB-denominated deals.

Trapped by State Creation

The fundamental impasse between Washington and Cupertino is philosophical. Apple views CXMT as a commercial counterweight to a broken global oligopoly. Washington views CXMT as an arm of the Chinese state.

And Washington is not wrong. Unlike Western start-ups, CXMT was willed into existence by state capital. The Hefei municipal government bankrolled 80% of CXMT’s initial Phase 1 buildout in 2016, patiently absorbing an accumulated deficit of nearly $5 billion over a decade before the company turned a profit.⁷ Its corporate prospectus reveals complex acting-in-concert arrangements that hand 73-75% of factory voting control to state-directed entities, while those same entities hold only a minority economic stake, a governance structure the prospectus itself lists as a formal risk.⁸ This is not how Western semiconductor companies are structured, and it raises legitimate questions about where commercial decisions end and state directives begin.

But Washington’s security concerns extend beyond corporate governance. In 2023, ten former Samsung employees were arrested for allegedly stealing trade secrets — technology that had cost Samsung over $1 billion to develop across five years, and which represented the exclusive process used for mass-producing 10nm class DRAM at the time. A former Samsung executive involved in the case subsequently took up a role at CXMT as head of development.⁹ This is not a passive DRAM chip failing to spy on consumers, this is the kind of state-adjacent technology acquisition that gives security hawks a legitimate argument, and it deserves more than a dismissal.

China’s motivations are not hard to read. It has seen what happens when the US pressures suppliers to cut off key technical components from adversarial states. Most recently it saw the chip restrictions applied to Huawei and it has responded by funding its own alternatives with extraordinary patience. CXMT is the direct product of that strategic calculation.

Beyond this, the state relationship carries a structural ceiling that Apple’s lawyers cannot contract around. As Jack Ma infamously discovered, the Chinese state retains ultimate supremacy over private capital. Washington knows that a commercial agreement signed with Apple today can be rewritten by a decree from Beijing tomorrow.

A Conditional Compromise

So, will Apple get the go-ahead?

The most realistic scenario is a calculated bureaucratic compromise: the US government will likely grant Apple a preliminary, rolling waiver subject to ongoing review. Washington will not give Apple a permanent green light, nor will it completely starve one of its most iconic corporate titans of the resources it needs to compete globally. Instead, the Department of Commerce will use the waiver as a strategic leash.

• The Regulatory Leash: A rolling 12-to-24-month waiver keeps a permanent “kill switch” armed. If Beijing escalates regional trade or political tensions, the waiver can be revoked instantly, ripping a premium revenue stream away from CXMT right as it navigates its upcoming domestic IPO.
• The Intelligence Hook: To maintain the waiver, Apple will likely be required to provide Washington with granular, ongoing supply chain audits, handing the US government unprecedented visibility into China’s internal manufacturing yields and equipment capacities.
• The Micron Bridge: A temporary approval perfectly manages the multi-year gap until US domestic capacity is ready. It allows Apple to survive the current memory drought but mandates a systematic wind-down of Chinese sourcing by 2029-2030, the exact window when Micron’s megafabs in Idaho and New York are scheduled to achieve high-volume production.

At the end of the day, neither Washington nor Beijing will make Apple happy, leaving the company marooned in the middle of a geopolitical battlefield. The uncomfortable reality is that the modern tech supply chain is just another front line in the cold war for silicon sovereignty.

while (trump.isHappy()) {
    apple.buyMemory(supplier: "CXMT");
}
// warning: loop may terminate without notice

Footnotes

1. Financial Times, Apple lobbies Trump administration to protect Chinese chip supplier from US blacklist, June 2026. https://www.ft.com/content/apple-cxmt-lobbying (paywalled; reported widely including by Reuters and Bloomberg)
2. SemiAnalysis, China’s CXMT Is Set to Challenge DRAM Incumbents, June 2026. https://newsletter.semianalysis.com/p/chinas-cxmt-is-set-to-challenge-dram
3. South China Morning Post, CXMT’s blowout earnings outlook fuels rally in Chinese memory chip stocks, May 2026. https://www.scmp.com/tech/tech-trends/article/3353988/cxmts-blowout-earnings-outlook-fuels-rally-chinese-memory-chip-stocks
4. SemiAnalysis, China’s CXMT Is Set to Challenge DRAM Incumbents, June 2026. https://newsletter.semianalysis.com/p/chinas-cxmt-is-set-to-challenge-dram
5. SemiAnalysis, China’s CXMT Is Set to Challenge DRAM Incumbents, June 2026. https://newsletter.semianalysis.com/p/chinas-cxmt-is-set-to-challenge-dram
6. Douglas Research, Initial Thoughts on the Changxin Memory Technologies (CXMT) IPO, May 2026. https://douglasresearch.substack.com/p/initial-thoughts-on-the-changxin
7. Gasgoo / Autonews, CXMT Unveils the Aggressive Expansion of Domestic Memory Chips, May 2026. https://autonews.gasgoo.com/articles/news/cxmt-unveils-the-aggressive-expansion-of-domestic-memory-chips-2056381885041795073
8. SemiAnalysis, China’s CXMT Is Set to Challenge DRAM Incumbents, June 2026. https://newsletter.semianalysis.com/p/chinas-cxmt-is-set-to-challenge-dram
9. Wikipedia, ChangXin Memory Technologies, citing Reuters. https://en.wikipedia.org/wiki/ChangXin_Memory_Technologies