🧊 The AI Dollar: Part 3/6: China's Dominance Is Fake

February 3, 2026Bloomberg: $35/month. Financial Times: $42/month. The Economist: $17/month. Original analysis by Tatsu with 40+ footnotes: $8/month.Share this preview with others.China looks terrifying on satellite imagery.Scroll through defense analyst Twitter and you'll see pictures of new destroyers sliding down shipyard ramps, rows of J-20 stealth fighters parked on tarmacs, carrier battle groups conducting exercises. The Pentagon's annual China Military Power Report dutifully catalogs the growth: more ships than the US Navy, more missiles, more everything. The narrative writes itself: the rising power will inevitably eclipse the declining hegemon.Look closer. The picture changes.This is Part 3 of our series on the AI Dollar, and it addresses the elephant in every analysis of American hegemony: what about China? If the United States is supposed to decline like the Dutch and British before it, China is supposed to rise like Britain and America did. Dalio says so. The Pentagon says so. Conventional wisdom says so.Conventional wisdom is wrong. Not about China's ambitions, which are real, but about its capabilities, which are hollow.Full investigation below. $8/month for novel, footnoted deep analysis.The Brittle Peer ConceptLet me introduce a framework that better captures China's military and technological position than either "paper tiger" (too dismissive) or "peer competitor" (too credulous).China is a brittle peer.In static metrics, the numbers look like parity:* Ship counts: PLAN has more hulls than the US Navy* Missile counts: PLARF can saturate any target in the Pacific* Aircraft counts: PLAAF fields hundreds of fourth and fifth-generation fighters* Chip counts: SMIC can produce 7nm semiconductorsBut static metrics don't fight wars. Dynamic capabilities do. And in the dynamic domains that actually matter, systems reliability, propulsion maturity, acoustic signatures, manufacturing yields, operational experience, China faces gaps that aren't closing. They're widening.A brittle peer is dangerous. It can impose costs. It can contest. But it breaks under stress in ways a true peer does not. Let me show you where the fracture lines are.Aerospace: The Engine ProblemThe Chengdu J-20 "Mighty Dragon" is China's pride, a twin-engine stealth fighter that looks vaguely like a stretched F-22. Production has ramped impressively; the PLAAF is fielding approximately 100-120 airframes per year, matching or exceeding F-35 deliveries to the US Air Force. The "paper tiger" thesis clearly doesn't hold for raw numbers.But look at what's under the skin.The canard compromise. The J-20 uses a canard-delta configuration, small forward wings that improve low-speed handling and lift. This design choice is revealing: it compensates for inadequate engine thrust. The problem is that canards move during flight, and moving surfaces increase radar cross-section. The J-20's stealth is optimized for frontal aspect only. Come at it from the side or above and its signature blooms. Compare this to the F-22's all-aspect stealth or the F-35's carefully shaped surfaces with no forward canards.The WS-10 reality. For most of its operational life, the J-20 has flown with WS-10C "Taihang" engines, domestically produced but derived from Russian designs. These engines work. They're reliable enough for training and patrol. But they don't enable "supercruise," the ability to sustain supersonic flight without afterburners. Without supercruise, the J-20 must light afterburners to go fast, which burns fuel at prodigious rates (reducing range) and creates a massive infrared signature (making the aircraft visible to heat-seeking missiles and infrared sensors).The F-22 has had supercruise since 2005. The J-20 doesn't have it in 2026.The WS-15 mirage. China's intended solution is the WS-15 "Emei," an engine that would theoretically match the F119 engine's performance in the F-22 Raptor. Reports suggest it's entered flight testing, possibly low-rate production. But there's a reason it's taken this long: high-bypass military turbofans require single-crystal turbine blades that can withstand temperatures exceeding 1,600 degrees Celsius. This is metallurgy at the atomic level. The US spent decades developing these materials. China is still struggling.Even if the WS-15 enters service, its durability is questionable. Western analysts expect significantly shorter Mean Time Between Overhauls (MTBO) than American equivalents, meaning more frequent engine swaps, more logistics burden, lower fleet readiness rates.The J-20 looks like a fifth-generation fighter. It flies like a 4.5-generation fighter with stealth coating.Naval: The Fujian QuestionIn November 2025, the PLA Navy commissioned the Fujian (CV-18), China's third aircraft carrier and its first with electromagnetic catapults (EMALS). The propaganda value was immense: China had leapfrogged the older steam catapult technology to match the USS Gerald R. Ford.The propaganda was accurate about the technology. It was silent about the physics.The power problem. EMALS requires enormous electrical power, massive current draws to accelerate aircraft from zero to 170 mph in two seconds. The USS Ford generates this power from two A1B nuclear reactors with effectively unlimited energy output. The Fujian generates it from diesel engines and steam turbines.This creates what naval architects call a "parasitic load dilemma." When the Fujian runs its catapults at combat tempo, it drains stored energy that the conventional power plant must regenerate. This forces tradeoffs: top speed versus launch rate, radar operation versus catapult cycling. The Ford doesn't face these tradeoffs. Its reactors produce more power than it can use.The sortie gap. The metric that matters for aircraft carriers is Sortie Generation Rate (SGR), how many aircraft you can launch and recover in 24 hours of sustained combat operations. The Ford-class is designed for 160 sorties per day. Analysts estimate the Fujian's capacity at roughly 70-80, about half.Why? Two deck elevators versus the Ford's three. Conventional logistics tether (the Fujian has to refuel itself, consuming time and requiring vulnerable replenishment ships). Limited crew experience with high-tempo operations.The Shandong, China's second carrier, achieved a milestone of 10,000 cumulative sorties over five years since commissioning. That averages to fewer than 6 sorties per day, a training tempo that wouldn't last one morning in a high-intensity conflict.The air wing question. The Fujian will eventually carry the J-35, a medium-weight stealth fighter. But the J-35 faces the same engine constraints as the J-20, using the WS-21, a derivative of the WS-13, with similar performance limitations. And the KJ-600 airborne early warning aircraft, critical for extending the carrier's radar horizon, is still in development.The Fujian is real. It's impressive. It's roughly 60% as capable as its American equivalent.Undersea: The Acoustic GapIf you want to understand the difference between static metrics and dynamic capability, examine submarines. On paper, China has a substantial undersea fleet. In the water, it's loud.The Type 093B Shang-class is the backbone of China's nuclear attack submarine force. Its acoustic signature, the sound it makes that allows enemies to detect it, is roughly comparable to US Los Angeles-class Flight I submarines or Russian Akula I boats. Those are 1980s designs. The Type 093B entered service in the 2010s.Sound is everything in submarine warfare. The side that detects first kills first. A multi-decibel disadvantage isn't a minor technical issue; it's a death sentence in combat. US Virginia-class submarines, with advanced pump-jet propulsors and sound-isolation mounting, are literally a generation quieter.The Type 095 Tang-class is supposed to close this gap, with Chinese defense publications describing pump-jets, shaftless drives, and hybrid propulsion systems. The projected capability approaches Virginia-class quietness. The projected timeline keeps slipping. Quietness requires extreme precision in manufacturing, perfectly machined gears, vibration-damped shafts, hydrodynamically flawless hulls. This is the same metallurgical challenge that plagues the engine programs.The Wuhan incident. In mid-2024, satellite imagery confirmed that a new nuclear submarine sank pier-side at the Wuchang Shipyard near Wuhan. It sank while being fitted out, before ever going to sea. The causes remain unclear, but the incident points to quality control failures at the most critical phase of submarine construction. A navy that can't keep its newest boats afloat at the dock faces questions about keeping them alive in the crushing pressures of deep-ocean operations.US submarines operate with relative impunity in the First Island Chain. Chinese submarines are routinely detected transiting chokepoints like the Miyako Strait. The undersea domain remains America's decisive advantage in any Western Pacific conflict.Strategic: The Water Missile ScandalIf the aerospace and naval gaps are concerning, the strategic missile gap is existential.China's PLA Rocket Force (PLARF) is supposed to be the regime's ace, the massive missile arsenal that can overwhelm any defense and hold American bases and carriers at risk. The DF-27 can reach Hawaii with a hypersonic glide vehicle. The DF-17 can strike regional targets on depressed trajectories that evade missile defense radar. The quantity and variety are genuinely impressive.Then came the purges.In 2024 and 2025, the Chinese Communist Party launched the most significant anti-corruption campaign in the military since the 1990s. The Rocket Force was the primary target. Defense Minister Li Shangfu was removed. Multiple generals were prosecuted. The leadership was gutted.Why?U.S. intelligence leaks provided the answer: some missiles had been filled with water instead of fuel.Let that sink in. The silo-based strategic deterrent, the force that's supposed to guarantee China's security against American attack, contained missiles that could not fly. The corruption was so pervasive that basic maintenance, filling a missile with propellant, had been subverted for graft.Other reports indicated silos with lids that couldn't open due to mechanical defects or poor maintenance. The missiles inside were effectively buried alive, unusable even if ordered to launch.This isn't a minor procurement scandal. It's a crisis of institutional integrity. If you're a Chinese general planning a Taiwan operation, can you trust that the missile salvo you're counting on will actually launch? If 10% of your missiles are water-filled, your calculus changes. If 20% are?The Rocket Force purges suggest the corruption was systemic, reaching the highest levels of command. The "Paper Tiger" thesis is most applicable here: not to the technology (which is world-class in design) but to the institution (which is rotten).Silicon: The 20% ProblemWe covered the semiconductor gap in Part 2, but it's worth restating in the context of China's broader "brittle peer" status.SMIC can produce 7nm chips. This is true. What's also true:Yield rate: ~20%. Eighty percent of SMIC's advanced wafers are waste. TSMC yields exceed 90% on mature processes. SMIC requires roughly 4-5 wafers to produce what TSMC produces with one.Cost per chip: 4-5x. The yield problem translates directly to cost. Every working Huawei Ascend 910B costs the Chinese state multiples of what an equivalent Nvidia chip costs on the open market. This isn't commercial competition; it's strategic subsidization.Capacity ceiling. SMIC's total 7nm capacity is roughly 45,000 wafers per month. TSMC's total advanced-node capacity dwarfs this by orders of magnitude. China can make the chips; it can't make enough of them.No path to 5nm. Without EUV lithography (which ASML will not sell), SMIC cannot advance to 5nm or 3nm nodes. The multi-patterning workaround becomes exponentially more difficult at smaller nodes. Each generation, China falls further behind, not closer.AI: The Software MirageDeepSeek's V3 and R1 models shocked Western observers. Competitive with GPT-4 on benchmarks. Trained at a fraction of the reported cost. Chinese AI, it seemed, had arrived.Look closer.DeepSeek trained its models on Nvidia H800 chips, acquired before export controls tightened, and legacy A100s. Not on domestic Huawei Ascend hardware. The innovation was architectural, clever software optimizations that squeezed maximum performance from a constrained compute budget. The software engineers are legitimately elite.But the hardware underneath is American. When the smuggled chips run out, when the grey-market supply dries up, what then?Huawei's Ascend 910B exists. It performs on benchmarks. But its software stack (CANN) is immature compared to Nvidia's CUDA ecosystem, prone to crashes and compatibility issues that slow development. Chinese AI labs use Ascend for demonstration and Nvidia for production, when they can get Nvidia.The talent drain. Per Carnegie Endowment research, 87% of Chinese AI researchers who publish at top-tier conferences (NeurIPS, ICML) choose to remain in the United States. The very best leave. China keeps the second tier.This is the software mirage: impressive results achieved with unsustainable infrastructure, optimizing around constraints that will only tighten as export controls mature.The Corruption TaxUnderlying all these specific capability gaps is a systemic problem: you can't trust your own numbers.The water-missile scandal isn't an isolated incident. It's a symptom of an incentive structure that rewards reported metrics over actual capability. When promotion depends on hitting targets, targets get hit on paper even when they're not hit in reality. When anti-corruption campaigns are political weapons (used against rivals, not applied systemically), corruption metastasizes in the spaces where no one is looking.A Chinese commander cannot know, with certainty, that:* His missiles have fuel* His submarines' hulls were welded properly* His aircraft engines will last their rated hours* His chips will perform to specThis uncertainty imposes what I call a "corruption tax" on military planning. Every capability must be discounted by some unknown factor representing possible graft. A force that looks like 100 units might function like 70 or 80 or 60.The United States has procurement scandals and cost overruns, but it does not have missiles filled with water. The institutional rot is qualitatively different.The Brittle Peer SummaryLet me synthesize:Where China is genuinely strong:* Shipbuilding tonnage (50% of global capacity)* Missile quantity and variety* Industrial mass for wartime reconstitution* Certain software domains (AI architecture, app development)Where China is brittle:* High-performance propulsion (jet engines, submarine drives)* Acoustic stealth (submarines remain 1980s-equivalent)* Semiconductor fabrication (yields, costs, capacity)* Institutional integrity (corruption undermines reliability)* Talent retention (top researchers leave)The gap isn't closing. At the bleeding edge, where competitive advantage is determined, China faces physics problems that money and determination can't quickly solve. Engine metallurgy takes decades to master. EUV lithography can't be stolen or reverse-engineered. Quiet submarines require manufacturing precision that's cultural as much as technical.The "rising power eclipses declining hegemon" narrative assumes China is on a trajectory toward parity. The evidence suggests a different trajectory: impressive on static measures, struggling on dynamic ones, and falling further behind at the frontier.In Part 4, we'll examine what happens when this brittle peer tries to execute the most complex military operation in history: an invasion of Taiwan. Spoiler: the physics don't cooperate there either.NotesNotes[1] J-20 production estimates from Military Watch Magazine and DoD annual reports.[2] WS-10C and WS-15 status from DoD China Military Power Report 2025.[3] Fujian commissioning and EMALS analysis from Naval Technology and USNI News.[4] Shandong sortie data from Global Times, Chinese state media.[5] Type 093B acoustic comparison from defense analysts and Congressional Research Service.[6] Wuhan submarine incident confirmed by ORF Online and Western intelligence assessments.[7] Rocket Force corruption from Arms Control Association; water-filled missiles from US intelligence leaks reported in multiple outlets.[8] SMIC yield analysis from Georgetown CSET.[9] DeepSeek hardware from RAND and Tom's Hardware.[10] AI researcher retention from Carnegie Endowment. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit tatsuikeda.substack.com/subscribe