Dwarkesh PodcastArthur Kroeber on Dwarkesh Patel: Why BYD Beat Tesla at Home
How BYD beat Tesla after China spent 200 billion in EV subsidies; Beijing treats technology acquisition like a VC fund willing to lose for decades.
CHAPTERS
- 0:00 – 9:43
Why China getting richer feels destabilizing: trade surpluses, politics, and the social compact
Kroeber unpacks why anxiety about China’s rise is less about China becoming wealthy per se and more about how it gets rich—via an export- and surplus-driven manufacturing model—and what that does to other countries’ politics. He emphasizes the legitimacy gap between democratic self-identity in the US and China’s successful authoritarian system, and how that friction shapes global rules and narratives.
- •Critique of China’s model: manufacturing dominance, persistent trade surpluses, and perceived unfairness
- •Welfare gains from cheap goods vs political backlash in deindustrialized societies
- •Systemic political differences: US democratic identity vs China’s authoritarian legitimacy
- •Need for a shared ruleset to integrate China’s scale into the global economy
- •US domestic policy failures (redistribution, social contract) often get scapegoated onto China
- 9:43 – 14:57
Why a Cold War frame fails: deep economic integration and no “end state” of collapse
The conversation argues that the US–China conflict can’t be treated like the US–USSR Cold War because trade and investment ties are orders of magnitude larger. Kroeber stresses there is no realistic endpoint where China ‘goes away’ or transforms into a fundamentally non-Chinese strategic actor.
- •Cold War analogy breaks empirically: China is deeply embedded in US trade and corporate investment
- •A true Cold War would imply near-zero trade/investment—politically and economically difficult
- •China’s model is ‘successful on its own terms’ and beneficial to many global stakeholders
- •Even regime change wouldn’t erase China’s core strategic goals (security, tech self-reliance)
- •Chinese nationalism could be even harder to manage under full electoral politics
- 14:57 – 22:07
What a workable bargain could look like: Chinese manufacturing investment in the US—and the data/dual-use barriers
Dwarkesh proposes a division of labor (China gets EVs/solar; US gets AI/chips), but Kroeber argues China won’t accept being ‘pigeonholed.’ He offers a more realistic win-win: allow regulated Chinese direct investment in US manufacturing, while recognizing Washington’s security concerns around data flows and universal dual-use tech.
- •Grand bargains that cap China’s ambitions are not credible; China wants the full tech stack
- •Proposal: welcome Chinese FDI in US manufacturing (EVs, green energy, automation) to rebuild industrial capacity
- •Why Washington resists: data as a strategic asset; modern manufacturing as data-creation
- •Shift from ‘some dual-use tech’ to ‘everything is dual-use’ complicates openness
- •A regulated framework (data localization/rules of the road) could make selective openness feasible
- 22:07 – 29:34
BYD vs Tesla: how China’s EV ‘leapfrog’ strategy finally worked
Kroeber tells the arc of China’s long attempt to build an auto industry, its failure in conventional internal-combustion vehicles, and the pivot to EVs. Tesla’s Shanghai entry becomes the catalytic shock that forces Chinese firms to master consumer design and product appeal on top of batteries and software.
- •1990s JV strategy in conventional autos largely failed to create globally competitive Chinese champions
- •Late-2000s pivot: ‘leapfrog’ into EVs with broad subsidies and supply-chain focus
- •Tesla’s wholly-owned Shanghai factory (2018–2019) changes consumer expectations and competitive dynamics
- •Chinese firms strengthen design capability (including hiring global talent) and catch up quickly
- •China behaves like a patient ‘giant VC fund’: tolerate losses until a few bets pay off
- 29:34 – 37:58
Why China’s industrial policy can succeed: export discipline, domestic competition, and tolerance for failure
They debate whether China is uniquely good at picking winners; Kroeber reframes success as ecosystem-driven rather than clairvoyant central planning. Export orientation forces global competitiveness, domestic competition tests ideas, and scale allows long-run subsidy cycles that democracies often abandon after a single failure.
- •Strategic Emerging Industries list is broad and often ‘obvious’ rather than precise prediction
- •Export orientation prevents comfy import-substitution rent-seeking; you must compete globally
- •China’s domestic market is less protected than Japan’s; foreign competition validates/pressures upgrades
- •Scale and patience: accept many failures (contrast with US ‘Solyndra proves industrial policy fails’)
- •Industrial policy works best when embedded in a broader competitive manufacturing ecosystem
- 37:58 – 45:26
Will China face a Japan-style crisis? Why the balance-sheet mechanics differ
Dwarkesh raises Japan’s 1980s convoy/cross-shareholding model and asks whether China will ossify into champion-led conglomerates. Kroeber argues China’s financial and corporate sectors are ‘quarantined’ by design, reducing the chance of Japan-style debt-deflation dynamics—even though China has serious property and local-government debt issues.
- •Japan’s crisis mechanism: cross-shareholdings + land bubble → system-wide debt deflation
- •China bans key Japan-style linkages: industrial firms can’t own banks; banks can’t load up on industrial equity
- •China’s debt problems are real but more compartmentalized (developers, local governments)
- •Private industrial firms historically relied more on retained earnings; lower leverage than Japan then
- •Key difference: China’s geopolitical/security incentives create stronger pressure to keep the system dynamic
- 45:26 – 58:07
Local government debt: land finance origins, post-2008 excess, and why it’s ‘manageable but taxing’
Kroeber explains the rationale of land-backed local financing vehicles: it worked early when China was underbuilt and land was undervalued. The 2008 stimulus supercharged it via looser commercial-bank lending, creating waste and a long unwind; he disputes the highest debt estimates, stressing structure, servicing, and double counting—while conceding debt drags growth.
- •Land-backed local financing was centrally encouraged and initially effective for infrastructure/housing buildout
- •Post-2008 ‘spend, spend, spend’ turned it into a sprawling, poorly underwritten machine
- •Debt metrics debate: formal vs contingent liabilities and substantial double counting risks
- •China’s total debt ~300% of GDP; high for a middle-income country even if not collapse-prone
- •Contained, local-currency debt reduces crisis risk but depresses growth via leverage overhang
- 58:07 – 1:14:58
Why China isn’t richer per capita: the ‘efficiency fallacy,’ scale constraints, and Xi’s shift from GDP to tech KPIs
Dwarkesh challenges the apparent competence-versus-poverty puzzle; Kroeber responds that China’s growth was historically extraordinary, and ‘waste’ can be a byproduct of effectiveness at scale. He then argues growth has slowed below potential recently because Xi shifted incentives away from GDP maximization toward technology goals and tighter control.
- •China’s 1990s–2010s growth was historically unprecedented; claims it ‘could’ve been much faster’ may be naive
- •Effectiveness vs efficiency: loose capital allocation helped mobilize resources and spread development geographically
- •Per-capita catch-up is arithmetically harder with 1.4B people despite massive urbanization
- •Xi era: reoriented governance from growth-at-all-costs to technology targets and control
- •Result: slower growth and a trade-off between average living standards catch-up and strategic tech goals
- 1:14:58 – 1:25:32
How China keeps tech under control: re-regulating platforms, finance, and services—and the demand shortfall risk
Kroeber details the post-2020 regulatory turn: Ant’s halted IPO, platform crackdowns, and re-tightening finance and services. He argues these political choices constrain service-sector dynamism and jobs, weakening aggregate demand and contributing to deflationary pressure even as hardware-focused industrial policy advances.
- •Ant/Alibaba crackdown exemplifies fear of US-style financial risk and securitization
- •Platform regulation closes off fintech/telehealth/social expansion paths on stability and political-risk grounds
- •Broader re-regulation channels finance back into state banks; reduces private-sector credit and entrepreneurship
- •Hardware-tech fetishism plus service suppression can yield demand deficiency and deflation risk
- •Recent top-level signals suggest Beijing is beginning to recognize the need for a demand strategy
- 1:25:32 – 1:34:31
Two enabling bets China got right: informatization (internet) and electrification (cheap abundant power)
Kroeber argues China’s edge is less about picking sectors and more about building enabling infrastructure. He explains how early internet expansion enhanced both innovation and state control, then draws a parallel to electrification—massive generation, grid buildout, renewables scale—creating a structural advantage for manufacturing and power-hungry computing.
- •‘Informatization’ bet: internet as growth enabler and governance panopticon (control + capability)
- •China’s approach defied the 1990s belief that the internet would doom authoritarian regimes
- •Electrification: China now has >2x US generation capacity; renewables capacity alone rivals total US capacity
- •Ultra-high-voltage grid and low electricity prices as enduring competitive advantage
- •Cheap power becomes a key input to future industries, including data centers and AI deployment
- 1:34:31 – 1:44:20
Does China win AI? Scale advantages vs walled-garden limits and fragmented compute
Dwarkesh lays out a China-dominates-AI case: talent, eventual chip catch-up, and unmatched scaling of energy and deployment. Kroeber counters that AI’s economic returns may accrue mainly in applications, where openness and global market access matter; he also notes China cannot easily centralize compute into a single national champion.
- •Pro-China case: scale of talent, manufacturing, power buildout, and ability to deploy massive AI labor
- •Kroeber’s skepticism: biggest value may be in applications driven by consumer feedback loops
- •China’s ‘walled garden’ limits external experimentation compared to US firms’ broader global canvas
- •DeepSeek as an example of non-state innovation that can make Beijing uneasy
- •China is unlikely to force a monopoly compute allocation; major firms will pursue parallel efforts
- 1:44:20 – 1:55:09
AI safety and the ‘red telephone’: why communication channels are collapsing and why that’s dangerous
They discuss AI risk as a diffuse, multi-dimensional problem that doesn’t map neatly to Cold War nuclear hotlines. Kroeber warns that US–China communication structures have been dismantled since 2017, worsening crisis response (COVID as example) and raising the risk of miscalculation across domains, including AI.
- •AI governance is not a simple leader-to-leader deterrence problem; it’s decentralized and unpredictable
- •Trump-era removal of ~100 working-level dialogues left a thin communication architecture
- •COVID illustrates costs of missing backchannels (e.g., CDC-to-CDC links) during early outbreaks
- •Both countries hardened narratives: US distrust and China’s post-COVID inward turn/self-sufficiency
- •Restoring communication is urgent but politically difficult in both systems
- 1:55:09 – 2:18:19
What foreigners get wrong: extreme ‘China collapse’ vs ‘China dominates’ narratives and the cost of ignorance
Dwarkesh argues Americans’ direct exposure to China has dropped; Kroeber agrees and gives examples from teaching and from reduced journalism/consular presence. They criticize the magnetism of extreme narratives, the tendency to scapegoat China for domestic failures, and the policy errors that flow from caricatures.
- •Reduced travel/exchange leads to cartoonish perceptions; direct experience improves nuance
- •Information asymmetry is worsening: many Chinese students in US vs very few Americans in China
- •Policy own-goals: journalist expulsions, consulate closures, fewer ‘eyes on the ground’
- •Extreme narratives (total collapse/total domination) thrive under ignorance and drive bad policy
- •US manufacturing decline is long-run/technological; blaming China often substitutes for domestic reforms
- 2:18:19 – 2:27:29
Future of the China–US relationship: coexistence over containment, investment as a bridge, and a realistic negotiating agenda
Kroeber closes with a pragmatic vision: the relationship has no endpoint, so the goal is bounded competition and workable coexistence. He argues containment-by-bloc is nonviable because China has made itself indispensable as a trade partner to much of the world; instead, the US should strengthen its own system, rebuild selectively, and pursue safer channels of engagement—especially investment—while pushing China toward stronger domestic demand.
- •US lacks a coherent agenda; constructive negotiation requires clarity about goals
- •Best-case path: create permission structures for regulated Chinese investment in US manufacturing
- •China’s demand weakness forces export surpluses, fueling global trade frictions; rebalancing helps everyone
- •Cold War bloc strategy won’t work: China is top trade partner for ~140 countries and preempted ‘encirclement’
- •No ‘win/lose’ endpoint—only long-term management to avoid hot conflict (especially over Taiwan)
