The 2018-2025 "cold war," in reverse. Gave frames 2018's US semiconductor embargo as the first deliberate attempt by one economy to "trip up" the second-largest. Rather than breaking China, it triggered a systematic de-Westernization of supply chains — state power and household savings redeployed to move up the industrial value chain. He dates it precisely: "2018 is the start... 2025 is the end." Evidence: the Busan summit, four planned US-China leader meetings within 12-14 months, and Trump inviting Xi to Mar-a-Lago rather than visiting Beijing first.
What the push produced. Manufacturing efficiency — Tesla Shanghai produces roughly twice as many cars per worker at ~$14,000/year labor cost vs. $88,000+ plus ~$20,000 healthcare in California; extends to nuclear plants, high-speed rail, turbines. The domestic price: real estate fell ~a third, equities ~two-thirds — a deliberately accepted deflationary shock that also purged fraud and weak businesses. Rare earths as leverage — the Raytheon CEO said the company couldn't produce missiles within three weeks without them; Ford/GM warned of factory shutdowns within two weeks without magnets. Export dependence, not import dependence — a car contains ~$3,000 of chemical products now largely supplied by Chinese producers rather than DuPont, Dow, or BASF.
Two different AI strategies. US: chasing AGI, "get 500 million customers paying $20 a month," proprietary closed models threatening white-collar service work. China: not chasing AGI — applying AI to factory automation and robotics, releasing models (DeepSeek, Alibaba's Qwen) open-source and free; a claim that ~80% of startups going through Andreessen Horowitz are already building on Chinese open-source models. His framing of the DeepSeek shock: it showed China could be competitive even at the pinnacle of technology despite the embargo. His broader line: "When China enters a room, profits walk out."
Energy as the hidden variable. China produces more electricity than all of Europe, the US, and Canada combined, the product of ~20 years of infrastructure buildout. A Chinese data center pays ~3 cents/kWh versus 7-8 cents in the US (citing the WSJ). He uses a WWII analogy: Nvidia is a "Tiger tank" (best-in-class), but the US's comfort in having the better tank echoes not out-producing the enemy before Normandy — the inferior-but-abundant Sherman tank won on volume.
The US capex bubble he's watching. Oracle raised FY2026 capex guidance to ~$50bn against ~$65bn expected revenue (capex near 75% of revenue), on top of a headline $300bn data-center commitment tied to OpenAI; CDS spreads widened on funding doubts. SoftBank committed ~$30bn to OpenAI but deployed only ~$7.5bn so far; the stock is down ~40% since the announcement. His read: bull markets reward aggressive capex early, then punish overpaying — and the market has already crossed that line; he calls large chunks of the buildout "a huge misallocation of resources." Physical efficiency concern — whether today's ~4-million-square-foot data centers end up oversized within a few years. Market concentration — the S&P 500 was up ~25% on the year, with ~25% of that attributable to Nvidia alone, which fell ~17% on the DeepSeek news.
The "bezzle": late bull-market capital allocation. Gave invokes Galbraith's "bezzle" — the gap between perceived and actual wealth that grows during a prolonged bull market as fraud multiplies. His view: the US is 15 years into a bull market and, aside from the COVID crash (too quick to purge anything), hasn't had a "proper" bear market of the kind that resets behavior. He points to private credit as where this is starting to show. China, by contrast, went through its own reset — reinforced by the Nvidia export ban, which prevented a wasteful capital-spending spree.
US fiscal position and a convergence toward state capitalism. He flags US "twin deficits" at roughly 11% of GDP, meaning the US still needs foreign capital even while leaning harder into industrial policy — structurally low-return-on-capital and prone to corruption risk. He notes the Buffett indicator (market cap to GDP) sitting around 2x, arguing "the stock market is now the tail that wags the dog." His broader claim: the US is "becoming more like" China's state-directed model, not the reverse.
The Latin America pivot. He sees the Trump administration's strategic attention shifting from the Indo-Pacific toward Latin America — a "new Monroe Doctrine" partly to secure commodities and labor and keep China from picking up distressed assets. Argentina: US underwriting on the order of $30-40bn+ in support. Venezuela/Cuba: he frames Venezuela's relationship to Cuba as effectively a reverse takeover, expects US pressure on Venezuela to precipitate a Cuban collapse with knock-on effects on far-left-aligned groups in Colombia and Peru. The bull case: Latin American bond yields have already fallen 200-300bp this year, with another ~300bp of compression expected (Brazil from ~13.5% toward 10%).
Europe and Japan: worse off than the US. China's rise is a bigger problem for Europe and Japan given manufacturing's larger economic share. Their main lever — currency devaluation — is politically constrained in Germany and already underway in Japan even as domestic inflation accelerates. Chinese exporters have displaced Western suppliers in Indonesian trains, a Saudi nuclear bid, and vehicles in South Africa and Brazil at the expense of Volkswagen and Peugeot.
China's own fragility. Not one-sided triumphalism — Chinese consumer and business confidence remain "crushed" after seven years of belt-tightening, exports are the one clear bright spot, and China has little incentive to trigger a US market crash since that wouldn't help rebuild its own domestic confidence. A compliance culture among Chinese business leaders persists, using the 2019 action against Jack Ma as the precedent.
