China, AI Immigration, Rare Earths & Chips, Tariffs, Markets | BG2 w/ Bill Gurley & Brad Gerstner

Summary

• The dot-com lesson for AI: Gerstner, prepping his HBS 25th-reunion talk, ran the split-adjusted math — Amazon is up ~440X from its 2000 high and ~1,800X from the low, and the Nasdaq is still up 5X from the bubble peak. “We overestimated what it was going to do in the short term… but dramatically underestimated the long term” — and AI’s short term is already running faster than the internet’s ever did.
• Data walls are going up in hyper speed: Reddit sued Anthropic despite a deal, Anthropic cut Windsurf off its models, and Salesforce changed its ToS so you can query but not train on your own CRM and Slack data. Gerstner predicts enterprise apps will have to declare themselves “open data or closed data” — and challengers will declare open and steal customers. With OpenAI hitting 400B annual searches 8 years faster than Google, the TripAdvisor-in-Google’s-underbelly dynamic is replaying in fast-forward.
• China’s success is Darwinian, not Soviet: Beijing seeds 500–1,000 competitors per strategic industry at the provincial level and lets markets whittle them down — producing $130/car solid-state lidar vs Waymo’s $5,000, and 500+ EV startups vs a US handful. Corollary for investors: China deprioritizes market cap — BYD’s 30% price cut may have been state-encouraged, but Gurley says he has no proof — so betting on Chinese Mag-7-style $3T winners may misread “the objective function of what a win is.”
• The visa 180 is a self-inflicted wound: Rubio’s move to revoke visas of Chinese students “in critical fields” reverses Trump’s All-In promise to staple green cards to diplomas, at a moment when ~40–50% of top US AI researchers are of Chinese origin. Gurley warns of “a McCarthy-like perspective that could be very dangerous to our long-term competitiveness”; Gerstner: build the championship team — “we should not care where in the world the basketball player comes from.”
• Gerstner’s trade proposal ahead of the Trump–Xi call: rare earths for the deprecated Blackwell 30 — no HBM, no CoWoS, degraded for training but keeps China’s researchers and developers in CUDA, avoids accelerating Huawei, generates billions for the US and Nvidia, and buys years to stand up domestic rare-earth supply. The tail risk of the ban: if AGI is existential and China is locked out, “we just have to go take Taiwan.”
• Market call: land the plane or give back 10–15%. Nasdaq +20% off the lows makes sense if blended tariffs settle ~10–15% and the reconciliation bill passes (failure = a $4T tax increase); core PCE supports two Fed cuts in the back half, and 10-year hand-wringing ignores that yields have sat in a 4–5% range for two years. “Don’t take yourself out of the game” — but the proof is in the pudding.
• Governance lightning round: Gurley’s Delaware warning is landing — high-profile and private companies are reconsidering incorporation after evidence of 66x lawyer fee multipliers — and the ISS/Glass Lewis proxy duopoly (97% share, >80% foreign-owned) is an open incubation target: “reach out to us. We could help fund it.”

Deep dive

1. 25 years after the bubble: short term overestimated, long term dramatically underestimated

• Gerstner’s reunion-slide shocker: Amazon peaked at $243 in 1998, sat at $150 at the start of 2000 when Blodget called $400 — top-ticking it before the fall to $26 and public ridicule. But split-adjusted, that ridiculed 2000 high equals ~47 cents today: Amazon is up ~440X from the high, ~1,800X from the low, and the Nasdaq is still up 5X from the 2000 peak, 10X from the trough. Gurley’s detail: it went public at 17 and broke issue for two months.
• The frame he brings to “the dawn of the age of AI”: “we overestimated what it was going to do in the short term” — slow broadband, terrorist attacks in 2001, recession — “but what’s probably even more surprising is how dramatically we underestimated the long term. It blew away all of our estimates.”

2. Data walls: declare yourself open or closed

• Gerstner’s strategic flag from just the past week: Reddit sued Anthropic this morning despite having a deal; Anthropic cut Windsurf off all its models after the OpenAI acquisition; and Salesforce changed its terms of service so CRM and Slack data can be queried via an MCP connector but not trained on. Gerstner’s reaction: “I can’t fathom how upset I would be if I were paying a seven, eight-figure license to Salesforce and they told me I couldn’t train on my own data.”
• Gerstner’s prediction: enterprise apps will split into “open data or closed data,” and Salesforce competitors will immediately declare open to steal customers — “battle lines are being drawn.” Meanwhile Salesforce will surely train on it themselves and sell you their agentic view, “which is why they’re doing this.”
• Gerstner’s speed point: this is the Zillow-board Google lesson — “you can’t build a brand in the underbelly of Google,” TripAdvisor went $20B → $1B when Google took recommendations back — except compressed: OpenAI reached 400B annual searches 8 years faster than Google, over a billion a day. Hence Windsurf and Cursor building their own models as open-model dependencies reorder.
• Both are “loud and proud on the side of keeping this all open”: Deep Research now reaches GitHub, Google Docs, Gmail, Outlook, HubSpot, Dropbox via MCP; the open question is whether Google weaponizes its phone/Gmail/office stack and walls OpenAI out. “At the end of the day, it should be my data.”

3. Why China wins: 500 competitors first, Darwin does the rest

• Gurley had ChatGPT (version 4.5, “the $200 a month version”) produce the research, which they’ll post. Core finding: in industries China wants to win, the state ensures 500–1,000 competitors exist, then lets the market whittle them down — not the Soviet single-state-champion model. Gerstner’s contrast: Soviet state actors were inefficient, full of grift, and bankrupted themselves; China’s products are globally competitive, “so there has to be something systematic and structurally different.”
• The side benefits per Gurley: massive optionality — Chinese solid-state MEMS lidar is priced ~$130 a car vs Waymo’s ~$5,000 because 500 startups tried different approaches — and a far more robust supply chain, with 50–100 competitors birthed at each step. Over 500 EV startups in China vs “four legitimate ones” in the US. Seeding happens at the provincial level; subsidies (BYD’s ~$2B) come after winners emerge — part two of the research covers that criticism head-on.
• No US equivalent is known, he concedes: American government involvement looks like Solyndra — “regulatory capture… at the end of the game. This is at the beginning of the game.” And on the IP-theft narrative: “if you narrow it down to one derogatory action, you’re ignoring this system… you could make a bad policy decision because you think if we just protect IP then this won’t keep happening.”

4. Beijing doesn’t care about your market cap — reprice Chinese equities accordingly

• Part three’s finding: a deep deprioritization of market cap of successful companies. BYD took prices down 30% in the past three weeks — “that may have had encouragement from the government. I don’t have proof of that, but it looks synonymous with the kinds of things that matter to them.” If your government optimizes for employment and durable global competitiveness, you encourage “the Amazon approach… your margin is my opportunity.”
• The investor warning: if you’re buying Chinese stocks “hoping they turn into these three-trillion-dollar entities, the Chinese government may not consider that part of the objective function of what a win is” — Gerstner adds it belongs in any US-China relative-valuation comparison. Once winners emerge the hand gets heavy: golden shares, preferential procurement, a disappeared Jack Ma, ByteDance still private.
• Gurley wouldn’t be shocked if the four deep-pocketed Chinese open-source AI players are the same playbook as EVs and solar — plus one more objective function: affordability to the citizenry. “BYD selling a car for 10 grand is better for the consumer in China, as might be four open-source AI models.”

5. The visa 180: “the best way to stay ahead of China is to poach their talent”

• The whiplash: Trump on All-In promised green cards stapled to diplomas — “that includes junior colleges too… that is going to end on day one” — then Rubio tweeted the US will revoke visas of Chinese students with CCP connections “or studying in critical fields.” AI is obviously critical; both hosts read it as a 180 and hope it was “misinterpreted, that it’s very narrow.”
• Gurley’s slippery-slope warning: he’s “all for not having spies,” but the next step is “studying LinkedIn for every single AI company for anyone of Chinese origin… a McCarthy-like perspective that could be very dangerous to our long-term competitiveness.” The stakes: ~50% of AI researchers are of Chinese origin, China’s AI patent count now exceeds the US, and skilled immigration stuck at 200–250K/year for 20 years “is insanity.” The Manhattan Project itself was heavily immigrant-built.
• Gerstner’s brand-USA framing: it’s not just students already here but the next generation deciding whether America is “capricious and can just change on a dime” or the land of opportunity — and his championship test: “we should not care where in the world the basketball player comes from… we ought to take the same approach to AI.” A few weeks ago they proposed an AI visa for Chinese researchers’ families to relieve pressure to return.
• The policy-backfire through-line: “the export controls on China probably is what caused Huawei to catch up so quickly,” the Biden diffusion rule would have handed China the global AI stack, and — per Sacks — decelerationists are now “positioning themselves as China hawks so they can gather a bigger alliance.”

6. The trade Gerstner would make Friday: rare earths for the deprecated Blackwell 30

• Backdrop: WSJ reports China playing tough on rare-earth magnet exports, with carmakers risking factory shutdowns — the “kill shot” Gerstner flagged earlier this year, since China is the global monopolist in magnets. His parallel: both sides view their chokepoint as existential — chips for China, magnets for the US military and auto industry. His proposal ahead of the Trump–Xi call: trade rare-earth access for the deprecated Blackwell 30.
• The four-point case: the Blackwell 30 has no HBM and no TSMC CoWoS — competitive in-market but degraded for training — keeping half of the world’s researchers and developers who are in China in CUDA, allowing Nvidia to compete and avoiding acceleration of Huawei’s ecosystem; it generates billions in taxes, cuts the trade deficit, and funds Nvidia’s frontier lead (“a proxy for the US staying in front”); it secures rare earths now while both sides spend years unwinding dependencies; and the ban raises Taiwan risk — “if that is so existential as a risk to our country, we just have to go take Taiwan” — hugely risky when the US needs 3–4 years to move its supply chain.
• Gurley’s three reactions, all in agreement: economics 101 — “if you take our best stuff off the table, they’re not going to buy our crap,” so trade balance is unreachable; the Taiwan point (he and Jeffrey Sachs have made it); and telling Europe and South America they can’t buy Huawei is beyond what our government can enforce — he expects ASML to just ignore us. “I worry that rather than build a wall around China, we’re going to build a wall around America.”
• The endgame math: ~0% of leading-edge chips are fabbed outside Taiwan today, headed to 15–20% US capacity by 2030 — and a UAE–TSMC joint fab with US sovereign influence could push US-influenced advanced-node share to 40–50% in 4 years. “But we’re not going to do that if we’re in a war over Taiwan.” Credit to Sacks and Lutnick for repealing the diffusion rule; the Blackwell 30 is “a closer call” he hopes they take.

7. Market check: land tariffs and the bill, or give back the bounce

• Gurley’s setup — “as much uncertainty as there’s ever been this year” — vs Gerstner: “the market’s clearly not agreeing with you.” Nasdaq +20% from intraday lows, the likely Russell 2000 ~+1% YTD, S&P +2%; the bounce fits his early-May flight-path change on the “Bessent consensus.” Two things must land: blended global tariffs ~10–15% (“hundreds of billions, not trillions” — China is “the big enchilada”), and the reconciliation bill — failure means a $4T tax increase and markets a lot lower; passage adds ~$300–400B of stimulus via no tax on tips/overtime and the Social Security deduction, driving back-half acceleration.
• On Elon opposing the “big beautiful bill”: the debt worry is legitimate, but “you can’t just cut $2 trillion in a single year — that would be an 800 basis point headwind to GDP,” recession-if-not-depression territory. He wants Bessent to lay out the multi-year “3-3-3” path (3% growth, “a debt-to-GDP ratio of 3%” as he put it, per Dalio’s piece); discretionary cuts come via the rescission act just sent to Congress, not reconciliation.
• Rates hand-wringing overdone: the 10-year’s move from 4.2 to 4.4–4.5 sits inside a 4–5% range that’s held for two years — recall Summers calling 7% in late 2022. Core PCE came in below expectations, the Fed admits it’s restrictive, and the market prices two cuts in the back half. His asymmetry: sign the bill, land China, get the cuts, and “this market’s going to be a lot higher… don’t take yourself out of the game” — miss, and it’s back down 10–15%.
• Probability of a China deal: “pretty high” — his door-one/door-two framing from the crisis stands: “he wrote a book called The Art of the Deal… he does not think he wants to slam the brakes on the global economy,” but “China’s got to step up and be willing to deal as well.”

8. Governance lightning round: exit Delaware, disrupt the proxy duopoly

• Gurley’s Delaware clip drew 2.5M views and a Fortune headline on lawyers winning fee multipliers up to 66x their normal hourly rate. Two sharpenings since: the highest-profile companies face the most risk — “an activist judge is going to want to make an example out of them” — and staying private is no shield: if shares trade in secondaries, a Delaware-incorporated unicorn can be sued, and an activist plaintiff might be “particularly excited” to target one. Delaware’s whole predictability rationale “is no longer true.” Gerstner confirms companies are already asking him about reincorporating.
• On Hagerty’s tweet that ISS and Glass Lewis hold 97% market share and have “hijacked corporate governance”: Gurley reached out to ISS during the SBC episode and found “no one there thinking from a first-principles perspective” about shareholder alignment; per the senator, both firms are now over 80% foreign-owned. The irony he flags: super-voting shares — usually deemed worse governance — exist precisely so these two firms can’t dictate to companies.
• The incubation call: index giants like BlackRock shouldn’t lazily vote the duopoly’s line, and an AI-driven alternative scoring genuine shareholder alignment is buildable — “if you are already working on an alternative to these two companies, reach out to us. We could help fund it.”