Odd Lots · · 55 min

The Impossible Trilemma That Hangs Over the Nvidia Crisis

With Tian Yang

Summary

Tian Yang argues that the AI buildout, Nvidia's circular financing deals, and US-China chip competition are all shaped by a trilemma of sovereignty, domestic cohesion and sound money, which he sees as the force behind the current Nvidia-related strain. He expects AI financing to become harder as the Fed hikes and inflation rises, with SoftBank's collateral tied to OpenAI's ability to keep raising. He also argues that guns-and-butter deficits, financial repression tools such as stablecoin yield limits and bill issuance, and the Iran war point toward inflation, gold strength, and capital controls, with China's household consumption sacrificed to strategic technology. Yang concludes that a crisis, rather than voluntary reform or productivity gains, is probably needed before the trade-offs are resolved, and that arms control may eventually follow.

Key takeaways

  • Tian Yang on US deficit & government debt: Guns-and-butter spending without tax rises means sound money gives way, as in the 1960s.
  • Tian Yang on Gold & precious metals: Gold's rally since 2023 warns of fiscal deficits and a breakdown in sound money.
  • Tian Yang on China's economy: China prioritises guns and sound money, suppressing household consumption ('butter').
  • Tian Yang (uncertain on NVIDIA): Nvidia's circular financing reflects AI ecosystem sprinting for funding, harder as rates rise.
  • Tian Yang (uncertain on SoftBank Group): SoftBank's collateral depends on OpenAI's next private fundraising round succeeding.

What was said, by company and theme

30 views from 1 speaker. Every quote is checked against the episode's transcript (how).

NVIDIA

Across the shows: 146 speakers in other episodes, 131 bullish and 110 bearish calls, past 30 days

Where everyone stands on NVIDIA

  • Tian Yang · Variant PerceptionUncertain

    Nvidia's circular financing reflects AI ecosystem sprinting for funding, harder as rates rise.

    Yang frames AI buildout financing as a marathon in which every marginal fundraise becomes a sprint. He reads Nvidia's circular financing deals as a sign of that sprint: the ecosystem lacks time to bring in outside investors, so it uses inverted balance-sheet arrangements. The Fed has hiked and inflation has risen, so each sprint is getting harder against higher rates. The structural AI theme still holds, but the financing can keep working only with difficulty.

    “And in a way, if you look at the circular financing around Nvidia, I think it's because they're sprinting. You don't have time to bring people in.”at 26:57

SoftBank Group

Across the shows: 5 speakers in other episodes, 0 bullish and 6 bearish calls, past 30 days

Where everyone stands on SoftBank Group

  • Tian Yang · Variant PerceptionUncertain

    SoftBank's collateral depends on OpenAI's next private fundraising round succeeding.

    Yang describes a two-way risk in AI financing. A strong IPO or a successful OpenAI private round would make SoftBank's collateral 'money good' again and make pre-commitments real money. The obverse is that SoftBank's collateral depends on OpenAI's ability to keep raising, so it is exposed if that funding stalls. He presents this as a mechanism rather than a firm call on the outcome.

    “Then in two weeks, OpenAI announces, hey. We close our next private round. Yeah. Suddenly SoftBank's collateral is money good again.”at 30:06

US–China AI & chip race

Where everyone stands on US–China AI & chip race

  • Tian Yang · Variant Perception

    Huang's exit reflects sovereign demand for total trust in strategic AI assets, Oppenheimer-style.

    Yang draws an analogy to Oppenheimer, a central figure in building the atomic bomb who later came under suspicion over his loyalty. He argues that in the sovereignty lens, anyone at the centre of a strategic technology must have total trust from the state. The Jensen Huang headlines are one instance of that pressure. Yang says he does not know what was disclosed, but the pressure is real at the margin.

    “You can see this overarching lens of sovereignty today. We don't know what [Jensen] divulge or not, but at the margin, these are the pressures that come through. There has to be total trust from the sovereign to strategic assets.”at 9:00
  • Tian Yang · Variant Perception

    Whoever leads in AI will dominate rivals, splitting the world into blocs governments must back.

    Yang says that if you believe more compute yields smarter models, you must back the largest players and put substantial resources behind them. The reason is that whoever wins and has the best AI will be able to dominate other countries. The world would then split, and every other country would have to choose a side. So the AI race is backed by governments and has a large addressable market.

    “If that's true, then you have to back the big guys and put a lot of resources behind it. Because clearly, whoever wins and has the best AI will hack the other country. Right? And then the world will split.”at 32:45
  • Tian Yang · Variant Perception

    Escalating AI rivalry will create physical constraints forcing an arms-control accord, as in the Cold War.

    Yang's most likely circuit breaker for the trilemma is arms control. He expects the US and China to keep escalating until physical constraints on both sides, such as compute and energy limits, force them to acknowledge the need for an agreement. He compares this to the Cold War arms-control process.

    “It's probably the arms control analog that we keep escalating to a point where you create so much physical constraint both here and in China that there's acknowledging, okay, we have to do something about it.”at 38:53
  • Tian Yang · Variant Perception

    Security competition will keep a floor under AI spending regardless of market returns; no US-China rapprochement.

    Yang agrees that the US and China will keep investing in AI for national security and geopolitical reasons, even where the investment is not economic on market terms. He says markets tend to underweight this. He cites the long history of US-China tech conflict, from the Huawei sanctions onward, as evidence that countries think very long term, and he sees little sign of rapprochement between the two.

    “I think that's the angle markets typically don't tend to know, give a lot of weight to... the countries need to think very long term.”at 39:58

US–China relations

Where everyone stands on US–China relations

  • Tian Yang · Variant Perception

    Both powers pursue self-sufficiency, as China's decade-long push followed Western blocks on its tech and food access.

    Yang argues that sovereignty means being able to say no to rivals, which requires not being dependent on them. He traces China's push for self-sufficiency back to the ChemChina–Syngenta deal and its concern about dependence on Western seed and food technology. He notes that the West blocked acquisitions and later imposed ZTE and Huawei sanctions, which revealed acute vulnerability on both sides. The result is a decade-long mobilisation towards self-sufficiency in tech, food and energy.

    “But if you actually look at China, this has been like a decade long process... China was worried, hey. They were dependent on the West for food.”at 10:06
  • Tian Yang · Variant Perception

    Capital controls and export bans are creeping in and are almost certainly negative for asset markets.

    Yang argues that if governments need to finance debt through financial repression, they cannot let money flow out freely. Capital controls are the uncomfortable outcome for asset owners. He points to a slow creep already visible: China and Russia have banned diesel exports, and Trump has floated similar measures in the US. Other countries already impose capital controls, and he sees the trend as moving away from free movement of money and goods, which is almost certainly bad for markets.

    “the angle maybe we don't talk about as much that's gonna be uncomfortable for asset owners is the potential for capital controls.”at 18:51
  • Tian Yang · Variant Perception

    Geopolitical shocks will likely be inflationary and push toward capital controls that distort valuations.

    Yang argues that sovereign-driven constraints are one-off, hard-cutoff events that markets have not had to price for a century. A US investor in China faces the risk that capital is locked up, repatriation is blocked or an investigation begins. He thinks each such shock is likely inflationary and pushes toward capital controls that distort valuations, so portfolio resilience depends on understanding the direction of these shocks.

    “the direction of these shocks is always gonna be likely inflationary and likely towards some kind of capital control type thing that distorts valuations.”at 43:14

Frontier AI labs & model race

Where everyone stands on Frontier AI labs & model race

  • Tian Yang · Variant Perception

    Evidence so far says more compute yields smarter models, so top frontier labs deserve backing.

    Yang treats the claim that adding compute produces smarter models as the key assumption behind the AI buildout. He calls it an open question, but says the evidence so far supports it. If it holds, the top two frontier labs, which can show models improving with compute, will attract backing. Marginal parts of the AI ecosystem are less likely to get government financing.

    “I would say that's an open question. The evidence so far is still yes. If that's true, then you have to back the big guys and put a lot of resources behind it.”at 32:45
  • Tian Yang · Variant Perception

    Under sovereign-backed AI, domestic suppliers win big while sub-scale frontier labs without compute are poor bets.

    If AI investment becomes sovereign-backed rather than market-driven, the gains concentrate. Domestic critical-resource providers and domestic equipment makers move to the front of the line and could become monopoly suppliers. By contrast, a frontier AI lab that lacks enough compute to keep up is a bad investment. Yang sees this as a sharper split between winners and losers than in the rising-tide phase, and he points to China's markets as the example.

    “But if we move to a more sovereign backed investment AI... domestic equipment makers, they're front of the line. They might become even the monopoly supplier. But if you bought... the fifth frontier AI lab that doesn't have enough compute, I don't think that's a good investment.”at 40:40
  • Tian Yang · Variant Perception

    Evidence so far shows more compute yields smarter models; a slowdown would signal the spending cycle ending.

    Yang's key assumption is whether more compute produces a much smarter model. Two years ago many expected returns to compute to flatten, but the evidence since then has shown that adding compute makes models much smarter. He says the sign that the cycle is turning would be evidence of that trend slowing or stopping, which would indicate that AI spending is nearing an end.

    “But the problem is we've had a lot of evidence. So you put a lot more compute, it gets a lot more smarter. When does that end?”at 48:46

US deficit & government debt

Where everyone stands on US deficit & government debt

  • Tian Yang · Variant Perception

    Guns-and-butter spending without tax rises means sound money gives way, as in the 1960s.

    Yang argues that when a country is committed to both military and social spending, the leg that gives way is historically sound money. He compares this to LBJ's Great Society and Vietnam War. LBJ said the country could pay for both without tax rises, the deficit grew, and inflation pressure built until the Fed chair was pressed to hold rates down. The 1968 gold-pool break followed. Yang sees the same shape today.

    “Then the tension becomes, if you're committed to your guns spending and you're committed to your butter spending, then history shows that typically the leg that gives way is sound money.”at 11:18
  • Tian Yang · Variant Perception

    Financial repression with 4-5% inflation and suppressed rates can amortise away US debt.

    Yang sets out the mechanism for dealing with heavy debt through financial repression. Authorities run inflation of about 4-5% and suppress interest rates, which erodes the real value of the debt. He notes the US has its own sovereign currency, unlike France, so the tool is available to it. The approach works mechanically, which is why he expects it to be used.

    “the base idea is, you know, run inflation of four to 5%. Right? Trying to suppress interest rates. You... Or amortize away the debt.”at 18:51

Gold & precious metals

Where everyone stands on Gold & precious metals

  • Tian Yang · Variant Perception

    Gold's rally since 2023 warns of fiscal deficits and a breakdown in sound money.

    Yang draws a parallel to 1968, when gold's peg broke as inflationary policy spread. He sees the 2023 gold rally against a background of persistent fiscal-deficit concern as a similar warning. Governments cannot cut entitlements or butter spending while mobilising for tech and military competition, so the market is signalling that something has to give.

    “If you go back to 2023, we've had this gold rally, right, come through. You've had this persistent concern about fiscal deficits, I. E. You can't cut the butter, you can't cut entitlement, yet we are in this mobilization.”at 13:15
  • Tian Yang · Variant Perception

    Gold's resilience signals underlying pressures; a structural breakout may take a while despite cyclical rate headwinds.

    Cyclically, rising interest rates are the biggest headwind to gold, since higher rates raise the opportunity cost of holding it. Structurally, Yang sees a long-cycle analog in which gold has counter-trend moves. Gold holding up despite higher rates is evidence of underlying pressures, but he thinks a truly explosive move may take a while to arrive.

    “But the fact that the gold has behaved is telling you there's underlying pressures. It may take a while before that truly explodes.”at 38:00

China's economy

Where everyone stands on China's economy

  • Tian Yang · Variant Perception

    China prioritises guns and sound money, suppressing household consumption ('butter').

    Yang argues China faces the same trilemma of sovereignty, domestic cohesion and sound money as the US. Its choice is different: it keeps the military and sound money, which means suppressing household consumption. So household 'butter' is what gives way in China. This is a deliberate trade-off, not a policy failure.

    “China has the same trilemma. But their choice is sound money and guns, and then they suppress the households. They're suppressing butter.”at 24:33
  • Tian Yang · Variant Perception

    Chinese consumer tech is out of step with Beijing, which is suppressing consumption in favour of sovereign priorities.

    Yang says China has been suppressing consumption, which he calls 'butter', to prioritise sovereign and strategic technology. Consumer tech companies are therefore not aligned with the state's priorities. He ties this to the weakness of the Chinese market and the investor lesson that you need to be aligned with the sovereign.

    “They've been suppressing butter. So if you own consumer tech, it's not exactly aligned to the sovereign.”at 42:11

AI: overall outlook

Where everyone stands on AI: overall outlook

  • Tian Yang · Variant Perception

    AI buildout remains a secular story, but its financing is now tied to macro interest rates.

    Yang argues that non-AI parts of the economy show a classic macro cycle, with fast-cycle raise growth slowing. AI has been the secular trend. What connects the two now is AI's financing need, which depends on macro interest rates and the current position of the cycle. So the market is bifurcated, but the AI build-out thesis itself has not changed.

    “But AI has been the secular thing. What connects them now is this financing need. And the financing need is ... a function of macro interest rates and where we are.”at 31:07
  • Tian Yang · Variant Perception

    AI market-size assumptions look optimistic once governments mandate local sourcing.

    Yang says some optimistic total-addressable-market assumptions for AI are likely wrong. Through a sovereignty lens, the question is where value leaks. If governments finance the AI buildout, they may attach conditions such as requiring purchases from local suppliers. That would keep stimulus at home and remove foreign suppliers from the market, so the addressable market for those suppliers shrinks.

    “I think clearly there's some optimistic TAMs for the markets. But I would say, if you put a sovereignty lens on, there's a lot of assumptions about which of the markets and where things are gonna leak.”at 31:54

US inflation

Where everyone stands on US inflation

  • Tian Yang · Variant Perception

    Iran war and simultaneous physical constraints are turning into a major US inflation problem.

    Yang lists the forces combining at once: the fiscal guns-and-butter pressure, the Iran war, and physical supply constraints. Together these push up prices and turn into a big inflation problem that policymakers will have to confront. He points to the hot CPI report that week as evidence that hopes for a downward glide path are fading.

    “And finally, you have a Iran war. You have these things that force a lot of simultaneous physical constraints to hit, and then suddenly becomes like a big inflation problem, and we'll have to confront it.”at 13:30

Stablecoins & crypto policy

Where everyone stands on Stablecoins & crypto policy

  • Tian Yang · Variant Perception

    Stablecoins barred from paying interest act as a financial-repression tool funnelling demand to Treasuries.

    Yang compares the stablecoin rule that issuers cannot pay interest to Regulation Q in the 1960s, which capped the interest banks could pay. Both suppress what savers earn and steer funds elsewhere. Tracy Alloway notes stablecoins are a large source of demand for Treasuries. Yang sees it as a hint of financial repression, a tool that could be used to finance government debt.

    “So in a sense, if you look at stablecoins, it's a, hey. You can't pay interest. You have hints of that.”at 20:38

Banks

Where everyone stands on Banks

  • Tian Yang · Variant Perception

    SLR reform and bill issuance are tools already in place to channel demand into government debt.

    Yang lists tools of financial repression that are already in place. Supplementary leverage ratio reform frees up bank balance sheets. The Treasury is issuing more bills rather than long-duration bonds, echoing Operation Twist in the 1960s. Together these can channel bank and other demand into government debt, and the tools could be used further if politics shifts after an election.

    “You mentioned, obviously, a SLR reform, right, freeing up bank balance sheets. So there are... At least the tools are in place.”at 21:10

US elections & politics

Where everyone stands on US elections & politics

  • Tian Yang · Variant Perception

    Polarisation makes sweeping fiscal reform hard; it usually needs a crisis first, as with Reagan and Thatcher.

    Yang points to deficit-focused reform efforts such as the DOGE initiative, the America Party, and Ross Perot's third-party run. None has yet produced real change. Polarisation makes sweeping bipartisan fiscal reform very hard. Reagan and Thatcher did implement reform, but only after the crises of the 1970s. So a crisis is usually needed before the population will accept the pain.

    “But the problem is you usually need a big crisis first Yeah. For that to actually be implemented.”at 22:39

Wealth, inequality & billionaires

Where everyone stands on Wealth, inequality & billionaires

  • Tian Yang · Variant Perception

    Elite stake-giving pledges like SpaceX's and Anthropic's are marginal signs, not enough redistribution.

    Drawing on Peter Turchin's long-cycle work, Yang says a peaceful reform requires economic elites to voluntarily give something up to support the bottom. He points to Elon Musk talking about giving away SpaceX shares and to AI companies talking about giving the public a stake. He calls these token signs. The redistribution needed is far larger, so he thinks a crisis is probably needed first.

    “You can see, you know, there's token signs. Right? Elon talking about giving away SpaceX shares. Right? We've had anthropica companies talking about giving the public a stake. Though I think those are the marginal signs.”at 23:49

US rates & the Fed

Where everyone stands on US rates & the Fed

  • Tian Yang · Variant Perception

    Fed under new chair could mirror Martin's 1960s playbook: hike rates, tame inflation, then normalize.

    Yang draws an analogy to Fed chair William McChesney Martin. Martin adjusted at the margin and hiked rates to bring inflation down, which caused a bit of a credit crisis before things normalized. Yang argues policy is still moving in that direction, so the new chair could follow the same path. This is an analogy, so the conviction is moderate.

    “[Warsh] could be this generation's Martin, right, where at the margin, you do adjust. He did hike rates back then. Inflation comes down.”at 37:12

Middle East & Iran

Where everyone stands on Middle East & Iran

  • Tian Yang · Variant Perception

    A prolonged Iran war could act as a catalyst, crowding out spending and forcing a populist response.

    Yang wrote his report because the Iran war looks like a catalyst that forces the trilemma's trade-offs. Wars are hard to stop once they start, and they are an extreme version of what the sovereign needs. If the war is dragged out, it creates sovereign crowding-out, which combined with AI spending forces either a populist response or extraction from the population.

    “the Iran war started giving me the vibes that it's acting as a kind of a catalyst... if we drag this out, then it's gonna have these crowding out impacts.”at 43:33

AI & jobs

Where everyone stands on AI & jobs

  • Tian Yang · Variant Perception

    A productivity boom would relieve debt but delay reform, as historical gold-flood analogs suggest.

    Yang argues that a productivity boom, the wild card many are relying on, first releases debt constraints but kicks the can down the road. He cites Spain and Portugal in the age of exploration, where a flood of gold and silver eased debt but amplified existing institutions and delayed reform, drawing on Acemoglu's work. Nick Bloom's research on electrification suggests productivity gains of only about 1.5 to 2% of GDP a year.

    “Obviously, the first order impact is it releases your debt constraints. It kicks the can down the road... But when you have it, it doesn't force you to reform.”at 45:27

Manufacturing & industrial policy

Where everyone stands on Manufacturing & industrial policy

  • Tian Yang · Variant Perception

    Sovereign mobilization is inflationary at first, then overbuilds duplicate capacity and turns deflationary.

    Yang describes the endgame of state-led mobilization. Countries duplicate everything, and the early phase is inflationary because everything is bottlenecked. Domestic problems then set in, arms control eventually comes, and the economy overbuilds. That overcapacity becomes very deflationary at the end. He says the current phase is still in the ramp-up, which is why inflation persists.

    “in the original phase of mobilization, we're essentially doing duplicates of everything. So initial phase is inflationary, everything's bottlenecked... That's gonna suddenly become very deflationary at the end.”at 46:25

US Treasuries & yields

Where everyone stands on US Treasuries & yields

  • Tian Yang · Variant Perception

    Long-end yields are hard to trade: AI is long-term deflationary but geopolitical shocks keep adding inflation.

    Yang's view is that AI is secularly deflationary, since the evidence suggests it lowers costs over the long term. However, every geopolitical and domestic shock adds inflationary pressure against that trend. The tug-of-war between the two is why he thinks the market will struggle to trade long-dated bond yields.

    “I think the evidence probably suggests AI in the long term is, I think, is deflationary. But the problem is every geopolitical shock, domestic shock is gonna be inflationary against that.”at 47:19

Across the shows

What other episodes Chorus covered said about the same companies in the 30 days up to this one.

  • NVIDIA146 speakers · 131 bullish, 110 bearish calls
  • SoftBank Group5 speakers · 0 bullish, 6 bearish calls

Topics

  • AI addressable market
  • AI arms control
  • AI arms race
  • AI buildout thesis
  • AI financing
  • AI leadership and national security
  • AI productivity and debt relief
  • AI scaling returns
  • AI spending floor from security competition
  • China consumer tech vs state priorities
  • Fed policy path
  • Iran war as a catalyst for the trilemma
  • OpenAI funding and SoftBank collateral
  • bank balance sheets and Treasury demand
  • capital controls and export bans
  • capital controls and foreign investment risk
  • compute scaling
  • decoupling and self-sufficiency
  • elite redistribution
  • financial repression
  • fiscal reform politics
  • gold as sound-money signal
  • gold price drivers
  • guns and butter deficits
  • household consumption
  • inflation drivers
  • long-end yield outlook
  • overbuilding under sovereign mobilization
  • stablecoin yield and Treasury demand
  • winners and losers of sovereign AI investment

More from Odd Lots

All Odd Lots episodes

Ask your AI about this episode

Chorus gives Claude, ChatGPT and other assistants every view from this episode and every other show Chorus follows. Ask who agrees, who changed their mind, and what else was said about the same companies.

Setup guide

A weekly email with the calls that mattered from the most important voices in investing. Unsubscribe any time. Privacy Policy

Chorus reports what speakers said. It isn't investment advice. How Chorus works · About Chorus