Ray Dalio: AI Is Eating Everything — and It Might Eat Itself

Ray Dalio isn’t here to talk about AI products or model capabilities. He’s here to explain why the entire financial substrate on which AI companies depend may be cracking, and why AI’s economic logic might collapse under global competition before the technology itself fails. His framing: “AI is eating everything and it might eat itself.”

The Episode

Dalio joins the All-In Podcast for the third time, hosted by David Friedberg. A year ago he warned that the US deficit needed to fall to 3% of GDP to stabilize the debt cycle. It hasn’t. The CBO now estimates 6%. Gold has climbed from $2,900 to $5,200 per ounce, validating his thesis in real time. This 49-minute conversation covers the five forces driving the global order, why DOGE’s reforms were structurally doomed, the monetary nature of gold versus Bitcoin, interest rate dilemmas in a K-shaped economy, the misunderstanding of tariffs, America’s descent into “Stage Five” of historical cycles, and why AI companies may not survive even as AI technology thrives.

The US Government as a Failing Company

Dalio frames the federal government’s finances with corporate clarity. It spends roughly $7 trillion per year and takes in about $5 trillion: a 40% operating deficit. Cumulative debt stands at 600% of annual revenue. Half of the $2 trillion annual deficit is just interest payments. The government must roll over $9 trillion in maturing debt while financing $2 trillion more.

About a third of debt buyers are foreign holders, and the situation is deteriorating on their end too. Dollar-denominated debt is already an oversized share of foreign portfolios. Geopolitical tensions raise the specter of sanctions disrupting debt service. Dalio draws the explicit parallel: “You could imagine that with China. You could imagine that with Europe even.” Europeans now wonder whether US sanctions could freeze their own holdings.

The 3% deficit-to-GDP target he proposed last year was a stabilization floor, not a cure. At 6%, the US is running at double that floor, and the political system shows no path to closing the gap.

Why DOGE Was Doomed

When Friedberg asks whether DOGE failed because of wrong actions or structural impossibility, Dalio leans firmly toward the latter. Reforming a massive, inefficient government requires executive authority that moves fast. But democracy demands public approval on short electoral timelines. Cutting programs like school lunch spending triggers immediate backlash, and in a society where “no matter what you do, you’re criticized and torn down,” the political math doesn’t work.

“Structurally a little difficult at this stage? What an understatement.”

He’s not blaming Musk. He’s saying the cycle itself makes reform nearly impossible. The Minnesota daycare fraud case, where billions flowed to nonexistent facilities, exemplifies what happens in a system too large, too complex, and too politically constrained to self-correct. The federal workforce shrank by 317,000 (roughly 14%), but whether those people become productive private-sector workers or get absorbed into other government-adjacent roles remains an open question.

Gold Is Money, Not Speculation

Dalio pushes back hard on the common framing of gold as a speculative asset. Gold, he argues, is the second-largest reserve currency held by central banks. It’s money.

The reasoning is structural. Money is fundamentally debt: holding currency means holding someone’s promise to deliver purchasing power. When governments accumulate too much debt, central banks print money to service it. This has been the fate of every fiat currency. Gold stands apart because it can be transferred between nations, can’t be printed, and doesn’t depend on a third party’s promise.

He draws a critical distinction between wealth and money. Wealth exists in physical assets (buildings, companies) but can’t be spent directly. To spend it, you must convert it to money by selling assets. Right now, global wealth relative to money is massively out of balance. Any large-scale wealth-to-cash conversion, whether from a bubble bursting or wealth taxes, would trigger a liquidity crisis because there isn’t enough “money” to absorb the selling.

Gold allocation across central banks has moved from “extremely small” to “a less small number,” approaching but not reaching the historical average. Dalio’s recommendation: 5-15% of a portfolio in gold, derived from optimization analysis, not a directional bet.

“Gold is not a precious metal that’s speculated on. It is the second largest reserve currency that central banks hold.”

Bitcoin’s Institutional Ceiling

Over the past year, gold rose 80% while Bitcoin fell 25%. Dalio explains the divergence through institutional, not technical, logic.

Bitcoin transactions lack privacy and can be monitored and indirectly controlled. Central banks won’t hold it as reserves, which caps its ceiling as a monetary instrument. Quantum computing introduces additional uncertainty. And Bitcoin holders tend to overlap with tech stock investors, creating correlated sell-offs: when one asset gets squeezed, the other follows.

The market is also relatively small and controllable. Dalio’s verdict is blunt:

“There is only one gold.”

The Interest Rate Trap and the K Economy

Interest rate policy faces an almost impossible balancing act. Rates too low encourage borrowing and fuel bubbles (good for debtors, bad for creditors). Rates too high squeeze debtors who can’t service their debt. With massive debt on both sides, the margin for error has nearly vanished.

The “K economy” makes it worse. At the top, people debate who becomes the first trillionaire. At the bottom, 60% of Americans read below a sixth-grade level. AI is replacing lower-skill labor, widening the gap further. Setting a single interest rate for what amounts to two separate economies is a nearly unsolvable problem.

The Fed is currently shortening debt maturities to hold down long-term rates, but this increases rollover risk. On new Fed Chair Kevin Warsh, Dalio says he’s a practical man who understands both sides, but calls it “a very, very big challenge.”

“One man’s debts are another man’s assets.”

Tariffs: What Economists Got Wrong

Dalio offers a contrarian reframing. Economists, he argues, make a fundamental error by excluding taxes from inflation calculations. If your taxes go up, that’s money leaving your pocket, functionally identical to rising housing costs. Why shouldn’t it count?

Historically, tariffs were the primary revenue source for most governments throughout most of history. They remain, in Dalio’s view, “a totally valid way of raising money.” The strategic case is stronger: the US has hollowed out its manufacturing base and middle class, trade deficits are unsustainable, and in a world of escalating great-power competition, economic dependence equals strategic vulnerability.

“Everybody’s threatening to cut off everything.”

However, Trump’s proposal at the State of the Union to replace income tax entirely with tariffs is “not feasible”: the scale is insufficient and the impact is regressive. Dalio sees the fundamental issue as the productivity gap, not the tax mechanism. Tariffs are a valid tool within a broader plan to rebuild domestic industries and independence, but they can’t carry the whole load.

Stage Five: When the System Itself Is in Jeopardy

Dalio places the US in Stage Five of his six-stage national cycle model, drawing from 500 years of historical data. The defining characteristic: irreconcilable differences.

“When the causes people are behind are more important to them than the system, the system is in jeopardy.”

He draws parallels to Plato’s writings on democratic fragility (circa 350 BC) and Julius Caesar’s assassination in the Roman Senate. The pattern: internal divisions escalate until one side decides the system itself is less important than winning.

His prediction is grim. Democrats will likely take the House in midterm elections, leading to complete gridlock. “Nobody can succeed because everybody’s going to be fighting.” The country needs a strong leader to force painful reforms, but democracy inherently resists the concentration of authority needed to execute them. The unspoken choice between “some form of socialism and some form of fascism” is quietly emerging.

Three things, he says, have made countries successful throughout history: educating children well (both in skills and civility), maintaining an orderly civil environment where people can compete productively, and staying out of wars, both civil and international. The US is having problems with all three.

AI Is Eating Everything — and Might Eat Itself

Dalio’s AI analysis is unusual in Silicon Valley discourse. He doesn’t discuss model architectures or capabilities. He discusses competitive dynamics and profit logic.

His historical pattern: in every technology bubble, investors believe they’re betting on the technology. They’re actually betting on companies. The technology survives; most companies don’t. The dot-com bubble and the late-1920s technology speculation followed exactly this pattern. AI companies face the same dynamic.

But there’s a structural twist specific to AI. The US operates on a profit-recovery model: massive capital investment that must eventually generate returns. China may operate differently. Dalio describes the Chinese philosophy:

“In China, they would say usage of AI is fantastic. So, it should be like electricity. Let’s make it free for everyone and open source for everyone.”

China captures productivity gains through usage rather than through vendor profits. Their technology is “almost as good” and “not far behind,” and it’s available for free. This creates a systemic threat to the American AI business model: how do you earn returns when a near-peer competitor offers comparable products at zero cost?

This isn’t a technology gap problem. It’s an institutional mismatch. Two different systems with incompatible economic logics competing in one global market. Dalio doesn’t predict which wins, but he’s clear that the assumption of inevitable profit recovery that underpins AI stock valuations may be wrong.

The Marshmallow Test

Asked what he’d change in the Constitution, Dalio invokes the marshmallow test: the child who chooses two marshmallows in 20 minutes over one marshmallow now tends to have a better life. America’s fundamental problem, he suggests, is a preference for instant gratification at the national level.

But he acknowledges the system’s remarkable adaptability. Throughout history, debts have been written off, crises navigated, recoveries achieved. Legislating the balance between fiscal prudence and innovative risk-taking is inherently difficult; rules strict enough to enforce discipline would stifle the entrepreneurship that drives growth.

His ultimate advice is simple: read history, understand these cycles, and strive for balance.

Afterthoughts

Dalio brings something genuinely rare to the AI conversation: a macroeconomic and historical-cycles perspective that most technologists don’t have and most economists don’t apply to AI. His “AI eating itself” framing isn’t about technology failing; it’s about the business model failing because of global competitive dynamics. This angle is almost entirely absent from Silicon Valley discourse.

• The China-as-free-electricity framing is the most consequential claim. If China treats AI as public infrastructure and captures value through productivity gains rather than vendor profits, the entire US AI investment thesis, which requires profit recovery, faces a structural challenge that no amount of model improvement can solve.
• “Investors bet on companies, not technologies” has sharp teeth. Every technology cycle produces this confusion. The internet survived the dot-com crash; most dot-com companies didn’t. AI will survive whatever comes; the question is which companies survive with it.
• The K-economy observation connects to AI in ways Dalio only hints at. If 60% of Americans read below a sixth-grade level and AI is replacing lower-skill labor, the technology accelerates the very inequality that destabilizes the system. AI may be both the most powerful technology and the most destabilizing force in the same economy simultaneously.
• Stage Five is a framework, not a prediction, and that’s its strength. Dalio isn’t forecasting a specific event. He’s saying the structural conditions that precede collapses in his 500-year dataset are present. The value isn’t in the prediction; it’s in recognizing the pattern.
• The gold thesis was tested and passed. A year ago Dalio recommended gold allocation. Gold went from $2,900 to $5,200 while Bitcoin dropped 25%. Whatever you think of his broader cycle theory, the specific trade worked, and his reasoning (institutional, not speculative) proved more durable than the “digital gold” narrative.