Goldman's 27% AI Profit Forecast Is a Liquidity Claim Wearing an Earnings Costume

Goldman Sachs forecasts a 27% jump in S&P 500 profits, with AI spending carrying the load. The forecast is not the story. The funding mechanism is. A 27% aggregate profit expansion — roughly three times the index's long-run norm of 8% to 10% — does not materialize because American corporations suddenly discovered efficiency. It materializes because someone, somewhere, is willing to finance the capital expenditure that manufactures it. AI spending carries the load. Fine. Then answer the only question that survives contact with a balance sheet: who is carrying the load, and with whose money? I have audited enough sell-side models to know what this genre omits. A research desk does not publish a 27% number and then walk you through the discount rate that makes it coherent. It publishes the number, attaches a technology theme, and lets the reader supply the conviction. The AI narrative is doing two jobs at once — it is the revenue story and the alibi for the capex. That dual role is a structural tell, not a compliment. To read this forecast correctly, you have to stop reading it as an earnings document and start reading it as a liquidity document. Earnings are the dependent variable. Liquidity is the independent one. Always has been. The capex that carries the load is not self-funding. It moves through three channels: hyperscaler operating cash flow, corporate debt issuance, and a rotating cast of vendor-financing arrangements that keep the accounting clean while the cash moves in a circle. Every one of those channels is rate-sensitive. When the cost of capital falls, the buildout accelerates and the profit forecast validates itself. When the cost of capital rises, the same buildout becomes a drag on margins, and the 27% number quietly becomes 19%, then 11%, without anyone issuing a correction. M2 money supply and the Federal Reserve's balance sheet are the two variables I track before I look at any crypto chart. Not because they are predictive in a clean, linear sense — they are not — but because they set the boundary conditions for everything else. When M2 expands, the marginal dollar has to go somewhere, and it flows down the risk curve until it reaches the highest-beta assets. Crypto sits at the far end of that curve. AI equities sit one rung above it. The Goldman forecast is, functionally, a bet that the curve stays lubricated long enough for the profit to be realized. That is a monetary bet masquerading as a corporate one. Here is where crypto enters, and where most equity analysts stop thinking. Crypto is not a separate asset class that occasionally correlates with the Nasdaq. It is the highest-beta expression of the same global liquidity impulse that funds AI capex. When the money printer runs — and I use that phrase precisely, not rhetorically — both AI equities and digital assets inflate together, because they draw from the same pool. When that pool contracts, they deflate together, and the correlation traders dismiss as risk-on, risk-off reveals itself as a shared dependence on the same marginal dollar. I mapped this in 2020, during DeFi Summer, when I built a Python model correlating Compound's on-chain interest-rate volatility against the Fed's balance sheet. The finding was uncomfortable and I have never stopped using it: DeFi yields were not decoupled from monetary policy. They were a leveraged proxy for it. The same logic now applies to AI equities. The 27% forecast is a statement about the future supply of capital, not the future productivity of chips. So the question for anyone holding Bitcoin, ETH, or a tokenized version of the AI trade is not whether Goldman is right. It is whether the liquidity that validates the forecast is the same liquidity that supports the position. The answer is yes, with a lag and a leverage multiplier attached. Let me be concrete, because abstraction is where forecasts hide. Start with the physical constraint. AI datacenters and Bitcoin mining compete for the same three inputs: electricity, silicon, and capital. In 2021, I watched hash rate migrate out of China and reprice the global cost of production for Bitcoin. The same dynamic now runs in reverse. Hyperscalers are signing fifteen-year power purchase agreements at prices Bitcoin miners cannot match, and the miners are being priced out of their own grid interconnects. This is not a metaphor for competition. It is competition, denominated in megawatts. The on-chain consequence is measurable. When miners lose the power bid, hash rate growth stalls, and the security budget of the network comes under pressure. This is the point most people miss. Bitcoin's security model depends on fee revenue plus block subsidy, and the subsidy is scheduled to halve. I have argued for two years that without the inscription wave injecting genuine fee demand into the base layer, Bitcoin's long-term security model would already be in structural trouble. Ordinals and inscription activity did not just create a speculative fad — they proved the base layer could monetize blockspace in ways the original monetary thesis never anticipated. AI's demand for power now threatens to squeeze the very mining economics that subsidize that security. The two stories are not adjacent. They are wired into the same circuit. Now the capital side, which is where the Goldman number actually lives. If AI capex is funded by corporate debt, then the marginal buyer of that debt sets the cost of capital for the entire risk complex, crypto included. In 2024 I spent six months inside the custody structures of BlackRock's iShares Bitcoin Trust, tracing how the underlying storage mechanisms introduce subtle regulatory and operational risks the marketing material does not surface. The lesson generalized: institutionalization did not insulate crypto from macro. It wired crypto directly into the same fiduciary plumbing that funds AI. When a sovereign wealth fund allocates to digital assets — and I have advised on exactly these allocations — it is not making a crypto bet. It is making a duration bet on the global cost of capital, and crypto is the instrument it chose. Consider the second-order effect on DeFi. Liquidity fragmentation is not a real problem. It is a manufactured narrative that venture capital uses to justify funding the next chain, the next rollup, the next bridge. There are dozens of Layer 2s now competing for the same small base of users, and the result is not scaling — it is the slicing of already-scarce liquidity into fragments too thin to price efficiently. AI capex does not fix this. It accelerates it, because the capital that might have funded a new DeFi primitive is now absorbed into GPU clusters and power contracts with a fifteen-year lockup. The opportunity cost of the AI trade is measured in abandoned crypto infrastructure. That is the integration nobody on either side of the aisle wants to admit. Equity analysts think crypto is a casino. Crypto natives think equities are boomer assets. Both are wrong in the same direction. They are two tranches of the same capital structure, and the AI profit forecast is the senior tranche telling you what the junior tranche is about to feel. Here is the contrarian read, stated as a hypothesis rather than a conviction, because the data is thin and I refuse to pretend otherwise. The consensus assumes AI and crypto rise together on the same liquidity tide. The decoupling thesis says the opposite: that AI capex becomes the mechanism through which liquidity is drained from crypto. Consider the mechanics. If hyperscalers and their suppliers absorb an ever-larger share of global capital formation, they crowd out the speculative margin that funds digital assets. The AI trade becomes the exit liquidity for the crypto trade — not through malice, but through arithmetic. There is only so much risk capital, and a 27% profit forecast is a claim on a very large slice of it. Exit liquidity is a social construct — until it isn't, and the bid disappears. I have watched this movie before. In 2021, I spent three months pulling on-chain transaction data for Art Blocks and Bored Ape Yacht Club and calculated that 85% of secondary volume was wash-trading bots, not collector demand. I called it a liquidity illusion and published a report the mainstream ignored. The structure was the same then as it is now: a compelling narrative, a self-reinforcing price signal, and a thin base of genuine marginal buyers underneath. When the narrative rotated, the floor did not hold. The AI trade is not the NFT trade — the cash flows are real, the capex is real, the chips are real. But the marginal buyer is just as narrative-driven, and that is the fragility the 27% number conceals. There is one more layer. The yield on a staked position is not compensation for risk in the way the brochure claims. Yield is just rent for your ignorance — the fee you pay for not understanding where the return actually comes from. In a liquidity-driven regime, that rent is cheap, because the tide lifts everything. In a liquidity-constrained regime, the rent becomes the entire story, and the positions that survive are the ones whose cash flows exist independently of the narrative. Very few crypto positions clear that bar. Almost none of the AI-adjacent ones do. So where does this leave a crypto portfolio in a bull market being repriced by an AI capex cycle it does not control? Watch the funding, not the forecast. If corporate debt spreads tighten and hyperscaler capex guidance rises, the liquidity tide is still coming in, and crypto rides it. If spreads widen and capex guidance gets trimmed by even one major player, the tide turns, and the 27% number becomes the exit sign. Algorithms don't panic. They just reprice the marginal buyer, and the marginal buyer is you. The real question is not whether Goldman is right about profits. It is whether you are positioned as the liquidity or the exit. Answer that before the next earnings revision answers it for you.