A liquidity contraction is underway. The financial engineering holding the AI buildout together is beginning to fray. And the popular narrative that artificial intelligence is supercharging American growth may be, in the words of Totem Macro's Whitney Baker, "a total inversion of the truth."
Baker's July 2026 research note, "Financial Chicanery and the Reality of the AI Bubble,”1 is not a polite critique. It is a systematic dismantling of the dominant market thesis, constructed from first principles and traversing macro liquidity flows, corporate accounting structures, memory chip dynamics, and the mechanics of collateralized debt cycles. The argument is dense, interconnected, and deliberately contrarian. It deserves to be taken seriously.
The Liquidity Problem
Baker's organizing thesis is that dollar liquidity and cross-border lending are tightening acutely, even as domestic equity markets recorded a historic blow-off top. She frames this dichotomy plainly: "steep underperformance of the market's former leadership (inc. Mag 7, PE and dollar-sensitive indices like H-shares), and near-record equity funding costs in what is typically the loosest quarter of the year speaks to the dichotomy between liquidity on the one hand, and leverage on the other."
The proximate driver of the blow-off, in Baker's reading, was not fundamental improvement but a brief window in which government bond and bill supply went negative, easing repo markets and enabling "the fastest and biggest spike in equity margin borrowing on record." That window has now closed. The RRP, which Baker identifies as the hidden fuel of the entire post-Covid "K-shaped" expansion, ran dry in October 2025.
AI as a Growth Drag, Not a Boost
The more provocative claim is that AI spending is actively harming the broader economy. Baker contends that AI capex "crowds out domestic income and ships it to other economies" via imported chips and equipment. The consequence is not a productivity boom but an inflationary drag that keeps rates elevated and compresses real household income. "AI will accelerate the recession before it can generate revenue," she writes, "because AI-related activity is smothering its own potential customers."
This is not a theoretical abstraction. Baker points to prime-age labour force participation declining faster than in any non-Covid period, consumer delinquency rates at crisis levels across all non-mortgage loan types, and first-quarter PCE revised down to approximately 50 basis points annualized. The labour market, she argues, is "clearly weak."
The Accounting Mirage
Baker reserves particular scrutiny for the financial structures enabling the buildout. Datacentres are ringfenced in special-purpose vehicles, held off the balance sheets of hyperscalers and neoclouds, with lease commitments totalling approximately $850 billion, of which only around $200 billion is currently recognized. Chip vendors, private credit providers, and hyperscalers have entered into tripartite off-balance-sheet arrangements that allow each party to avoid full recognition of liabilities while booking revenue.
"This is why index earnings look strong," Baker states plainly. "It's the same profit illusion and structure as the Dot Com." Her estimate is that delayed or ringfenced depreciation has inflated index earnings by roughly 15%. Free cash as a share of profits sits at just 75%, and net debt per share has risen inexorably even through reported "strong earnings." The price-to-sales and price-to-cashflow multiples, she notes, tell a very different story than the headline PE.
The Memory Price Question
Baker raises a pointed question about the memory chip shortage narrative. Korean chip producers, she observes, are cutting output even as margins are reportedly 75 to 80 percent. "If demand is so insatiable," she asks, "why is Korean chip production falling?" Her alternative reading: the apparent shortage reflects seasonal consumer tech demand hitting an AI-depleted market in the usual pre-holiday cycle, not structural scarcity. The parallel to the late-1999 memory squeeze, which unwound "just as quickly and just as unexpectedly," is drawn explicitly.
Five Key Takeaways for Advisors and Investors
1. Liquidity leads fundamentals. The contraction in cross-border flows and repo markets historically precedes sharp deteriorations in global growth. Positioning should reflect that sequence, not the trailing earnings narrative.
2. Index earnings quality is deteriorating. Headline profit figures are materially inflated by deferred depreciation and off-balance-sheet structures. Price-to-cash-flow and price-to-sales tell the more honest story.
3. The AI capex cycle is crowding out domestic demand. Spending that flows primarily into imported goods and equipment creates income abroad, not at home. The macroeconomic math is not additive.
4. Consumer stress is not a tail risk. Delinquency rates, declining real incomes, and contracting PCE are concurrent phenomena, not lagging indicators. The lower branch of the K-shaped economy is already in distress.
5. Financial engineering defers, it does not eliminate. SPV structures, vendor financing, and minority-interest accounting have extended the cycle. They have not altered its destination.
Footnote:
Baker, Whitney. "Financial Chicanery and the Reality of the AI Bubble." Totem Macro, 10 July 2026, x.com/TotemMacro/status/2085335558004392317.
