Shall We Repeal The Laws Of Economics

The fiscal deficit is running near 6% of GDP despite 4% unemployment, an unusually elevated level during an economic expansion. Separately, McKinsey estimates more than $5 trillion will be spent globally on AI-related data centers through 2030, underscoring a major infrastructure-capex requirement that could add to financing and fiscal pressures.

Analysis

The relevant transmission channel is not the nominal size of AI capital expenditure, but the collision between private infrastructure financing demand and Treasury supply. If long-end real yields remain elevated, hyperscalers can absorb higher funding costs from operating cash flow, while leveraged data-center developers, merchant power projects and smaller AI infrastructure vendors face a materially higher hurdle rate. This favors balance-sheet-rich platforms (MSFT, GOOGL, AMZN, META) over higher-duration infrastructure equities and REIT structures such as DLR and EQIX, whose valuation support is more sensitive to cap-rate expansion.

Over the next 1-3 months, auctions, inflation prints and term-premium moves matter more than incremental AI spending announcements: a renewed rise in the 10-year real yield above recent highs would pressure long-duration software, data centers and unprofitable AI beneficiaries simultaneously. Over 6-18 months, fiscal dominance could create a second-order winner in defense, grid equipment and domestic industrial capacity, but only where backlog pricing offsets labor and financing inflation; ETN, PWR and GEV are better positioned than fixed-price EPC exposure. The contrarian view is that sustained public deficits can support nominal revenue growth and reduce the real burden of fixed debt, so the key differentiator is contractual pricing power rather than simple leverage.

The missing variable is power availability. Capital spending may shift from GPU procurement toward transmission, generation and cooling once local grid constraints bind; this would slow near-term data-center revenue conversion even if headline AI budgets remain intact. A decisive decline in real yields, evidence that hyperscaler capex is translating into AI revenue, or accelerating utility interconnection approvals would falsify the relative-duration thesis.

Key Decisions for Investors

• Express a 3-6 month quality-duration pair: long MSFT or META / short a basket of DLR and EQIX, sized modestly. The thesis is that cash-funded capex and diversified revenue withstand higher real rates better than cap-rate-sensitive infrastructure owners; stop out if 10-year real yields fall materially and data-center leasing spreads accelerate.
• Build a 6-18 month watch-list long in ETN, PWR and GEV rather than chase semiconductor beta. Enter on broad rate-driven pullbacks, contingent on backlog growth and margin guidance holding; grid bottlenecks create scarcity value, but fixed-price backlog deterioration would invalidate the setup.
• Use TLT puts or a short IEF overlay against AI-infrastructure longs through the next major Treasury refunding and inflation releases. This is a hedge, not a standalone fiscal-collapse trade; reduce if auction demand improves and real yields retreat.
• Avoid adding to highly levered data-center, power-development or pre-profit AI infrastructure names until financing terms, signed customer commitments and utility interconnection timelines are independently verified. The risk/reward is unfavorable if capex announcements are not matched by contracted revenue and available power.