U.S. economy hits pivotal milestone: Spending on data centers and other information-processing hardware now exceeds housing investment

U.S. inflation-adjusted spending on information-processing equipment reached $752 billion in Q2, surpassing $748 billion of residential fixed investment after AI-related spending rose 51% from early 2021 while housing investment fell 18%. S&P Global expects capex by Alphabet, Amazon, Microsoft, Meta, Oracle and SpaceX to exceed $1.3 trillion in 2027, but warns investment is outpacing revenue and forecasts collectively negative operating cash flow in 2026-27. Elevated Treasury yields and nearly 7% 30-year mortgage rates continue to suppress housing activity, with August housing starts down 2.6% to a 1.275 million annualized pace, while local political resistance to data-center development is increasing.

Analysis

The relevant equity risk is not headline capex but the transition from self-funded investment to externally financed capacity. As hyperscaler free cash flow turns negative, incremental debt raises the hurdle rate for marginal AI projects and creates a valuation split: MSFT and GOOG retain the strongest distribution-led monetization paths, while ORCL carries greater execution and financing sensitivity because its AI infrastructure build is more concentrated relative to its cash-generation base. Suppliers with contracted backlog—NVDA, AVGO, VRT, ETN and PWR—remain the near-term beneficiaries, but their 6-18 month risk is an abrupt order digestion once utilization, rather than announced capacity, becomes the binding KPI.

The second-order macro effect is a crowding-out of rate-sensitive construction and a regional power bottleneck. Data-center load growth supports regulated utilities and grid equipment, especially CEG, VST, GEV, ETN and PWR, but local opposition and interconnection delays can shift spend from new greenfield campuses toward existing powered sites and retrofit-heavy deployments. That favors VRT and electrical-equipment vendors over land-heavy data-center developers; it also makes announced capex a poor proxy for revenue timing.

Consensus still treats hyperscaler capex as uniformly bullish for AI beneficiaries. The more investable distinction over the next 1-3 months is whether management can show AI revenue/backlog conversion faster than depreciation and interest expense growth; absent that, capex guidance is likely to trigger multiple compression even if aggregate spend holds. Residential weakness is more likely a rates trade than a direct AI offset: a sustained Treasury-yield decline would improve homebuilder earnings expectations well before housing starts recover.

Key falsifiers are hyperscaler commentary on GPU utilization and lease cancellations, quarterly capex-to-revenue ratios, investment-grade credit-spread widening, and evidence that power/interconnection constraints defer—not merely relocate—deployments. A meaningful upward revision to cloud/AI revenue guidance would invalidate a near-term monetization short thesis.

Key Decisions for Investors

• Prefer a quality pair over a broad hyperscaler short: long MSFT / short ORCL for 3-6 months. MSFT has more diversified software monetization and balance-sheet flexibility; ORCL is more exposed to any delay in AI capacity ramp. Reassess if ORCL reports accelerating cloud revenue with stable or improving free cash flow.
• Maintain selective long exposure to AI physical-infrastructure bottlenecks via ETN and PWR over 6-18 months, rather than adding to high-multiple compute suppliers. Size for execution risk: utility permitting/interconnection delays can push revenue recognition out 2-4 quarters.
• Use a 1-3 month tactical long XHB or ITB only on a sustained decline in the 10-year Treasury yield, preferably after a weekly close below the prior yield range. Housing equities can rerate ahead of activity; avoid the trade if mortgage rates remain near 7% and permit trends continue deteriorating.
• Set an earnings-season alert for negative free-cash-flow guidance, rising debt issuance, or capex growth materially exceeding cloud revenue growth at AMZN, GOOG, META and MSFT. That combination supports reducing hyperscaler beta and adding downside hedges through QQQ puts rather than shorting suppliers with booked backlog.