Is AI the new China Shock?

U.S. hyperscalers are projected to spend roughly $800 billion on capital expenditures in 2026, up 83% year over year; global AI spending is estimated to exceed $2 trillion this year, with hard AI investment potentially reaching $10–15 trillion over the next decade. The article argues that AI is becoming a macro-scale investment regime, with implications for productivity, disinflation, power demand and asset prices, while warning that labor displacement and inequality could intensify. The IEA projects data-center electricity use will more than double by 2030 to 945 terawatt-hours.

Analysis

The investable issue is not the headline capex total but who earns an adequate return on it. Hyperscalers commit cash before customers’ willingness to pay for AI is proven; if utilization or monetization lags, depreciation and power costs can pressure free cash flow and eventually slow orders across chips, networking and data-center infrastructure. Conversely, labor savings captured by adopters could fund further deployment—but broad cognitive-labor displacement may also weaken household income and enterprise demand, creating a second-order drag on the same AI revenue pool.

Near term (days to weeks), treat aggregate spend estimates as narrative, not earnings evidence. Over 1–3 months, watch hyperscaler capex guidance alongside cloud growth, AI-related revenue disclosures, backlog conversion and free-cash-flow trajectories. Divergence—capex accelerating while monetization or utilization fails to follow—would be a negative signal for infrastructure suppliers, including NVIDIA, and a potential multiple risk for spenders such as AMZN, GOOG, META, MSFT and ORCL. It does not establish that any one company’s investment is uneconomic.

Over 6–18 months, grid connection delays, electricity prices and equipment lead times may redirect value toward power and data-center infrastructure rather than model providers. Utilities are not automatic winners: regulated returns and the timing of rate-base additions matter. The contrarian point is that the analogy to China may overstate a uniform macro shock: market-funded investment can be cut faster than state-directed construction, while AI’s productivity gains could offset labor disruption. The thesis weakens if capex growth moderates alongside improving AI monetization and stable returns; it strengthens if spend rises while cash generation and customer adoption disappoint.

Key Decisions for Investors

• Do not buy AI exposure solely on aggregate spending forecasts. For the next earnings cycle, compare each hyperscaler’s capex guidance with cloud growth, AI revenue or usage evidence, and free cash flow; trim exposure if investment keeps accelerating without improving commercial indicators.
• Watch for a relative-value opportunity in power-grid and data-center infrastructure versus AI application or labor-intensive software, but wait for evidence of order conversion and project returns. Verify supplier backlogs, grid interconnection timelines, electricity pricing and utility rate-base treatment before sizing.
• Avoid treating NVIDIA as a direct proxy for the macro thesis or shorting it on capex concerns alone. A more actionable downside trigger is repeated evidence of weaker infrastructure demand or utilization alongside hyperscaler capex cuts; improving monetization with sustained orders would falsify that bearish setup.
• Track labor-exposed employment and enterprise software pricing over the next 6–18 months. Faster job deterioration or falling per-seat pricing would raise the risk of a demand feedback loop; stable employment and evidence that AI complements rather than replaces paid work would argue against it.