Can AI Outrun America’s Deficit Problem?

National Economic Council Director Kevin Hassett said AI's economic benefits may be materially understated in official data, likening the measurement challenge to the early internet era. He also argued that federal workforce reductions can generate budget savings and remained optimistic on growth despite rising debt, while acknowledging that achieving a federal deficit of 3% of GDP will require greater focus.

Analysis

The investable issue is not whether AI ultimately lifts productivity, but whether it appears in earnings before it raises the discount rate. Over the next 1-3 months, a stronger-growth/fiscal-slippage narrative is more likely to steepen the Treasury curve than to re-rate broad software: higher long-end yields compress duration-sensitive multiples even if nominal GDP expectations improve. The clean near-term beneficiaries are firms monetizing AI infrastructure now—NVDA, AVGO, ORCL and data-center power/thermal suppliers—rather than application software names whose labor-savings claims remain difficult to verify in reported margins.

A leaner federal payroll creates a second-order headwind for government-services contractors and federal-office real estate, but the direct budget effect is unlikely to materially alter Treasury supply or the term premium without a broader entitlement/revenue package. Leidos (LDOS), Booz Allen (BAH), SAIC and CACI should be monitored for procurement delays, contract rebids and lower headcount-based revenue growth; the risk is more acute over 6-18 months than in the next quarter. Conversely, automation vendors with credible public-sector implementation channels could gain share if agencies are pushed to maintain service levels with fewer employees.

Consensus may be too quick to treat unmeasured productivity as disinflationary. In the transition phase, AI investment is capital-, power- and talent-intensive, potentially sustaining nominal demand while productivity gains diffuse slowly; that combination is bearish long-duration bonds and selectively bullish industrial power infrastructure. This thesis is falsified if core services inflation and wage growth decelerate while productivity revisions improve, or if 10-year real yields fall despite continued heavy Treasury issuance.

Key Decisions for Investors

• Maintain a 1-3 month curve-steepener bias: short TLT versus long SHY, sized modestly. Risk/reward improves if 10-year real yields remain above recent ranges; exit if weaker payroll/CPI prints drive a sustained decline in 10-year yields and the curve bull-steepens.
• Pair long AI monetizers/infrastructure (NVDA, AVGO, VRT, ETN) against a basket of high-duration software exposure (IGV) over 3-6 months. The catalyst is earnings evidence that infrastructure revenue converts faster than enterprise-seat productivity savings; cut the trade if hyperscaler capex guidance is reduced or IGV earnings revisions turn positive relative to semis.
• Watch—not yet initiate—a short basket in federal-services contractors (LDOS, BAH, SAIC, CACI). Require evidence of contract award delays, lower funded backlog, or FY guidance cuts; absent those data, workforce reductions may simply shift spending toward contractors and make the short structurally wrong.
• Add selective long exposure to grid and data-center electrification (ETN, PWR) on pullbacks rather than chase AI software beta. A 6-18 month thesis depends on utility interconnection backlog and hyperscaler capex remaining intact; invalidate on a broad data-center construction slowdown or materially weaker power-load forecasts.