BofA says AI productivity boost visible in narrow tasks, not yet economy-wide

BofA Global Research estimates AI is currently adding only about 0.1% per year to aggregate macroeconomic productivity, with potential upside to 1.0% annually over the next decade if adoption broadens and costs fall. The report also says U.S. Q2 GDP tracking remains steady at 2.6% saar, while final May University of Michigan consumer sentiment came in at 48.2 versus 48.0 expected. Overall, the piece is mostly macro commentary with limited near-term market impact.

Analysis

The important read-through is not that AI is weak, but that the market is still pricing the wrong part of the adoption curve. Near-term macro lift is too small to matter for broad cyclicals, yet the winners are concentrated in the picks-and-shovels layer where customers are forced to spend before they can monetize productivity. That argues for continued relative outperformance of infrastructure and application names with direct usage-based demand, while the broader “AI beta” trade should be treated as a dispersion trade rather than a blanket long.

SMCI remains a higher-beta beneficiary if capital spending on AI servers stays elevated, but the second-order risk is that the market increasingly distinguishes between shipment growth and durable gross margin expansion. If enterprise adoption slows, hardware names are the first place where order normalization shows up, and the multiple can compress faster than revenue. APP is the cleaner expression because ad optimization and targeting benefits can monetize faster and with less dependency on broad economic diffusion.

The contrarian takeaway is that weak macro translation is actually bullish for select AI vendors: it extends the period in which companies justify spend on efficiency tooling even without a visible GDP payoff. The market may be underestimating how long budgets stay allocated to AI even if the economy does not “feel” the productivity gains. The biggest reversal risk is a capex pause triggered by tighter financial conditions or disappointing enterprise ROI disclosures over the next 1-2 quarters.

I would avoid extrapolating this into a broad risk-on call for consumer or housing-linked names. If AI gains remain localized, then the winners are the firms selling productivity, not the firms relying on economy-wide throughput. That keeps the trade focused on earnings compounding and away from cyclical beta.

Key Decisions for Investors

• Long APP / short a basket of lower-quality software names with weaker AI monetization over the next 3-6 months; APP has a better path to near-term earnings leverage if AI spend remains concentrated in measurable ROI use cases.
• Trade SMCI tactically, not structurally: buy pullbacks only, using 1-3 month horizons, because the upside is tied to capex intensity while the downside is a multiple reset if hyperscaler ordering slows.
• Pair long AI infrastructure/applications vs short consumer-sensitive cyclicals for a 2-4 month relative-value trade; if AI adoption is not boosting broad productivity, macro-exposed names should lag while direct beneficiaries keep compounding.
• Use call spreads instead of outright longs in SMCI for event-driven exposure into the next earnings cycle; the setup offers convex upside, but margin and demand volatility make the left-tail sharp.
• If you want lower-beta AI exposure, prefer APP over hardware for a 6-12 month hold: the risk/reward is better because monetization is closer to product usage than to the next hardware refresh cycle.