AI productivity upside could be 10x the current estimates: BofA

Bank of America estimates AI could lift global economic growth by up to 1 percentage point annually over the next decade, taking growth from about 3.5% to 4.5%. The report says AI is already boosting task-level productivity, with software development up as much as 55% and writing tasks about 40%, though economy-wide productivity remains near 0.1% due to slow adoption and implementation constraints. It also argues faster AI adoption could eventually push neutral rates higher in leading economies as investment demand rises.

Analysis

The market is still underpricing the fact that this is not just an AI demand story; it is a capital-intensity and compliance-story layered on top of it. The near-term beneficiaries remain the compute stack and the banks financing the buildout, but the composition of winners is shifting toward suppliers with clean export channels, diversified end-markets, and less regulatory overhang. That makes the biggest risk to the trade not “AI demand slowing,” but procurement friction: if buyers start treating compliance risk as a supplier-selection criterion, order flow can rotate away from vendors exposed to customs scrutiny even if end demand stays intact.

Second-order, a higher-quality compliance regime is mildly bearish for the most expedient parts of the GPU supply chain and bullish for the audited, enterprise-facing ecosystem. Expect some pull-forward of orders into compliant channels, but also longer sales cycles, more documentation costs, and potential mix shifts toward direct purchase relationships and hyperscaler procurement teams. This should modestly improve pricing power for the dominant platform owner over time, while compressing margins for distributors and gray-market intermediaries that depend on speed and geographic arbitrage.

The macro angle matters for rates: if AI adoption keeps translating into even a partial productivity impulse, the market should continue to reprice the terminal-rate path higher in the US relative to slower adopters. That is supportive for financials with levered exposure to loan growth and capital markets activity, but only if credit stays stable; the cleaner expression is via long-duration productivity beneficiaries rather than broad cyclical beta. The main contrarian miss is that the current productivity gains may be front-loaded into software and knowledge work, with much slower spillover into capex-heavy sectors, so the earnings upgrade cycle could be narrower than the narrative implies.

Key Decisions for Investors

• Maintain a tactical long NVDA for 1-3 months, but finance it with upside calls rather than outright common; the best risk/reward is a call spread into any compliance-driven weakness in the supply chain, since end-demand should remain resilient while the market reprices quality of execution.
• Add BAC on a 3-6 month horizon as a relative beneficiary of a firmer U.S. neutral rate and stronger capital-demand environment; use it as a low-volatility way to express the higher-for-longer AI productivity thesis, with downside limited if rates stall.
• Short a basket of lower-quality AI hardware intermediaries or non-compliant channel names versus long NVDA over the next 1-2 quarters; the edge comes from procurement normalization, not from an AI demand collapse.
• If the tape sells off on export/compliance headlines, buy NVDA 1-2 month call spreads on weakness rather than chasing the rebound; the catalyst is that enterprise and hyperscaler buyers usually re-route rather than cancel orders.
• Avoid broad long duration or rate-sensitive cyclicals as the primary AI expression; the cleaner trade is selective longs in beneficiaries of higher productivity and stronger investment demand, not a blanket bet on growth re-acceleration.