AI Is Already in the Price: What Goldman's Macro Math Means for Your Equity Book
Is the market correctly valuing AI's benefits? The short answer: close enough to justify the hype, but pricing is pressing against the high end of what the economy can realistically deliver. Valuations aren't at bubble levels, yet optimism is running well ahead of the macro impact.
The macro math: $8T baseline, $5T-$19T plausible
Goldman's estimate for the present discounted value of generative AI-driven capital revenues lands around $8 trillion, with a range of $5 trillion to $19 trillion. That's big enough to validate heavy AI capex across chips, data centers, and software. In other words, the spending wave has a rational anchor-even if the distribution of who captures it is uncertain.
For a deeper look at the broad research backdrop, see Goldman's work on generative AI's economic impact: Goldman Sachs Intelligence: Generative AI.
Market pricing: $19T+ added since late 2022
Since ChatGPT launched, AI-adjacent market value has climbed by more than $19 trillion. That includes semiconductors, hyperscalers, and roughly $1 trillion added to the top private model providers. Put simply: market gains now sit at the upper bound of the projected macro benefits and are far above the $8 trillion baseline.
Two traps that inflate expectations
- Fallacy of aggregation: Extrapolating standout earnings from a few players across the entire stack pushes joint value (chips + models + cloud) beyond what they can collectively capture.
- Fallacy of extrapolation: Early profit spikes rarely persist. Competition, pricing pressure, and input costs tend to grind excess returns down over time.
The productivity upside is real-but pacing matters
AI could lift U.S. productivity by about 1.5 percentage points annually for a decade, raising GDP and earnings by roughly 15% over time. As long as the economy and the capex cycle hold, sentiment likely stays buoyant. But outside of hardware, profit pools are still thin, which leaves the tape exposed if revenue ramps slip.
Portfolio playbook: Price discipline over narrative
- Anchor to PDV bands: Stress test exposures against the $5T-$19T range and the $8T base case. If your AI basket implies the upper tail, demand upper-tail execution.
- Underwrite value capture, not headlines: Map who actually monetizes each dollar of AI spend-chip designers, manufacturers, cloud platforms, model providers, workflow software, integrators.
- Watch unit economics: Training/inference costs, cloud gross margins, depreciation schedules, and energy intensity will determine durability.
- Follow the cash: Separate capex-fueled revenue from recurring, high-margin software and services. Look for payback periods and customer ROI evidence.
- Prepare for price compression: Assume ASP declines and efficiency gains pass through to customers over time; fade permanent-high-margin assumptions outside scarce assets.
- Balance the stack: Pair bottleneck assets (compute, advanced packaging, high-bandwidth memory, top-tier foundry) with downstream beneficiaries that can compound.
- Risk budget the story: Size positions so missed product cycles, chip supply shifts, or regulatory hits don't blow through drawdown limits.
Signals to track
- Hyperscaler AI capex run-rates, capacity adds, and the mix shift between training and inference.
- Disclosed AI revenue and attach rates in software and services; proof of net-new budgets vs. wallet share shifts.
- Cloud and chip gross margin trends versus energy, networking, and depreciation costs.
- GPU/accelerator availability, lead times, and supplier concentration risk.
- Model performance per dollar and per watt; pricing for API usage and enterprise commitments.
- Policy and compliance developments that affect data access, model training, and enterprise adoption.
Bottom line
The AI trade is credible on fundamentals, but the market has front-loaded a lot of the payoff. Upside remains if adoption and cost curves beat the base case, yet the asymmetry now cuts both ways. Keep owning quality, capital-aware winners-just price in competition, cost gravity, and the time it takes for productivity to show up in earnings.
Related resource for finance teams
For a practical view of tools reshaping workflows and P&L impact, see this curated list: AI tools for finance.
Your membership also unlocks:
