Adam Grossman, Global Equity CIO at RiverFront, opens with a reframing. The question is no longer whether artificial intelligence is transformative — the productivity gains and cash flows generated by AI hardware companies have already answered that. The real question, as Grossman puts it, is whether "all these positive fundamentals and future earnings have already been fully priced into the return." That distinction — between a real theme and a fairly priced one — is the entire piece in miniature.
The Framework: Intrinsic Value
RiverFront's discipline sits deliberately between two failure modes. A pure value screen risks underestimating a growth theme's durable earnings power; a pure momentum chase risks overpaying for earnings already reflected in the price. The firm's answer is a four-step Intrinsic Value process: assess the macro backdrop, analyze the competitive framework, project earnings and identify catalysts, and finally assess valuation relative to those earnings. Grossman is candid about how demanding that last step is in practice: "A theme can be fundamentally intact, and the stock can still have appreciated too much. Walking away from a party still going strong is the hardest call in active management…this framework helps give us the discipline to do that."
Five Themes, Five Different Verdicts
Applying that lens across the AI ecosystem produces a genuinely split picture — not a blanket bullish or bearish call.
AI Model Providers — names like Anthropic, OpenAI, Palantir, and SpaceX — carry the most compelling stories and the least earnings history. Grossman notes "price discovery is at the mercy of sentiment, as the earnings and growth rates necessary to justify the valuations will take years or even a decade to materialize." He doesn't dismiss the opportunity cost of standing aside, but the valuation gap here is the widest in the piece.
Hyperscalers and Semiconductors get the most constructive treatment — "the sweet spot for stock pickers" — on the strength of durable earnings. But Grossman flags a real caveat: "security selection will become increasingly important as the AI Trade unfolds," since price appreciation in parts of this group has already outrun earnings growth, and he expects volatility even where fundamentals hold.
AI Physical Infrastructure — utilities, industrials, power and cooling equipment — earns a "still attractive" label because the earnings case is structural rather than narrative-driven: "You cannot add power generation or transmission capacity overnight, and the demand signal from hyperscalers is locked in for years." Risk here comes from a hyperscaler capex pivot, faster-than-expected supply relief, or regulatory backlash.
Early Disrupted Industries — principally software — is where Grossman pushes back hardest against consensus. Software has been broadly de-rated on disruption fear, but he draws a dot-com parallel: many threatened incumbents absorbed the new technology rather than being replaced by it. "A lot of the 'AI Hype' in our minds is the underestimation of the complexity of enterprise software." Crucially, "Neither analyst earnings estimates nor management guidance have confirmed a slowdown in earnings."
Early Non-Technology Enterprise Adopters — industrials, healthcare, consumer companies applying AI internally — remain a "wait and see." Grossman is careful here: "our general view is that layoffs attributed to AI could be related to post-COVID over hiring." The margin-expansion story hasn't shown up in the aggregate data yet.
The Divergence Is the Point
What ties the five themes together is precisely that they don't move together. Model providers are overpriced against a distant earnings horizon; infrastructure is fairly priced against a visible one; software may be underpriced against fear that hasn't materialized in the numbers. Grossman's framework exists to keep those three very different situations from being treated as one undifferentiated "AI trade."
Conclusion
Grossman closes with an important admission of fallibility: "A macro shift, earnings miss, a competitive threat moving faster than expected: any of these can upend a well-founded thesis." The framework's value isn't infallibility — it's that it forces the team to define in advance what evidence would change its mind, and act on it. As benchmark concentration in AI names grows, he notes, passive investors absorb that risk without ever choosing it — which is precisely the case for active management right now.
Five Key Takeaways for Advisors and Investors
- Intrinsic value is a continuous discipline, not a one-time screen — the goal is tracking price against evolving earnings potential, not making a single call and holding it.
- The AI trade is not monolithic; model providers, hyperscalers, infrastructure, disrupted software, and enterprise adopters each warrant separate treatment.
- Software incumbents represent the most asymmetric opportunity in the piece — the market has priced in disruption the earnings data hasn't confirmed.
- Grossman explicitly expects to be wrong about parts of this thesis, and treats predefined disconfirming evidence as the safeguard.
- Rising AI concentration in benchmarks means passive investors are absorbing risk by default; active positioning is framed as the deliberate alternative.
Footnote:
1 Grossman, Adam. "Finding Value in the Crowded AI Trade." RiverFront Investment Group, 30 June 2026, www.riverfrontig.com/insights/ai-trade/.
