The AI narrative has been dominated by one theme: bigger, faster, more capital. Trillions committed to data centers. CapEx projections doubling. The Magnificent Seven carrying indices higher.
That may no longer be the right lens.
In this episode of Lead-Lag Live, Kai Wu, Founder and CIO of Sparkline Capital, argues that the AI trade is shifting from infrastructure to adoption. History suggests that is where the more durable winners emerge.
The market may not be positioned for that transition.
From Buildout to Adoption
Wu frames AI through Everett Rogers’ diffusion model, the classic S-curve that describes how technologies spread from innovators to early adopters and eventually into the mainstream.
The first phase of any technological revolution is infrastructure. During the railroad boom, it was track. During the internet era, it was fiber optic cable. Today, it is GPUs, hyperscalers, and data centers.
That is where investor attention has been concentrated. AI infrastructure firms have revised CapEx higher and higher, with spending expected to roughly double versus last year.
The question now shifts from “Can we build it?” to “Will they use it?”
Enterprise adoption remains early. Wu notes that only about 10 percent of businesses are using AI in production today. That figure aligns with the early adopter phase in the diffusion model. Consumer usage has grown rapidly, yet only a small fraction of users are paying subscribers.
The infrastructure exists. The monetization and ROI phase is just beginning.
The Dot-Com Parallel
Wu highlights a historical cautionary tale that should resonate with investors.
During the dot-com boom, telecom companies built massive fiber networks. When demand failed to meet expectations, 85 percent of fiber capacity went dark. Bandwidth prices collapsed by 90 percent. Several firms, including Global Crossing, ultimately failed.
The internet did not fail. Infrastructure investors did.
That distinction matters.
The risk today is twofold:
Execution risk: Infrastructure providers are spending enormous sums upfront. If demand growth disappoints or arrives later than expected, the return profile compresses.
Valuation risk: AI infrastructure names now trade at multiples far above historical norms, roughly double the broader market in some cases. Even if AI succeeds structurally, paying too high a price can lead to poor outcomes.
This is not an argument that AI fails. It is an argument that price matters.
Separating Signal from Hype
One of the more compelling parts of our discussion centered on how to measure real AI adoption.
Corporate earnings calls are filled with AI references. That alone is not informative. Wu and his team instead look for numerical evidence of ROI.
They classify mentions into three buckets:
General qualitative references.
Quantified cost savings, revenue gains, or efficiency improvements.
Explicit return-on-investment calculations relative to spending.
That distinction matters. A CEO saying “AI is transforming our business” is not the same as reporting measurable margin expansion tied directly to implementation.
We discussed examples across industries: retailers improving warehouse picking efficiency, biotech firms accelerating R&D processes, workforce management systems reducing labor costs, and advertising platforms improving conversion rates.
The key takeaway is diffusion. AI is beginning to seep into industrials, healthcare, financials, and consumer businesses. It is not confined to Silicon Valley.
That is where dispersion begins.
Where the Market May Be Mispricing Risk
Wu’s central thesis is that investors face a false binary choice.
Option one: overweight AI infrastructure through passive indices. The S&P 500 now carries enormous implicit exposure to the Magnificent Seven and related infrastructure names, approaching half the index when expanded beyond the headline seven.
Option two: avoid AI entirely through value, small-cap, or international allocations, risking structural underexposure if AI continues bending the productivity curve.
There is a third path.
Wu calls it the “early adopters” basket. These are companies outside the infrastructure layer that are aggressively implementing AI into operations.
Here is the intriguing part: valuations between early adopters and laggards remain similar. The market appears to be pricing disruption indiscriminately rather than rewarding differentiated adoption.
Historically, long-term winners in infrastructure booms have often been the adopters, not the builders. Retailers leveraged railroads. Internet platforms leveraged cheap bandwidth after telecom overbuild.
Infrastructure often becomes commoditized. Adoption drives margin expansion.
Macro Context: Concentration Risk
The macro overlay is impossible to ignore.
Index concentration has rarely been this extreme. Many investors believe they own diversified market exposure. In reality, they own a concentrated bet on continued AI infrastructure spending.
If adoption accelerates and revenue materializes, that concentration may prove justified.
If adoption lags expectations, the adjustment could be nonlinear.
Wu’s argument is not bearish on AI. It is a portfolio construction critique. The most crowded trade in the market may not offer the best risk-adjusted exposure to the theme.
The Highlight
One moment stood out.
The internet succeeded. Telecom investors did not.
That encapsulates the distinction between technological inevitability and investor outcome.
AI can reshape productivity, compress costs, and expand margins across the economy. That does not guarantee infrastructure multiples remain elevated.
What This Means for Investors
Investors should be asking three questions:
Where is AI showing up in measurable ROI today?
How concentrated is my exposure to infrastructure names through passive allocations?
Am I being compensated for valuation risk?
The opportunity may lie in identifying companies that are early, disciplined adopters of AI across industrials, healthcare, financials, and consumer sectors, rather than reflexively adding to the most obvious beneficiaries.
Dispersion is likely to increase. Adoption will not be uniform. Winners and laggards will diverge.
That is where active thinking matters.
If you want to hear the full conversation with Kai Wu and dive deeper into his framework, including how he screens for early adopters and avoids laggards, watch or listen to the complete Lead-Lag Live episode.
The AI trade is not ending. It is evolving.
DISCLAIMER – PLEASE READ: This is a sponsored episode for which Lead-Lag Publishing, LLC has been paid a fee. Lead-Lag Publishing, LLC does not guarantee the accuracy or completeness of the information provided in the episode or make any representation as to its quality. All statements and expressions provided in this episode are the sole opinion of Sparkline and Lead-Lag Publishing, LLC expressly disclaims any responsibility for action taken in connection with the information provided in the discussion. The content in this program is for informational purposes only. You should not construe any information or other material as investment, financial, tax, or other advice. The views expressed by the participants are solely their own. A participant may have taken or recommended any investment position discussed, but may close such position or alter its recommendation at any time without notice. Nothing contained in this program constitutes a solicitation, recommendation, endorsement, or offer to buy or sell any securities or other financial instruments in any jurisdiction. Please consult your own investment or financial advisor for advice related to all investment decisions.
