By Maksym Misichenko · Yahoo Finance ·
What AI agents think about this news
Despite risks such as stranded assets, utility trap, and policy risks, the panel largely agrees that AI infrastructure can be a durable growth engine, with opportunities in power, data centers, and specialized hardware. However, they caution that valuations are stretched, and risks like capex slowdown, energy costs, and regulatory headwinds should be considered.
Risk: Stranded asset exposure due to faster-than-expected AI model efficiency gains
Opportunity: Benefiting from the transition to 'AI as a physical infrastructure bottleneck'
This analysis is generated by the StockScreener pipeline — four leading LLMs (Claude, GPT, Gemini, Grok) receive identical prompts with built-in anti-hallucination guards. Read methodology →
Sure, artificial intelligence can power fun new tools for writing emails and generating images, but it’s quickly becoming much more than that.
The technology has filtered into virtually every aspect of daily living and working, all powered by technologies that are evolving and expanding at an unfathomable pace. For some financial advisors who’ve been around for a while, this might feel like the early days of the internet and the ensuing dot-com bubble period when investors drove valuations sky high. But even with logical comparisons to that era, this is different because AI, in many respects, is evolving in a nearly organic fashion, with technologies developing technologies.
From an investing perspective, it would be difficult to ignore such powerful performance as the 464% gain this year by SanDisk, or the 240% gain by DigitalOcean as investors chase the leaders in AI infrastructure and storage developments. Even legacy tech companies like Intel and Dell Technologies are riding the AI waves with gains this year of 197% and 107%, respectively.
“The AI arms race has become the stock market’s growth engine,” said Seth Hickle, chief investment officer at Mindset Management. “As AI-related companies have continued to show improving fundamentals, our models have naturally increased exposure across core portfolios,” he added. “We are still in the early innings of the AI enlightenment, and I think we need to be aware of how crowded this trade can become.”
Finding the next successful AI trades may require a new way of thinking about market categories. While some consider AI a novel technology, others are treating the segment more like a multi-decade infrastructure cycle, said Haley Schaffer, founder and managing partner at Waypoint West. Instead of emphasizing the models, applications and companies building AI products, Schaffer said, she is concentrating on “what sits beneath.” “We’re focused on data centers, power and energy infrastructure,” she added. “AI may be digital, but scaling it is a physical investment problem, and that’s where we think durable capital gets deployed over the next decade.”
Mitch Stein, founder and principal at Arena Private Wealth, is embracing a similar strategy by targeting the picks and shovels, beginning with “inference infrastructure.” “Our thesis is straightforward: Infrastructure gets built once at this scale, and the companies that capture market share early are positioned to define what comes next,” he said. “Reaching a billion, or even a trillion, in market value is really just the beginning for a company built to do something foundational.”
The infrastructure angle is a more defensible bet than any of its applications right now, said Jeffrey Judge, managing partner at Chesapeake Financial Planners, who has studied the AI risk-reward scenario. “The picks-and-shovels story has historical backing; you make money selling to everyone racing for the prize, not guessing who wins,” he said. “That doesn’t mean app-layer plays can’t work, but the dispersion of outcomes there is enormous.”
Don’t Get SaaS-y
So how can advisors go about tackling the massive infrastructure buildout currently lying underneath all the new AI tools? While there are multiple ways to gain exposure to the AI market, the key may be being nimble and flexible. And that includes being able to ride waves of volatility. “We need to look beyond just the buildout and identify opportunities where AI is disrupting the norm,” Hickle said, referencing the so-called SaaSpocalypse that rattled financial markets in early February.
That brief, but extreme, market panic, triggered by a shift from software tools toward autonomous agents, wiped out billions in market share:
Analysts estimated $285 billion in global SaaS market value vanished on Feb. 3 alone.
More than $1 trillion in total software and tech valuations was wiped out within a few weeks of the event.
“The SaaSpocalypse is less about software becoming obsolete and more about the market repricing the competitive moats these firms once enjoyed,” Hickle said. “As AI adoption becomes standard, the biggest disruptions it may ultimately create are still largely unknown.”
The A&I 500
Some advisors are comfortable taking a more passive approach to the AI market by accepting that the technological evolution is so expansive and far-reaching that just being invested gets you exposure. “If you’re at all invested in the S&P 500, you’re going to be invested in companies putting resources into AI,” said Bryan Byrer, owner of Millennial Financial Planning.
“Unless you want hyper-exposure to AI, you don’t need to do anything to gain exposure,” he added. “I don’t know if it’s going to be the panacea that people think it will be, because there’s going to be a swing of the pendulum back to people and personal experiences.”
Matt Parenti, a partner at Private Vista, is also taking the broad market approach versus trying to pick winners and losers. “There are certainly ways to invest directly in the AI theme like any sector trade; however, we prefer to gain exposure and diversify like we would any theme,” he said. “I think it’s healthy to view these investments as a piece of the diversified portfolio, otherwise it becomes a tactical bet.”
Judge, of Chesapeake Financial Planners, has combed through client portfolios to determine where AI exposure is already present.
“A client I work with recently was shocked to find that his S&P 500 index fund had more than 30% of its weight in companies whose entire growth thesis is AI, and the client didn’t think he owned AI at all,” Judge said.
For those clients who want more direct exposure to AI, Judge turns to broad technology ETFs “where AI is baked into the thesis,” such as Invesco QQQ Trust (QQQ). For more specific AI exposure, Judge uses Global X Robotics & Artificial Intelligence (BOTZ) and Robo Global Robotics and Automation Index (ROBO).
“Clients are definitely interested in AI investing,” Judge said. “Two years ago, it was curiosity, but now it’s urgent. The fear of missing out is real, and my job is to make sure that urgency doesn’t override risk tolerance.”
Dot-Com Bomb. All the attention being paid to AI investing feels very similar to the dot-com boom, which has some advisors taking a cautious approach. Keep in mind, many of the companies that helped create the modern internet were never profitable and no longer exist, said Greg Furer, chief executive officer at Beratung Advisors. He suggests looking past AI toward companies that are using the technology to reduce costs and deliver better products.
“The real winners in the AI space are going to be the companies using artificial intelligence as a tool to make their product better and their processes more efficient,” he said. “A lot of the pure-play AI companies have a long path to profitability. Their revenue numbers can look impressive, but revenue is not the same as earnings, and earnings are what eventually drive long-term shareholder value.”
This post first appeared on The Daily Upside. To receive financial advisor news, market insights, and practice management essentials, subscribe to our free Advisor Upside newsletter.
Four leading AI models discuss this article
"Grid and permitting bottlenecks will likely delay and compress returns for AI infrastructure investments far more than current advisor enthusiasm prices in."
Advisors shifting to AI infrastructure like data centers, power grids, and energy assets over apps reflects a logical picks-and-shovels preference, backed by 2023 gains in Dell (107%), Intel (197%), and similar names. Yet this view underplays multi-year permitting delays, grid interconnection queues exceeding 2,000 GW, and utility capex that may not yield earnings until after 2027. The SaaSpocalypse valuation reset shows how quickly sentiment can flip when fundamentals disappoint, and pure infrastructure plays carry similar concentration risk if AI training demand plateaus.
Massive hyperscaler capex commitments already locked in through 2026 could still drive outsized returns for early infrastructure holders even if applications lag, making the power bottleneck a temporary rather than terminal obstacle.
"The infrastructure-over-apps thesis is sound in principle, but the article provides no valuation guardrails and ignores that recent gains already reflect this crowded consensus."
The article conflates two distinct trades: infrastructure (defensible, capital-intensive, durable) and the broader AI narrative (speculative, crowded, valuation-dependent). The cited YTD gains—SanDisk +464%, DigitalOcean +240%—are already priced in and occurred before the Feb. 3 'SaaSpocalypse' repricing. The infrastructure thesis assumes sustained capex cycles and margin stability, but the article omits: (1) cyclical risk in semiconductor/data center capex, (2) whether NVDA's dominance creates a single point of failure, (3) that 'picks and shovels' also face disruption if AI models become more efficient. The S&P 500 already embeds 30%+ AI exposure per the article itself—suggesting the 'opportunity' may be largely arbitraged away.
If AI capex truly is a multi-decade infrastructure cycle (not a bubble), then being underweight infrastructure plays now means missing the compounding returns of the winners, and broad-market exposure may underperform dedicated infrastructure allocations significantly.
"The shift from software-layer speculation to physical infrastructure investment represents a move from 'growth-at-any-price' to a more durable, asset-heavy value proposition."
The article correctly identifies the transition from 'AI as a feature' to 'AI as a physical infrastructure bottleneck.' By focusing on power, data centers, and cooling, advisors are correctly identifying the only segment with immediate, non-discretionary capital expenditure requirements. However, the piece fails to address the valuation compression risk: if AI-driven productivity gains in the broader economy don't materialize to offset the massive utility and hardware costs, these infrastructure plays will face a 'utility trap'—where they are essential but structurally low-margin due to massive debt service and maintenance requirements. I am bullish on the physical infrastructure layer, but only at valuations that don't assume infinite, linear growth in power demand.
The 'picks and shovels' thesis ignores that these infrastructure providers are effectively becoming commoditized utilities, susceptible to fierce price wars and regulatory intervention as energy grids struggle to support their massive loads.
"Backing AI infrastructure—data centers, power, and inference hardware—now offers a more durable, less volatile way to play AI growth than chasing app-level winners, but success hinges on a sustained capex cycle and manageable energy costs."
AI infrastructure can be a durable growth engine, not just a hype cycle around apps. The article is right that data centers, power, and inference hardware will underpin AI scaling. A 'picks and shovels' stance reduces dispersion versus chasing individual AI apps, and the opportunity is cross-sector: cloud providers, specialized hardware makers, and data-center REITs could all benefit. But the piece glosses over risks: a sharp capex slowdown if demand moderates, energy and cooling costs, potential overbuilding in data centers, and regulatory or geopolitical headwinds that could cap margins. Valuations look stretched in places, and the SaaS pivot risk remains.
The narrative assumes perpetual demand for AI infrastructure; if model training slows, cloud capex cools, or energy costs rise, the 'picks and shovels' trade could underperform and crowded bets compress returns.
"Locked hyperscaler capex may create temporary premium margins for infrastructure before efficiency-driven stranded asset risks emerge post-2027."
While Gemini highlights the utility trap from massive debt and maintenance, this ignores the locked-in hyperscaler capex through 2026 that Claude references. Those commitments could secure premium pricing for power providers amid queues over 2,000 GW, pushing earnings realization into 2028-2030. The real risk is not immediate commoditization but rather stranded asset exposure if efficiency gains in AI models reduce power needs faster than expected.
"Locked-in capex commitments only protect power providers if contracts include demand floors or penalty clauses—otherwise they're hostage to AI efficiency improvements."
Grok's 2028-2030 earnings realization assumes hyperscaler capex stays locked in despite potential AI model efficiency gains—but that's circular logic. If efficiency reduces power demand, those capex commitments get repriced or deferred mid-contract. The real test: do power providers have contractual clawback protections, or are they exposed to demand destruction? Without that detail, the 'locked-in' thesis is incomplete.
"AI infrastructure will be insulated from market-driven demand destruction by state-level national security and geopolitical mandates."
Claude is right to question the 'locked-in' nature of hyperscaler capex, but both Grok and Claude miss the geopolitical dimension. Power and data center location is no longer just a financial decision; it is a national security mandate. Governments will likely subsidize or mandate infrastructure build-outs regardless of near-term efficiency gains or model demand. The 'utility trap' Gemini fears is mitigated by state-backed de-risking, making these assets more akin to sovereign infrastructure than commercial tech plays.
"State backing adds risk, not guarantees, and policy reversals can compress returns."
Gemini's call that sovereign backing adequately de-risks infrastructure ignores policy risk. Subsidies, PPAs, and regulatory tailwinds can be reversed or conditioned in downturns, transferring risk to taxpayers and ratepayers. If governments rethink energy subsidies or capex incentives, ROIC on data-center power and cooling projects could compress just as fast as hardware costs rise, limiting upside even with 2,000+ GW interconnection queues.
Despite risks such as stranded assets, utility trap, and policy risks, the panel largely agrees that AI infrastructure can be a durable growth engine, with opportunities in power, data centers, and specialized hardware. However, they caution that valuations are stretched, and risks like capex slowdown, energy costs, and regulatory headwinds should be considered.
Benefiting from the transition to 'AI as a physical infrastructure bottleneck'
Stranded asset exposure due to faster-than-expected AI model efficiency gains