AI: Fundamentals, Valuation, and the Next Allocator Dilemma
I’m sitting in my classroom in disbelief.
Five years of training in value investing and a year-and-a-half at business school led me to a class called “Managing the Marketspace.” The old marketplace of revenues, margins, and cash flow driving shareholder value had suddenly been replaced by clicks and eyeballs.
Shortly thereafter, I attended a wedding and sat next to someone working at a technology company. The business had just gone public, sporting a $3 billion market cap and $3 million in revenue. The more questions I asked, the more confused I became with the answers.
Eventually, that recent Wharton graduate turned to me in frustration and said, “You just don’t get it.”
He was right. I couldn’t see the future or understand the present.
It was the spring of 2000.
A few months later, the dot.com valuation bubble burst, but the internet-powered economy roared on. Maybe I was proven right, or maybe I was early and wrong. That’s what made the period so difficult to navigate.
The enthusiasts were right about the technology.
The skeptics were right about prices.
Here we are again with AI.
AI is the next revolutionary technology and a centerpiece of every investment conversation. I won’t pretend to know how the technology, business models, or capital markets will play out, but it’s hard not to think about AI these days.
I tend to see the world through probabilities rather than certainties. Consistent with that thread, I see two sides of the AI discussion across investment prospects, winners and losers, and the next big allocator challenge.
Fundamentals vs. Prices
The AI supply chain is experiencing unprecedented adoption, revenue growth, and capital expenditure. The fundamentals of frontier models, compute, infrastructure, energy demand, and capital formation are off the charts.[1]
At the same time, the valuations of both public and private companies imply these trends will continue, creating unprecedented growth, returns on invested capital, and future profits.
Two recent podcast guests capture the two sides of the debate. Gavin Baker from Atreides describes the AI revolution as one of the most extraordinary moments in the history of capitalism. He sees real demand constrained by the supply of watts and wafers, compelling returns from productivity gains, and a market that underestimates the durability of AI spending.[2]
On the other hand, Rajiv Jain from GQG is avoiding hyperscalers and AI-related businesses. He worries about downside risks from massive CapEx without free cash flow follow-through, a lack of pricing power, and extreme valuations. While he agrees that AI is a revolutionary technology, he is skeptical that today’s business fundamentals justify today’s prices.
Both may be correct.
Much like the internet, AI may transform business and become ubiquitous throughout the global economy. But it’s also possible that markets have already priced in a decade or two of progress, just as happened with Amazon and Microsoft in 2000. Those companies ultimately fulfilled enormous expectations, but investors who bought before the bubble burst still endured years of disappointing returns.
The question isn’t whether AI matters. It’s how much of that future is already reflected in today’s prices.
Winners and Losers
During the internet boom, investors didn’t have to distinguish winners from losers. Everything went up.
The hard work started after the boom. Amazon became one of the most valuable businesses in history. Pets.com and hundreds of other online retailers vanished. The internet transformed the economy while simultaneously destroying enormous amounts of capital.
AI may prove similar.
Today, capital is abundant for private AI companies at every stage and in every layer of the stack – models, infrastructure, applications, tooling, and services. But capitalism eventually forces distinctions. Not every company can be a winner.
Technology can succeed spectacularly while many investments fail to meet expectations.
From Access to Positioning – The Allocator Dilemma
For allocators, this distinction matters for another reason. The biggest winners increasingly sit inside institutional portfolios, creating a different challenge than simply deciding whether AI is real.
Allocators are moving from how to gain exposure to AI to how to position portfolios when access is no longer an obstacle.
The power law phenomenon in venture capital is more pronounced than ever. The ten largest venture-backed winners are worth around $2.5 trillion and represent over 50% of venture capital value.[3]
Many institutional investors participated in these businesses through early, mid, and late-stage private investments. A few fortunate endowments are reported to have 10-15% of their entire pools in SpaceX.[4]
For years, “private for longer” created allocation imbalances across public and private markets. LPs rode extraordinary winners without much agency to adjust their portfolios. Ed Grefenstette spoke about this set-up on a recent podcast, articulating the unsatisfying options of selling in the secondary market at a discount, slowing venture commitments, or adjusting asset allocation targets to adjust for mega-cap venture winners.
But as mega-cap private companies begin offering meaningful liquidity to public investors, allocators face a new challenge: what do you do when yesterday’s private market winners become tomorrow’s public market index constituents?
A few years ago, in “Active Management Today is a Single Decision,” I wrote that active management came down to the choice of how much exposure to hold in the Mag 7. Replace “Mag 7” with “AI” (overlap acknowledged), and the dynamic isn’t much different. AI may be real, valuations may be high, and allocators still have to decide what and how much to own without hiding behind the constraints of long-ago seeded private investments.
The next decade may not only determine the winners and losers of AI. It may determine how allocators perform relative to each other. That’s a question of technology, economics, and market structure all at once.
The Hard Part Comes After Being Right
AI technology may exceed our expectations, while many investments fail to meet them.
The challenge is now deciding how much of that future is already reflected in today’s prices, and how much exposure to own when access is no longer a bottleneck.
Twenty-six years after that wedding conversation, I suspect the lesson is the same. The latest and greatest technology will exceed our lofty expectations, many investments will not, and both can be true at the same time.
The future is no longer the hard part to imagine.
The hard part is deciding what it’s worth today.
[1] No need for me to create a chart or add to the statistics. You can use Google or your favorite LLM as well as I can.
[2] Gavin was amazing on Capital Allocators; it’s his discussion on Invest Like the Best that walks through the recent case for AI.
[3] https://pitchbook.com/news/reports/2025-annual-us-vc-valuations-and-returns-report
This really captures the shift from "do I have access to AI?" to "how much should I actually own now that it's everywhere." The dot-com parallel lands, enthusiasts right on the tech, skeptics right on the price, and both true at once. I've been wrestling with the same tension from the operator side, building my own AI tooling to size and rebalance exposure rather than guess. Wrote up how I approach it here if useful: https://thiagopedicosaragiotto.substack.com/p/building-my-own-financial-ai-tools
Great piece and the comparison to the internet / dot com era is spot on.