PodcastExcess ReturnsChris Davis

The Winners That Get Away | Chris Davis and Robert Hagstrom on Holding Great Businesses

Chris Davis and Robert Hagstrom discuss the challenge of holding durable winners while distinguishing lasting competitive advantages from temporary ones. They question how much value AI providers will capture, how long current advantages will last, and whether the data-center investment cycle may peak around 2027–2028. The discussion also weighs position sizing, fund constraints, valuation, and the risks of selling successful investments too early.

Investment StrategyArtificial IntelligenceMarket Valuations

Preview · 21 more points in the app

Read the Full Summary

Key Takeaways

Chris Davis says Amazon’s retail scale historically reinforced lower prices, shorter wait times, and broader selection, but he worries that aggressive advertising could weaken that customer-centric experience and make its moat more vulnerable to a rival. The contrast is between scale that strengthens the customer value proposition and monetization that could erode it.

The speakers caution that today’s visible AI leaders may not become its durable winners, as Yahoo, AOL, and Cisco did not define the eventual outcome of the internet boom; Robert Hagstrom says current advantages in data centers and Nvidia GPUs do not reveal how long high returns will last. Chris Davis also says AI investment has made Meta’s advertising business—and, he says, Google’s existing business—more valuable, illustrating that returns may accrue to established businesses as well as infrastructure providers.

The speakers describe AI capital spending as a cycle, not a level certain to remain in the hundreds of billions or more indefinitely, and Robert Hagstrom says forecasts suggest hyperscaler reinvestment may peak in the second half of 2027, though that remains uncertain. If spending peaks and declines, margins, returns on capital, and hyperscaler cash could improve; the speakers also say installed infrastructure and high valuations could leave suppliers with slower growth or an extended period of sideways performance.

Chris Davis argues that if four or five cloud providers remain in 2028 or 2029 and switching costs are low, customers could capture more of the technology’s value, limiting providers’ pricing power. He also says opaque AI financing makes him nervous, recalling vendor financing, and emphasizes that providers’ economics depend on what they can charge for AI access and what other valuable businesses they have if fees disappoint.

Robert Hagstrom says the global scope of the data-center buildout makes it less directly comparable with a U.S.-only railroad boom, while the speakers use the railroad analogy to distinguish building infrastructure from eventually using it. They also describe a possible second wave of AI tools and agents that could benefit average consumers, even if identifying investable winners during the buildout is difficult.

TECHNOLOGY

Amazon’s Network Effects — Chris Davis contrasts network-effect businesses with those that may lack durable advantages. He says Amazon’s retail business historically built a feedback loop: greater scale enabled lower prices, shorter wait times, and more selection than competitors.

“And one of the concerns I have is that Robert and I were at an impressionable age when we saw the incredible value created in winner take all businesses and…”—

Cloud Switching Costs — Chris Davis argues that if four or five cloud providers remain in 2028 or 2029 and switching costs are low, customers could capture more of the technology’s value than its builders. That scenario could limit providers’ pricing power.

“So they is the big word there because if we end up in 2028 or 2029 with four or five cloud providers and low switching costs and we can pick…”—

Durable AI Winners — The speakers caution that today’s apparent leaders in a technology boom may not become its lasting winners. Yahoo, AOL, and Cisco stood out during the internet boom, they recall, while the eventual AI winners may not yet be identifiable.

“But, you know, remember it was the surefire winners of the internet boom were Yahoo, AOL, and Cisco…”—

MARKETS

The Cost Of Selling — Chris Davis argues that trimming or selling a successful investment, sometimes to lock in gains, can undermine its long-term result. He cites Bill Miller’s Amazon investment, which investors panicked out of, and Naspers’ $32 million investment for a 46% stake in Tencent in 2001, which he says grew to $235 billion. That contrasts with venture capitalists who sold early after tripling their money.

“Your question says they buy a great business. And so, of course, the only thing that could go wrong is they sell it. And that's what goes wrong…”—

Fund Constraints — Chris Davis says the 40 Act mutual-fund structure imposes restrictions that do not apply to someone managing only a personal account. Personal tolerance for volatility may also change as a balance sheet grows. Managers of client money must account for investors who enter after strong returns and leave after poor ones. A smoother return profile may produce a higher end return for mutual-fund investors, despite Buffett’s preference for a lumpy 15% over a smooth 12%.

“And I think what it took me 30 years to recognize was that being in the structure of the 40act mutual fund imposes all sorts of restrictions that you wouldn't…”—

Indexing And Momentum — The speakers argue that indexing can amplify momentum because rising stocks gain larger index weights, and that a market dominated by passive strategies could weaken price discovery. Chris Davis warns that feedback effects may eventually produce an unpredictable, potentially dramatic collapse. He also offers a speculative theory: venture capitalists may keep companies private longer because index funds could become price-insensitive buyers of large stakes.

“You say you get in more trouble with a good premise than a bad premise and indexing is a good premise and it worked for a long long time and…”—

FORECASTS

Hyperscaler Spending Outlook — Robert Hagstrom says management forecasts and analysis suggest the hyperscaler reinvestment cycle may peak in the second half of 2027. Cash-flow yield and return on capital could then recover, though he stresses that this remains uncertain until the forecasts are borne out.

“But it looks to me Michael Moesson wrote this in his latest piece on free cash flow and says it looks like the bottom of 2027, second half of 2027…”—

Holding great businesses and the case for position limits

Davis says selling a truly great business is often the biggest investment mistake, while position limits reflect both fund regulations and the need to protect clients from volatility-driven selling.

Robert Hagstrom on concentration and client risk

Hagstrom describes adopting a 10% limit amid client and regulatory constraints, while Davis and Hagstrom emphasize that even exceptional investments like Tencent and Nvidia require the ability to withstand volatility.

Index concentration, momentum and IPO incentives

Davis warns that index funds amplify momentum and concentration, and argues that IPO incentives may let venture investors exit into passive funds even as benchmark-relative pressure encourages risky behavior.

Trimming winners, margin of safety and the Costco lesson

Hagstrom links trimming to shrinking margins of safety and position constraints; Davis’s Costco experience shows how conservative assumptions about store growth, private label, and international expansion can lead investors to sell winners too soon.

Valuation, return on capital and the cost of growth

Davis and Hagstrom explain that valuation depends on returns on capital and the cost of growth, since businesses that reinvest profitably can compound value differently from those that must spend heavily to achieve similar earnings growth.

Is AI’s capital spending cycle temporary?

Hagstrom argues that hyperscaler capital spending could peak around 2027–2028 rather than remain permanently elevated, potentially restoring cash flow and returns if data-center investment transitions into an established infrastructure cycle.

AI pricing power: builders versus users

Davis questions whether AI builders will retain pricing power if customers can choose among competing providers, contrasting network-effect businesses with tools whose value may accrue mainly to users.

Competitive advantage, margins and customer trust

The investors focus on how long competitive advantages last, warning that high margins can invite competition and that aggressive monetization may erode customer trust and leave room for new rivals.

Balance sheets and circular AI financing

Davis cautions that opaque, circular AI financing deserves scrutiny, while emphasizing integrated financial statements and balance sheets; Hagstrom sees no clear evidence of fraud but remains focused on who ultimately pays for AI.

Evaluating the AI investment cycle

Hagstrom weighs management forecasts, global data-center demand, and a possible 2027–2028 capex inflection; both investors see the eventual returns and customer demand as crucial tests of the AI cycle.

Matt and Bogumil’s takeaways: position sizing and AI

In their post-interview discussion, Matt and Bogumil reflect on the humility required in position sizing, the trade-off between letting winners run and managing client risk, and the value of incorporating uncertainty into investment decisions.

Being an AI user versus an AI investor

Matt and Bogumil distinguish enthusiasm for using AI from confidence about investing in its providers, noting that the spending cycle may end, future winners may be hard to identify, and AI could benefit customers more than builders.

Related Episodes

Important Disclaimers

Not Investment Advice: For education only — not financial, tax or legal advice.

AI-Generated: Made by AI from source transcripts and may contain errors.

No Guarantees: Past performance does not predict future results; investing carries risk.

Our Commentary · Request Removal