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

Costco Wholesale

• Chris Davis · Davis AdvisorsBullish

  Costco's runway was underestimated: more stores, private label and international growth beyond what he modeled.

  Davis says Davis Funds sold Costco after its multiple rose from about 15 to 16 times to 28 times earnings, which looked too high for a non-high-growth retailer with limits on saturation. He admits this was a mistake. Munger warned they were unlikely to find a business this good again. The firm's store-count model was 'wildly too low,' because Costco stores could sit closer together than expected. They also underestimated private label as a higher-margin business that built the Kirkland Signature brand. They did not foresee Costco thriving internationally in Japan, Australia and Taiwan.

  “Our our models of how many stores there could be were wildly too low. ... It ended up you could put Costco's closer together than people realize. ... we never thought private label could be a higher margin business than their traditional business”at 27:49

• Chris Davis · Davis AdvisorsBullish

  Costco's thin margins make it unattractive to compete with, a durable source of its moat.

  Davis says Costco is a recurring example for him, since he made the mistake of selling it at around 80. He argues one source of its competitive advantage is that it has made the business unattractive for anyone to come in and compete. Its margin is about 2%, so there is little profit left for rivals to pursue. He uses this as a contrast with companies that maximise near-term monetisation, which he fears could weaken their moats.

  “I keep harping on, you know, my experience with Costco because it was such a mistake selling it at 80. But one of the sources of their competitive advantage was they made it so unattractive for anybody to come in and compete with them.”at 50:56

Amazon.com

Across the shows: 2 speakers in other episodes, 0 bullish and 2 bearish calls, past 30 days

• Chris Davis · Davis AdvisorsBullish

  Amazon's retail business has outgrown what models predicted, with private label and advertising adding leverage.

  Davis says his team modeled Amazon's retail business by following the Costco playbook, and was wrong because there could be more stores and more international locations. They never anticipated that Amazon's private label business would become strong enough to give it leverage over credit card companies, so they would pay Amazon to take a card. He says the model struggles to capture a dynamic management team with a great culture that adapts in ways no one expected. He notes Amazon's path from books to other categories, then fulfillment and advertising, as evidence it keeps finding new profit pools.

  “But we also never anticipated that the private label side could become so good, that they could gain so much leverage over credit card companies that they would be being paid to take somebody's credit card”at 29:58

• Chris Davis · Davis AdvisorsBearish

  Amazon's ad business risks degrading the customer experience and narrowing its moat.

  Davis says he thought it was a terrible mistake for Amazon to get into advertising, though he concedes his colleagues would say the ad business prints money. His core point is that Amazon is supposed to be the world's most customer-centric company, yet product recommendations now reflect what a paying party wants shown. He calls that a drug and says he has made the point to Jeff Bezos and Andy Jassy. They argued the ad dollars fund the reinvestment needed to keep the advantage, and Davis concedes Bezos has historically made better calls. He still worries that the companies will push so hard to monetise that they degrade the service, leaving the moat narrower for an insurgent competitor.

  “I thought it was a terrible mistake for them to get in the advertising business.”at 52:32

Tesla

Across the shows: 2 speakers in other episodes, 0 bullish and 1 bearish calls, past 30 days

• Chris Davis · Davis AdvisorsUncertain

  Tesla's value may hinge on licensing its self-driving software to other automakers, which he can't model as a car maker.

  Davis says he could not get the math to work for Tesla valued as a car manufacturer, since its share is too small to justify the price. As a thought experiment, he models a world where Toyota or Hyundai give up on self-driving and pay Tesla around $10,000 per car for its software, with buyers subscribing at roughly $2,300 a month for the feature. He says that if Tesla made $10,000 of profit on other automakers' cars, the software business would be a 'total gate.' He stresses he has never heard Tesla say it plans this and is only making the case up as a premortem exercise. He says Davis does not own Tesla.

  “Well, all of a sudden, you do begin to see, well, if they were able to make, you know, $10,000 of profit on everybody else's cars, then it's a total gate. Now I've never heard them say that they're gonna do that, so I'm not... I'm just making that.”at 31:34

Meta Platforms

Across the shows: 8 speakers in other episodes, 12 bullish and 3 bearish calls, past 30 days

• Chris Davis · Davis AdvisorsBullish

  Meta's ad business has exploded in value because of its AI investments; a good return on spending.

  Davis says Meta's AI investments have made its existing ad business far more valuable, so each reinvested dollar earns a return. He notes the same is obviously true at Google. He argues this reduces risk because these companies are generating enough cash that they do not need to issue lots of shares or debt, or enter convoluted transactions, to fund the buildout. He sees it as exposure to Hagstrom's scenario of rising returns on capital, and his preferred approach is owning companies that generate cash and reinvest it into their existing businesses.

  “The ad business at Meta has exploded in value because of the AI investments, obviously true at Google and so on.”at 46:58

Capital One Financial

• Chris Davis · Davis AdvisorsBullish

  Capital One is the most successful fintech ever yet trades at only 9-10x earnings.

  Davis says he reread Rich Fairbank's 2025 annual report for Capital One and was struck by how far ahead the company is. He calls it the most valuable and successful fintech company in history, still run by its founder. Yet it trades at nine or 10 times earnings. The report also spells out what it takes to be a business capitalising on AI, which Davis sees as the big question for companies. His point is that the valuation looks low given the company's position and its AI capability.

  “it's just amazing how far ahead they are as the most valuable and successful fintech company in history, still run by the founder, and yet it trades at nine or 10 times earnings.”at 47:46

NVIDIA

Across the shows: 11 speakers in other episodes, 12 bullish and 3 bearish calls, past 30 days

• Chris Davis · Davis AdvisorsBearish

  Nvidia keeps pricing up so hard it could erode the customer loyalty behind its moat.

  Davis worries that Nvidia keeps raising prices so much that it is taking every last dollar it could. He contrasts this with TSMC founder Morris Chang's ethos that a company should earn only a fair margin. He notes Nvidia could double or quadruple earnings by raising prices immediately, yet it appears to be pushing margins as hard as it can. He questions whether the chip customers are truly its friends, since Nvidia sells chips to them and then buys them back. He treats this as a risk to long-term competitive advantage in the current stampede.

  “NVIDIA, you know, continues to price up so much that as I recall, Morris Chang, you know, who built a company with a totally different ethos. His ethos was we need to earn a fair margin.”at 51:33

AI infrastructure & data centers

• Robert Hagstrom · EquityCompass Investment Management

  Hyperscaler capex may peak around 2H 2027, after which cash flow and returns on capital rebound.

  Hagstrom argues the hyperscalers' heavy capital spending may be cyclical rather than permanent. He cites Michael Mobison's free-cash-flow work, which suggests the second half of 2027 may be the peak of this reinvestment cycle, with cash flow yields and returns on capital then recovering. He says the forecasts are management's, not inside information, and that the analysis is provisional. He adds that the hyperscalers could become cash-flow machines within two to three years as capital intensity falls and subscription and rental revenue grows. He compares the buildout to Amazon's 1990s distribution-center spending, which critics called unsustainable. He also argues the AI buildout is spread across a much larger global GDP pie than the US railroad buildout. He calls the timing of the top of this capex cycle the most important data point for next year.

  “We may not be in a permanent condition with the hyperscalers, that they're gonna be capital intensive businesses forever. They will return back to being cash flow monsters.”at 42:12

• Chris Davis · Davis Advisors

  If AI cloud has four or five low-switching-cost providers, value flows to users, not builders.

  Davis's central worry is whether the AI infrastructure builders will have pricing power. He contrasts winner-take-all network-effect businesses, like Amazon's retail and cloud feedback loops, with a market of four or five cloud providers and low switching costs. In that market, customers can pick and choose, so the tool gets built but its value accrues to users rather than builders. He compares this to railroads, which were built on speculation, and says the question is whether the builders can earn pricing power on the tools they create. He says it is too early to know. He ties this to AI model pricing, asking how much companies can charge for access to their models and whether they have other addictive businesses if that is lower than expected.

  “if we end up in 2028 or 2029 with four or five cloud providers and low switching costs and we can pick and choose what works for us, we are going to have built an incredibly valuable tool, but the value of that tool is gonna go to the users of it rather than the builders of it.”at 45:04

• Chris Davis · Davis Advisors

  Opaque circular and vendor financing in AI deals makes Davis nervous.

  Hagstrom asked about the circular financing in AI, citing Google's $10 billion investment and the deals among AI companies and chip vendors. Davis says he cannot speak for the directors involved and that they may be comfortable with the risk-reward, so the picture is hard to generalise. He says the opacity of some of these financing deals makes him nervous and recalls vendor financing, where equipment makers lend money to customers to buy their equipment and generate growth. He ties the risk back to how much companies can charge for access to their AI models.

  “the opacity of some of the financing deals, of course, makes me nervous, and we all remember vendor financing. And equipment companies lending money to companies to buy their equipment, generating huge growth.”at 57:18

• Matt Zeigler · Excess Returns

  AI buildout spending has a finite limit; high valuations likely slow growth or stall.

  Zeigler argues the heavy capital spending on AI infrastructure must eventually hit a boundary. Companies spend large sums building out capacity, and if their valuations are pushed sharply higher, that will likely at least slow the rate of growth, or leave the stocks trading sideways for a long time. He says he has been thinking about this most. He also notes that some of these businesses benefit from the infrastructure being in place, and that recent agent and tool use cases are making the benefit feel closer. He frames the cycle as one that will come in its own waves.

  “there has to be some finite boundary on the other side of this. I don't think we just keep on expanding this. Maybe we do, but that's gonna come in its own cycles.”at 1:14:32

• Bogumil Baranowski · Blue Infinitas Capital

  Some AI infrastructure firms may return to high margins and low capex; current spending isn't permanent.

  Baranowski argues that investors, like the one he describes making the common mistake of projecting the most recent past into the future, are looking at the huge current capex numbers and assuming they are the permanent way of doing business. He says the AI wave is a short-lived cycle, though he cannot say how long. He believes there is a chance, of unknown size, that some AI-related companies will go back to very high margins, very high returns and a low capex model, though not necessarily all of them. He cites Robert Hagstrom's and Chris Davis's views that this is not forever as encouraging.

  “I think there is a chance, I don't know how big, that they will go back to very high margins, very high returns, and a low CapEx model. Some of them.”at 1:17:46

US stock market outlook

• Chris Davis · Davis Advisors

  Indexing is riskier than ever; it has become a large momentum fund, risky at the end of a bull market.

  Davis argues indexing has worked so well that people assume it always will, and that the world reacts to it. He says indexing amplifies momentum: the more a stock rises, the bigger its index weight, and the more money moves from active to passive, the more that effect is amplified. He calls the index 'a very large momentum fund' and says buying a momentum fund at the end of a bull market is a bad idea. He adds that as private giants such as SpaceX, Anthropic and ChatGPT enter the index, the index multiple will be much higher and sector concentration more amplified. He ties this to a warning that widows and orphans could get hurt.

  “I certainly think it's an enormously riskier position than it's ever been. You know, 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.”at 15:29

• Robert Hagstrom · EquityCompass Investment Management

  Tech overweight no longer justified: margin of safety has shrunk, so he cut tech to market weight.

  Hagstrom says his firm sold ASML after it rose 500%, and trimmed Google and Nvidia as they reached the 10% limit. Margin of safety declined as the market accelerated in mid-2026 during the Iran war episode, so payoff shrank and sizing was reduced in line with a Kelly-type logic. Tech was about 45% of the portfolio at the second-quarter change, now about 26% pure tech or 35% with Google and Amazon, which is market weight. He says the overweight was right for the prior two years but he is not sure it is right today, although clients like the AI tech boom and want exposure without being over their skis.

  “I'm just not so sure it is the right thing to do today. So being a market weight here based upon margin of safety and and 10% threshold has actually in some ways helped us out, I think.”at 24:07

Venture capital & startups

• Chris Davis · Davis Advisors

  VCs keep big private companies private to exit into price-insensitive index buyers, a dangerous setup.

  Davis recalls that in the 1990s, VCs got burned after bringing companies public: lockups ended, active managers found no business and the companies collapsed, leaving VC funds stuck with paper gains. His self-described 'tinfoil hat' theory is that today's VCs keep companies private longer because their best exit is a price-insensitive buyer, the index, which will automatically buy 50% or 60% of a company whatever the price. He says VCs own 80 to 90% of these companies, so the index provides a ready buyer. He links this to the index becoming a momentum vehicle that could hurt investors.

  “My really dark tinfoil hat conspiracy theory is that what's happening now is these companies, the venture capitalists are keeping them private for longer because they realize their best possible exit strategy is to a price insensitive buyer”at 17:22

AI bubble & tech valuations

• Robert Hagstrom · EquityCompass Investment Management

  Many tech firms' high returns on capital justify a premium multiple, but its size is hard to pin down.

  Hagstrom argues two companies with the same growth and earnings should not trade at the same multiple if one earns a 50% return on capital and the other 12%. He says many of the technology companies he is looking at have amazing returns on capital, so based on growth and returns on capital they deserve to trade at a premium to the market. He admits the premium is hard to quantify: a cash flow model gives a number, but high returns on capital add something extra. He notes it depends on forecasting how long high returns persist, which is a slippery slope.

  “it's so funny that many of the technology companies that we're looking at today have these really amazing returns on capital. And so you'd say to yourself, okay, they obviously deserve, between the growth rate and the return on capital, they deserve to trade at something premium to the market.”at 34:52

AI: overall outlook

• Chris Davis · Davis Advisors

  Too early to know AI winners; Internet-era surefire winners Yahoo, AOL and Cisco faded.

  Davis argues it is too early to say who wins in AI. He cautions against simply betting on the sure things, such as the semiconductor leaders and the model labs and cloud companies like Anthropic, OpenAI and Microsoft. He recalls that the surefire winners of the Internet boom were Yahoo, AOL and Cisco, which were head and shoulders ahead of everyone else, and that the eventual value went elsewhere. He says it is reminiscent of the Internet era. He also agrees that some chips on the table in companies reinvesting strong cash flows make sense.

  “the surefire winners of the Internet boom were Yahoo, AOL, and Cisco.”at 46:21