What’s moving in the markets
Are we in an AI bubble?
Two reasons this cycle is different.
The numbers are hard to ignore. Data centre investment is surging, semiconductor stocks are on a tear, and the comparisons to past hype cycles are getting louder: railroads, the internet, you name it. So is this time different, or are we just telling ourselves that?
Hyperscalers are spending more on AI investments than ever
We think there are two meaningful differences worth paying attention to.
The returns are already showing up. Yes, cash flows are being squeezed by historic levels of capital expenditure. But revenues and earnings at the hyperscalers are growing at the fastest rates ever recorded; remarkable for companies already this large. And it's not contained to the Googles and Microsofts of the world. Across the S&P 500, companies implementing AI strategies are seeing real improvements in earnings and productivity. Not just headcount reductions, either. New products are shipping faster, drug discovery is accelerating, data analysis is sharper…
The spending is chasing demand, not hoping for it. Previous bubbles were built on the anticipation of future demand arrived way later than expected. Today, the hyperscalers are supply-constrained: they cannot build fast enough to keep up with what customers are already asking for.
This doesn’t mean the market is bubble-free. Take memory. The Roundhill Memory ETF (DRAM), which tracks the memory chip industry, is up 25% since we mentioned it in last week’s newsletter. And we're starting to hear the words no investor should ever feel comfortable with: "It really is different this time." The argument being made is that memory businesses are transitioning from cyclical, boom-and-bust commodity sellers into long-term structural beneficiaries of AI demand.
Famous last words.
Buybacks are not an investment strategy
When cheap isn't cheap enough.
A value trap looks attractive on paper. The metrics suggest a bargain (like a low price-to-free-cash-flow) but the stock keeps going nowhere, or worse, keeps falling. PayPal and Fiserv are starting to look like textbook examples.
Both reported earnings last week, and neither made for comfortable reading. PayPal grew sales year-over-year, but margins and cash flows declined. Fiserv was worse: sales, margins, and cash flows all down. Both stocks fell 11% in a single week.
What makes this worth examining is that both companies are trading at what looks like a head-turning cheap valuation. PayPal sits at roughly 9.5x free cash flow. Fiserv is closer to 7x. On those numbers alone, you're buying a significant stream of cash for a low price. Which is, in theory, exactly what investing is supposed to be.
The bull case has been straightforward: low multiples plus high free cash flow equals aggressive buybacks, which shrink the share count quickly and amplify returns for remaining shareholders. The problem is that it only works if the free cash flow holds up.
Investors have been calling PayPal cheap since 2022
Think of it this way. Every year, you receive a few new ice cubes to add to your collection. The only issue is that every year, each cube is slightly smaller than the last. You keep adding, but the total amount of ice keeps shrinking. That's the melting ice cube problem, and no amount of new cubes changes the direction of travel.
If the FCF denominator keeps falling, and the market expects it to keep falling, then buybacks become noise. You're reducing the share count on a business that's worth less every year. The math doesn't work.
The real question for PayPal and Fiserv isn't whether they're cheap. It's whether cash flows can stabilise. Because here's the thing: at 7x free cash flow, all it takes is for the bleeding to stop. If cash flows flatten and hold, the upside from here could be spectacular. But until there's evidence of that turn, cheapness alone is not a thesis. Neither are buybacks.
Other Updates
Ratings
London Stock Exchange Group (LON:LSEG) got a rating downgrade
“Like many businesses in the financial data space, LSEG finds itself at an inflection point. On one hand, the company is adapting intelligently: partnering broadly, embedding its data into the platforms clients already use, and positioning itself as the trusted data layer for an AI-powered financial industry. On the other hand, the economics of this new distribution model remain largely unproven, and the competitive response from well-funded rivals is still taking shape. It is clear there is more uncertainty in the system and we believe the multiple coming down is justified to an extent.”
Click here to view the full update.Walmart (WMT) got a rating re-affirmation
“Walmart’s current valuation appears to be pricing in a very rosy scenario: AI-driven efficiency gains, the advertising and membership businesses scaling smoothly, and the higher-income customer cohort sticking around permanently… A business priced for high-quality growth tends to be punished harshly when the narrative wobbles, even temporarily.”
Click here to view the full update.
Dividend Cuts / Raises
Paychex (PAYX): 10% raise
FactSet (FDS): 5% raise
KLA (KLAC): 21% raise
Upcoming Earnings
Tuesday
Imperial Brands (H1)
Wednesday
Constellation Software (Q1)
Thursday
Applied Materials (Q2)
Want to see what we’re actually doing with real money?
Upgrade to unlock full access to the RatedA Portfolio (with weekly updates on every buy/sell and rationale), and the complete database of companies + ratings on www.rateda.co.
