TIP838: Google, Reddit, Amazon – Are Our Biggest Winners Still a Buy? w/ Daniel Mahncke & Shawn O’Malley

Alphabet has been the portfolio's largest holding almost from the start and has nearly doubled from an average cost basis of roughly 150 to 190 dollars per share. The original thesis was buying a fantastic business at a fair price, with a sum-of-the-parts valuation placing fair value between 180 and 200 dollars and a forward price-to-earnings ratio of about 18 times, at a time when the dominant narrative was that ChatGPT had made Google search obsolete. The doubling cannot be explained by multiple expansion alone because operating cash flow growth accelerated from zero percent in 2022 to nearly 40 percent today, while the price-to-operating-cash-flow multiple also moved from the mid-teens to the thirties.

The investment case has become materially more complicated. Alphabet reported its first quarter of negative free cash flow since its IPO, management is guiding for roughly 200 billion dollars in CapEx this year with estimates as high as 300 billion for next year, and the company has raised more than 50 billion dollars in long-term debt while announcing plans to issue approximately 85 billion dollars in new shares, effectively reissuing much of what it bought back. If CapEx reaches 300 billion dollars, Mahncke estimates operating margins could fall from the current 34 to 35 percent to as low as 20 percent by 2030. Alphabet is transforming from an asset-light software business into a capital-intensive infrastructure company and is no longer likely to function as the share cannibal previously anticipated.

Google Cloud growth accelerated to more than 80 percent year over year in the last quarter, and the Cloud backlog stands at half a trillion dollars growing 375 percent year over year, with about 50 percent expected to be recognized within 24 months. Roughly 40 percent of that backlog comes from Anthropic, representing meaningful concentration risk. O'Malley argues the real money in AI will come from serving businesses rather than consumers, that LLMs will be commoditized in the B2C market similarly to how cellular providers became interchangeable, and that Google's enterprise product suite positions it well for B2B AI. Despite this, Mahncke would not buy Google at today's prices given other available opportunities, though he would not sell an existing position.

Amazon was purchased in February at approximately 190 dollars per share and returned roughly 35 percent in five months. Like Alphabet, Amazon's Q2 free cash flow turned negative due to massive CapEx, and Amazon has raised close to 90 billion dollars in long-term debt this year while also guiding for approximately 200 billion dollars in CapEx. After adjusting for a 17 billion dollar markup on its Anthropic stake, Amazon trades closer to 17 or 18 times price-to-operating-cash-flow rather than the apparent 12 or 13 times. If forced to choose between Google and Amazon for new capital, Mahncke would choose Amazon, citing an asymmetry where AI success helps Amazon massively through AWS, chips, and efficiency gains, while failure would allow it to redirect capital back to its high-quality underlying business. Amazon has also built one of the largest chip businesses in the world at a 20 billion dollar annual run rate, and e-commerce still represents only about 20 percent of US retail spending.

Reddit was initially pitched at approximately 85 dollars per share, rose roughly 140 percent within four months prompting a trim, and was later partially repurchased. The position carries a realized gain of about 140 percent and an unrealized gain of about 60 percent on shares bought back. Reddit achieved a 30 percent operating margin by end of last year, years ahead of expectations, but logged-in user growth was only 7 percent year over year, management stopped reporting that metric, and more than half of daily visitors arrive via Google search without being logged in. Google currently pays Reddit only 60 million dollars per year for data licensing, a figure O'Malley described as shockingly low, and click-through rates from Google search to original sources have dropped roughly 60 percent as AI overviews answer queries directly. Reddit accounts for roughly 5 to 10 percent of raw LLM training data and up to 40 percent of real-time search citations in AI-driven answer engines, yet the technology is now mature enough that LLMs would not face a fundamental problem if Reddit stopped licensing its data.

TSMC was originally pitched at approximately 20 times earnings with revenue growing north of 30 percent, a fair value estimate of 320 to 340 dollars per share, and the hosts agreed to add it but failed to follow through. Approximately 15 months later the stock had risen roughly fivefold. Mahncke cautions this performance reflects current AI market enthusiasm more than TSMC's specific long-term outlook, and both speakers identify their mistake as treating the position too binary rather than taking a small two percent allocation to capture upside while surviving worst-case outcomes. O'Malley's primary concern is not geopolitical risk but competitive moat erosion, citing semiconductors as the textbook example of the fastest-changing industry where advantages can be most short-lived.

Mahncke draws a broader lesson that great companies surprise to the upside while mediocre companies surprise to the downside, and a rising price does not make a mediocre company great over the long term.