Travis Hoy, Lou Whiteman, and Matt Frankel of Motley Fool Hidden Gems Investing delve into the evolving landscape of artificial intelligence, particularly focusing on the existence of competitive "moats" and valuation challenges for AI companies.
Travis initiates the discussion by asking if moats truly exist in AI, noting that historical moats were driven by factors like strong brands in the 1980s or aggregation platforms like Amazon and Netflix in the 2000s. He observes that AI models appear to have fragile moats, with new models like Muse quickly challenging dominant players like ChatGPT.
Lou Whiteman asserts that the discussion of moats in finance is largely "balderdash," arguing that true moats are rare and only recognizable in hindsight. He cites a 2014 book on moats where half the featured companies are now out of business. For AI, he believes moats are non-existent because companies lack pricing power. He questions how a company can claim to be unconquerable if it cannot control its pricing terms with customers, concluding that he sees no permanent competitive advantages.
Matt Frankel is less pessimistic but agrees that AI moats "definitely appear fragile." He points out that ChatGPT's market share dropped from 79% to 54% within a year, while Claude grew from 1% to 9%. Matt argues that raw capability is not a moat, as frontier AI models operate within a narrow performance band. He identifies two potential, albeit still fragile, moats:
1. **Capturing enterprise workflow:** Companies ingrained in existing workflows, like Claude Code, show resilience.
2. **Distribution:** Gemini's usage surged by 450% because of its deep integration into Google's ecosystem (Maps, Gmail, Search). However, Matt questions the durability of even these moats if a superior product emerges, much like Chrome overtook other browsers.
Travis highlights a unique challenge in AI: unlike traditional industries where competitors are often distinct (e.g., a banana company not competing with a cereal company), AI sees tech giants like Meta, Google, and Anthropic all chasing the "same puck," making it difficult to build margins and moats. Lou agrees, noting that these established companies, built on dominance in specific sectors, now face unprecedented cross-industry competition. He concludes that the AI industry is too young to clearly identify durable competitive advantages.
The conversation then shifts to valuing AI companies, prompted by the anticipated Anthropic IPO (S1 filing) and the recent revelation that SpaceX's XAI is primarily an AI company. Matt admits, "I don't know how to value these companies and neither does anybody else." He points to several challenges:
* Anthropic's use of gross versus net revenue in private disclosures.
* Rapid, almost unbelievable, revenue growth rates (Anthropic from a $9 billion run rate to $65 billion by July).
* Lack of public S1 means unknown accounting methods, revenue breakdown (enterprise vs. consumer, with enterprise being stickier), CapEx needs, and current losses.
* No comparable public companies growing at this scale and pace.
Matt states he would scrutinize the S1 for "revenue retention" and "gross margin trends" to understand unit economics and real pricing power.
Lou echoes Matt's caution, suggesting current valuations are likely "less" than what's being hyped, especially given the lack of audited numbers. He argues that current valuations would place these companies among the top 10 globally, a prospect he believes is unlikely for all of them. He cautions against assuming "mega growth for generations to come," drawing a distinction from the dot-com bubble where companies like Pets.com, despite high mindshare, had surprisingly small market caps. He also highlights the potential for companies like Anthropic to report "revenue collected" rather than true net revenue (similar to Uber's gross bookings vs. actual revenue), which could inflate figures. Both agree that patience is warranted, especially given the "hundreds of billions of dollars cash that they do not yet have" for future obligations.
Finally, Travis asks about current investment opportunities in AI. Matt recommends **Modine Manufacturing (MOD)**, an "AI infrastructure stock" specializing in thermal management and cooling for data centers. Despite its "boring" nature, Modine trades at a reasonable 22 times forward earnings, has seen its data center revenue grow 90% year-over-year, and secured a $4 billion multi-year deal from a hyperscaler. He notes its diversified business reduces risk if AI build-out slows.
Lou, however, expresses less interest in "picks and shovels" at this stage, arguing that some primary beneficiaries (the companies using the picks and shovels) are now trading at better valuations. He suggests investing in established tech giants like **NVIDIA, Alphabet, and Microsoft** due to their "multiple ways to win" beyond just AI. He contends these companies have strong existing businesses even if their AI ventures don't fully materialize. His "even easier way" to gain AI exposure is simply to buy the **S&P 500**, given its significant weighting toward these AI hyperscalers.
The hosts conclude by anticipating the Anthropic S1 release for more concrete data on financials and competitive positioning.