AI investors are in for a rude awakening | Roger McNamee

The tech industry has invested an unprecedented $717 billion into LLM AI and infrastructure over the past three years, with comparable amounts expected, as major players each pursue a global monopoly to replace aging core products. Despite this colossal investment and forecasts for future revenue growth, current end-user LLM product sales are a tiny fraction of the capital deployed, creating an unsustainable gap where infrastructure providers like cloud services and Nvidia are the primary beneficiaries. This intense competition, coupled with reported AI failures for customers, suggests that many participants will inevitably fail, leading to substantial write-offs and a significant reckoning for investors as returns on this gargantuan capital allocation prove inadequate for most.

Analysis

The Large Language Model (LLM) AI sector is experiencing a capital investment boom of unprecedented scale, with an estimated $717 billion deployed over the last three years and comparable amounts anticipated. This expenditure is driven by a high-stakes competitive battle among five to six major tech players—including Google, Amazon, Meta, and Microsoft/OpenAI—each seeking a new global monopoly to replace their aging, and arguably deteriorating, core products. A profound and unsustainable imbalance exists between this massive capital outlay and realized end-user revenue, which for market leaders totaled only an estimated $4 billion in 2024. The majority of current AI-related revenue is circular; capital invested by firms like Microsoft, Google, and Amazon flows back to their own cloud services divisions and to critical infrastructure suppliers such as Nvidia. While this dynamic has fueled explosive revenue growth for infrastructure providers, it masks the fundamental risk that the underlying LLM technology has yet to prove its commercial value, with numerous reported failures across various industries. The market's current assumption that all major players will succeed is flawed. It is more likely that only one or two will emerge as profitable monopolies, forcing the others to write off hundreds of billions in investments, which could trigger a catastrophic reckoning for equity investors.

Key Decisions for Investors

• Investors should critically re-evaluate exposure to the primary LLM developers (MSFT, GOOGL, AMZN, META), as the current 'winner-take-all' dynamic implies a high probability of significant investment write-offs for the eventual losers.
• It is crucial to differentiate between the high-risk LLM application developers and the key infrastructure enablers like Nvidia, which are the primary near-term beneficiaries of the sector's massive capital expenditure cycle.
• Monitor the ratio of capital investment to end-user software revenue for AI-centric companies; a failure of this gap to narrow meaningfully is a leading indicator of an unsustainable market bubble.
• Consider the existential risk to big tech firms whose legacy products are maturing; a failure to establish a new monopoly in AI could severely impair their long-term market value, beyond the immediate impact of capital write-downs.