How Sam Altman fooled Sundar Pichai — and pushed Google into cannibalizing itself

The article argues that Google, OpenAI, Microsoft, Meta, and Amazon are collectively committing over $100 billion a year to AI infrastructure that may not generate enough revenue to cover its costs. It specifically claims OpenAI spends more than $5 billion annually on compute while producing only a fraction of that in revenue, and warns Google’s AI integration could cannibalize its core ad business. The piece frames the AI boom as a speculative overreach with growing downside risk if capital spending continues to outpace monetization.

Analysis

The market is still treating AI spend as an option on future monetization, but the more important second-order effect is margin compression in the incumbents that are funding the buildout. For the hyperscalers, every incremental dollar of capex raises near-term depreciation, power, and network costs before any durable pricing power is proven, which means the earnings risk is not the AI product itself but the dilution to core economics in search, ads, cloud, and operating margin. GOOGL looks most exposed because its AI strategy directly interferes with the highest-ROIC cash engine; the real risk is not that AI fails outright, but that it works just enough to shift user behavior without creating a comparable monetization layer.

The more interesting read-through is that the current spending regime creates a classic “winner’s curse” among the large platforms. The first mover can force competitors to match spend to avoid strategic irrelevance, but that also accelerates industry-wide supply bottlenecks in GPUs, power, data-center construction, and grid interconnects, which lifts costs for everyone and delays payback. If enterprise adoption decelerates even modestly over the next 2-3 quarters, the market will likely re-rate these names on free-cash-flow yield rather than narrative growth, and the most overextended multiple compression would likely hit the names with the least visible path to incremental monetization.

The catalyst path matters: this is not a same-week short because headline AI sentiment can keep overpowering fundamentals for months, but the setup becomes fragile when guidance season forces management teams to justify another year of elevated capex without commensurate revenue acceleration. The downside is convex if one or two product cycles show that users are willing to sample AI but not pay materially more, because then the market has to reconcile rising infrastructure intensity with stagnant ARPU/retention. Conversely, any evidence that AI meaningfully lifts search monetization, cloud deal sizes, or ads conversion would delay the thesis and support the bull case.

Consensus is probably missing that the downside is asymmetrically concentrated in Google, while MSFT, META, and AMZN are more insulated because AI is a feature layered onto broader monetization stacks. The move is likely underdone in relative-value terms rather than outright index beta: the right expression is to short the most strategically cannibalized business model and own the platform with the best ability to absorb capex through diversified enterprise exposure. The key question over the next 6-12 months is not whether AI is transformative, but whether it is accretive to the economics of the companies spending the most to deploy it.

Key Decisions for Investors

• Short GOOGL vs long MSFT in a 3-6 month relative-value pair; thesis is that Google bears the highest risk of self-cannibalization while Microsoft can monetize AI more cleanly through enterprise bundling. Target 10-15% relative underperformance, stop if GOOGL shows clear search monetization uplift or capex moderates meaningfully.
• Buy GOOGL put spreads 3-6 months out on rallies into AI-driven enthusiasm; prefer defined-risk structures because headline support can persist, but the payoff improves if the market starts penalizing capex intensity and ad deceleration. Risk/reward is attractive if implied vol stays below realized volatility.
• Overweight AMZN vs GOOGL on a 6-12 month horizon; AWS can translate infrastructure spend into a clearer enterprise revenue stream, while Google’s AI investment is more likely to pressure its core economics. This is a cleaner way to express the view that not all AI capex is equal.
• For more aggressive positioning, buy a basket short of GOOGL/META/MSFT against a long of software names with lighter capex exposure if market breadth weakens. The trade works best if the sector begins rewarding free-cash-flow durability over narrative growth.
• Avoid chasing the AI leaders after capex guide-ups unless there is evidence of margin stabilization; the risk/reward has shifted from upside optionality to execution risk. Reassess only after 1-2 earnings cycles confirm that incremental spend is translating into measurable revenue efficiency.