Google is hoarding TPUs to chase artificial general intelligence

Alphabet said it is prioritizing AI/AGI frontier compute—rather than renting out all TPUs—while confirming TPU sales to some customers. Despite very strong fundamentals (Google Cloud revenue +82% to $24.75B and profit +214% to $8.8B; $514B cloud backlog; Search +17%, YouTube ads +13%), free cash flow fell to -$5.9B (first negative since 2004) and shares dropped ~4% after hours. Management also raised planned AI infrastructure spending to $195B–$205B for the year (from $180B–$190B) and flagged a supply-constrained environment, planning to expand third-party “bridging” capacity in Q3.

Analysis

The near-term market issue is not demand for Alphabet’s products; it is the cash conversion profile of an AI strategy that is increasingly capex-intensive. That creates a clean mechanical headwind for the multiple: if free cash flow is negative while spend keeps stepping up, buyback support weakens and investors will pay less for every dollar of forward earnings, even if operating income is still expanding.

The second-order winner is the AI infrastructure supply chain. Accelerator vendors and adjacent bottlenecks are still the scarce assets in this cycle, so names tied to GPUs, networking, and power should keep seeing durable demand. On the competitive side, Google’s need to rent third-party capacity is a subtle bullish read-through for hyperscale peers with spare compute and for enterprises that can arbitrage multiple clouds; it also suggests GOOGL is more capacity-constrained than demand-constrained.

Over the next 1-3 months, the key catalyst is whether the market believes the capex step-up is a one-time bridge to monetize the $514bn backlog or the start of an extended FCF drought. The contrarian view is that consensus is overreacting to accounting optics: a negative FCF quarter from a fortress balance sheet is not a solvency signal, and AI Mode/search engagement gains imply the product risk is lower than the bear case. The thesis breaks if capex stays above ~$200bn without a corresponding cloud margin bridge in Q3 or if search monetization starts lagging query growth.

Key Decisions for Investors

• Tactically short GOOGL or buy a 1-3 month put spread on any post-earnings bounce; target a 5-8% downside move if the market keeps focusing on negative FCF and higher capex. Invalidate the trade if management reiterates capex without another downside surprise in cloud margins and the stock reclaims the pre-earnings gap.
• Keep a longer-dated bullish entry on the watchlist: buy GOOGL common or a 6-12 month call spread only after an additional 8-10% drawdown, when the cash-flow reset is better priced. Reward is a re-rating if backlog converts and AI/search monetization remains additive; risk is a prolonged capex treadmill.
• Overweight AI infrastructure beneficiaries on a 1-3 month horizon via NVDA, AVGO, or SMH against GOOGL if you want to express the capex cycle rather than the company-specific cash-flow optics. Falsifier: if Google internalizes more of its TPU stack and reduces third-party silicon mix faster than expected.
• Set an alert for the next quarterly capex guide and cloud margin disclosure; if spending stays elevated but operating cash flow does not re-accelerate, trim any long GOOGL exposure and rotate to the suppliers. If capex growth decelerates while backlog continues to expand, the setup shifts back to long.