2 Stocks to Buy as Google Starts Selling Its AI Chips

Broadcom's AI semiconductor revenue rose 221% year over year to $16.7 billion in fiscal Q3 and is guided to reach roughly $21.7 billion in Q4, while total revenue climbed 86% to $29.6 billion and free cash flow nearly doubled to $13.7 billion. Alphabet began recognizing TPU system-sales revenue in Q2, with most contracted revenue expected in 2027; Google Cloud revenue increased 82% to $24.8 billion and operating income rose to $8.8 billion from $2.8 billion. Alphabet's cloud backlog grew more than $50 billion sequentially to $514 billion, although its $195 billion-$205 billion 2026 capex plan and hardware-related costs pose margin and spending risks.

Analysis

The investable implication is not merely incremental TPU revenue, but validation that hyperscaler-designed silicon is moving from captive infrastructure into an external hardware channel. That expands AVGO's addressable content per AI cluster beyond accelerator ASICs into switching, interconnect and rack-level components, while creating a credible alternative procurement path to NVDA for a narrow set of scale customers. The near-term read-through is most negative for merchant AI accelerator pricing and NVDA's long-duration scarcity premium, not necessarily its absolute revenue base: external TPU deployments give large model developers leverage in negotiations even if they retain heterogeneous compute fleets.

For GOOG, the strategic value is customer lock-in and better utilization of a capex base that would otherwise be justified principally by internal demand and Cloud rentals. The economic risk is that system sales shift mix toward lower-margin hardware, increase working-capital intensity, and potentially cannibalize higher-margin Cloud consumption; the bullish case requires attached software, support, and recurring Cloud/network services to offset this. Over the next 1-3 months, investor focus should be on backlog conversion terms, customer prepayments, and whether capex guidance rises again; over 6-18 months, the key metric is Cloud operating-margin durability as external deployments scale.

Consensus may be over-extrapolating a single customer deployment schedule into a broad third-party TPU market. Custom silicon remains workload-specific, requires deep software integration, and may be less portable than CUDA-based fleets; a delayed deployment or a customer's decision to diversify back to NVDA would impair AVGO's AI-growth expectations disproportionately because the market is assigning a premium to sustained accelerator growth. The more underappreciated positive is that AVGO's networking exposure can monetize AI cluster scale even if accelerator share changes, making it structurally less binary than a pure TPU thesis.

Key Decisions for Investors

• Initiate a 3-6 month long AVGO / short NVDA pair at roughly equal AI-revenue sensitivity, not equal dollar beta. Thesis: external custom-silicon adoption compresses NVDA's marginal pricing power while AVGO captures both ASIC and networking content. Target 10-15% relative upside; stop if AVGO AI revenue guidance falls below the next-quarter run-rate implied by management or NVDA demonstrates renewed supply-constrained pricing.
• Accumulate GOOG on post-earnings or capex-driven weakness rather than chasing the TPU headline; use a 6-18 month horizon. Underwrite only if Cloud operating margin remains above the mid-30% range while backlog converts, as that would show hardware sales are attaching services rather than diluting the segment. Reassess on another material capex increase without corresponding Cloud revenue or backlog acceleration.
• Use AVGO put spreads or reduce gross exposure into the next earnings release if the position has appreciated materially: customer concentration makes a single deployment timing revision capable of producing a sharp multiple reset despite strong reported revenue. A 10-15% downside in the stock is plausible if the forward AI ramp is deferred by one or two quarters.
• Watch for disclosed customer prepayments, system gross-margin commentary, and third-party TPU wins beyond the initially identified buyer. Absence of additional contracts by the next two reporting cycles would argue that the market is niche rather than platform-like, favoring profit-taking in AVGO and a more neutral GOOG stance.