Anthropic Just Announced Huge News for Alphabet and Broadcom

Anthropic said it will use multiple gigawatts of next-generation TPU computing power starting in 2027, reinforcing demand for TPUs from Broadcom and Alphabet. Broadcom said its custom AI chip business could exceed $100 billion annually by the end of 2027, while its AI semiconductor division posted $8.4 billion in revenue last quarter, up 106% year over year. Alphabet’s Google Cloud TPU revenue rose 48% year over year last quarter, signaling continued AI-driven growth for both companies.

Analysis

The key second-order signal is not that TPUs are gaining share, but that hyperscale AI demand is broadening from a single-vendor architecture into a multi-chip sourcing model. That reduces Nvidia’s pricing power at the margin and, more importantly, makes custom silicon a strategic procurement category rather than a niche optimization. Over the next 12-24 months, that should continue to shift incremental AI capex toward vendors with design-win durability and high attach rates, which is structurally favorable for Broadcom’s mix and Alphabet’s cloud margin leverage.

Broadcom is the cleaner operating leverage story because each additional long-duration customer commitment improves manufacturing visibility and supports more aggressive capacity planning. The market may still underappreciate how much of the upside comes from revenue quality, not just revenue growth: once these programs reach scale, they tend to become sticky because migration costs, software integration, and performance tuning create switching friction. That said, the valuation risk is that investors extrapolate a straight-line ramp into 2027, leaving the stock vulnerable if order timing slips even modestly.

Alphabet’s TPU monetization is more subtle: the economics likely show up less as a discrete hardware P&L and more as cloud retention, higher workload density, and improved unit economics versus renting third-party accelerators. The bullish setup is that TPU adoption can expand Cloud margins faster than consensus without requiring a step-change in overall enterprise AI spending. The contrarian risk is that the market may already be discounting TPU success while underestimating capex intensity and the possibility that customers use TPUs as leverage in pricing negotiations rather than as a full migration path.

Nvidia is not being displaced, but the narrative is moving from monopoly-like scarcity to competitive abundance, which usually compresses multiple before it compresses earnings. If this multi-sourcing trend persists, the biggest beneficiaries may be adjacent ecosystem names that enable custom chip deployment, networking, and cloud orchestration rather than pure accelerator suppliers. The next catalyst to watch is whether other frontier-model customers publicly follow Anthropic’s long-dated commitment pattern over the next 1-2 quarters.

Key Decisions for Investors

• Long AVGO on a 6-12 month horizon; use pullbacks to add. Risk/reward favors exposure to recurring custom-silicon wins and operating leverage, with downside mainly from order-timing volatility rather than thesis breakage.
• Add to GOOGL for a 12-18 month horizon as a lower-beta AI beneficiary. The setup is attractive if Cloud margin expansion outruns consensus; risk is that capex remains elevated longer than expected.
• Pair trade: long AVGO / short NVDA as a relative-value hedge over the next 3-6 months. This captures the market shift toward custom silicon while reducing outright AI beta; close if Nvidia regains share through pricing or new product cycles.
• Sell out-of-the-money NVDA calls against existing long exposure for 1-2 quarters. The thesis is not bearish on earnings, but multiple expansion may be capped as customers diversify accelerator sourcing.
• Avoid chasing the names after a run-up; wait for post-earnings or market-wide AI de-ratings to build positions, since the main risk is sentiment overreach before the 2027 revenue inflection is visible.