Warren Buffett's Successor, Greg Abel, Buys an AI Stock That Poses a Threat to Nvidia

Alphabet’s TPUs are emerging as a meaningful AI accelerator alternative to Nvidia, with customers including Anthropic and Meta and a new plan to deploy TPUs directly in select customers’ data centers to expand its addressable market. First-quarter revenue rose 22% to $109.8 billion and operating income climbed 39% to $39.6 billion, while Wall Street’s median target price of $430 implies 11% upside from the current $386 share price. The article is constructive on Alphabet’s AI and cloud positioning, but it also notes Nvidia’s continuing dominance.

Analysis

This is less about TPUs becoming a GPU replacement and more about Alphabet monetizing an underappreciated control point in the AI stack: procurement flexibility. Letting select customers deploy TPUs in their own data centers removes a major adoption blocker for regulated, latency-sensitive, or capacity-constrained buyers, which broadens the TAM without requiring Google Cloud to win the full workload. The second-order effect is that Alphabet can attack Nvidia at the margin while preserving pricing power via tighter integration across cloud, model hosting, and hardware support.

The market’s likely underestimating how much this changes bargaining power. If large AI buyers can dual-source compute between Nvidia and Alphabet, Nvidia’s mix becomes more exposed to price competition in specific inference and training pockets, even if CUDA remains the default for frontier research. That said, TPU adoption is still a constrained wedge: the software moat and general-purpose flexibility keep Nvidia dominant, so the competitive damage is more about slower share gains and potential ASP pressure than outright displacement.

For Alphabet, the setup is attractive because hardware adoption can pull through higher-margin cloud revenue and deepen customer lock-in, but execution risk is real. The key watch item over the next 6-12 months is whether this becomes a repeatable enterprise product or stays limited to a handful of marquee accounts; broad rollout would be materially more bullish than isolated wins. For Meta and Blackstone, this expands optionality around compute sourcing and may lower capex per unit of deployed AI capacity, but it also increases dependence on Alphabet’s supply chain and service quality.

Consensus seems to be treating this as incremental. The more interesting read is that Alphabet is positioning TPUs as a strategic wedge in inference economics and enterprise deployment, not as a pure share-grab against Nvidia. If that thesis works, the re-rating for GOOGL likely comes from sustained cloud margin expansion rather than headline TPU revenue alone.

Key Decisions for Investors

• Long GOOGL vs short NVDA as a 6-12 month relative-value trade; thesis is not Nvidia collapse but incremental share pressure and pricing discipline risk as TPU deployment expands outside Google Cloud.
• Add to GOOGL on pullbacks into earnings or AI-capex selloffs; risk/reward improves if cloud margins continue to expand while TPU adoption broadens across regulated enterprise buyers.
• Buy NVDA downside protection 3-6 months out rather than outright shorting; use put spreads to express risk that TPU headlines compress sentiment without fighting the structural CUDA moat.
• Watch META and BX as secondary beneficiaries of cheaper/alternative compute sourcing; a tactical long in META vs broad tech can work if TPU access improves inference economics faster than expected.
• If TPU customer announcements accelerate over the next 1-2 quarters, rotate part of the AI basket from pure hardware beta into GOOGL and AVGO, which have better balance between AI exposure and earnings quality.