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

Alphabet’s TPUs are gaining traction as an AI accelerator alternative to Nvidia GPUs, with customers including Anthropic, Meta, and a Blackstone-linked cloud venture. First-quarter revenue rose 22% to $109.8 billion and operating income jumped 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 argues TPUs could expand Alphabet’s addressable market, though Nvidia remains the dominant AI accelerator provider.

Analysis

Alphabet’s TPU push is less about displacing Nvidia overnight and more about changing the buying criteria for a subset of AI workloads. Once a hyperscaler-grade alternative can be deployed inside customer-owned data centers, the procurement conversation shifts from “best training stack” to “good-enough performance at lower total cost plus supply assurance,” which can pull budget away from Nvidia in inference-heavy or vertically integrated deployments. The second-order winner is Alphabet’s cloud attach rate: TPUs can become the wedge that converts model builders into broader Google Cloud customers, lifting utilization on adjacent networking, storage, and managed services.

The market is likely underestimating the distribution advantage here. Nvidia’s moat is software, but the TPU story only needs to win a narrow slice of workloads to matter because AI capex is still expanding faster than any single vendor can absorb. The real risk to Nvidia is not a sudden share collapse; it is a slower mix shift where large buyers dual-source more aggressively, compressing pricing power and elongating refresh cycles over the next 12-24 months.

For Alphabet, the important catalyst is not the first TPU deal but whether on-prem TPU deployments become repeatable with large enterprises and sovereign/regulated customers that prefer control over data locality. That would expand TAM without requiring Google to build all the capacity itself, improving capital efficiency. The contrarian view is that TPU economics may be good enough to matter even if software parity never arrives, because many buyers will accept some developer friction in exchange for supply, cost, and deployment flexibility.

Berkshire’s incremental ownership matters because it validates the “pick-and-shovel plus platform” thesis: Alphabet can monetarily benefit from AI demand whether or not it wins the accelerator crown. However, the setup is not a clean long-NVDA/short-GOOGL pair; if TPU adoption broadens, the most vulnerable names are pure GPU-dependent inference vendors and any AI infrastructure supplier that lacks proprietary silicon or software lock-in.

Key Decisions for Investors

• Add to GOOGL on pullbacks over the next 2-6 weeks; thesis is a 12-18 month re-rating as TPU TAM expands beyond Google Cloud, with upside from multiple expansion if cloud margins continue to inflect.
• Reduce/hedge NVDA exposure into strength with a 3-6 month call-spread or partial stock hedge; risk is not headline share loss but margin compression from dual-sourcing and customer bargaining power.
• Pair trade: long GOOGL / short a basket of AI infrastructure pure-plays with no proprietary silicon or software moat over 3-9 months; the trade works if capex keeps growing but procurement shifts toward alternative stacks.
• Consider BRK.B as an indirect beneficiary of the AI hardware arms race; Berkshire’s position is a signal that large capital allocators see Alphabet as a durable AI monetization platform, not just an ad company.
• For event-driven traders, watch for the first announced on-prem TPU deployment with a marquee enterprise or sovereign customer; that is the next catalyst that could force sell-side estimates higher and justify a 5-10% re-rating in GOOGL.