Warren Buffett's successor just told the world's richest shareholders that AI at Berkshire means optimizing insurance claims, not building the next ChatGPT.
The Summary
- Greg Abel outlined Berkshire's "narrow AI" strategy at his first annual meeting as CEO: deploy AI for targeted internal efficiencies, not moonshots.
- Shareholders gave Abel positive marks for his pragmatic approach and directness, contrasting sharply with the hype-driven AI narratives dominating other corporate boardrooms.
- The real play: Berkshire's energy unit is chasing data center demand, building the infrastructure for other companies' AI ambitions while avoiding the technology risk itself.
The Signal
Greg Abel's first Berkshire Hathaway annual meeting post-Buffett wasn't about reinventing the conglomerate. It was about clarifying what AI actually means when you run $1 trillion in assets across railroads, insurance, and utilities. Abel told shareholders Berkshire will pursue "narrow AI" applications focused on specific operational improvements. Think automating insurance underwriting decisions, optimizing freight logistics, improving energy grid management. Not building frontier models or chasing AGI.
This is the anti-SoftBank approach. While venture capital pours billions into speculative AI startups and public companies rebrand themselves around large language models, Berkshire is doing what it always does: finding the boring, profitable angle everyone else ignores. Shareholders responded positively, signaling that at least one corner of the market still values execution over narrative.
"AI will be deployed narrowly" isn't a hedge. It's a strategy.
The more interesting move is what Berkshire Energy is doing. The utility subsidiary is actively pursuing data center contracts, positioning itself as infrastructure for the AI boom without taking technology risk. Data centers need massive, reliable power. Berkshire owns utilities across multiple states. The math is simple: let others burn cash training models, then charge them for the electricity to do it.
This is classic Berkshire positioning. During the dot-com bubble, they didn't buy pets.com. They bought brick manufacturers and carpet companies that would supply the warehouses Amazon eventually built. Now they're setting up to be the landlord and power company for AI's build-out phase.
Key strategic contrasts:
- Tech companies: Building AI models, taking technology and market risk
- Berkshire: Selling power and infrastructure to AI builders, taking utility-style margin
- Result: Berkshire captures AI growth without AI volatility
Abel's reception from investors suggests the message landed. He's not Buffett, and he's not trying to be. He's presenting a clear-eyed view of where AI creates actual value for Berkshire's specific portfolio, and where it's just expensive distraction. For a company built on insurance float and capital allocation discipline, that means operational AI in existing businesses plus infrastructure plays on energy.
The Implication
Watch Berkshire Energy's data center contracts over the next two quarters. If they're signing long-term power agreements with hyperscalers or AI-focused data center operators, that's the signal that Berkshire's narrow AI strategy is actually a wide infrastructure bet. The companies building AI need massive compute. Massive compute needs massive power. Berkshire owns the power.
For everyone else building in this space: there's a lesson in selling picks and shovels instead of panning for gold. The AI infrastructure layer, power and cooling and connectivity, might be the highest-probability bet in the entire stack.