Blog: Why Berkshire Hathaway Could Be The Ultimate AI
This weekend I was reading Warren Buffett’s comments on Tesla’s plan to sell insurance. It got me thinking about whether or not AI will lower the barriers for some kinds of companies to enter new markets. Sure, better data might mean better underwriting, but insurance is a balance sheet business. Tesla’s finances are already weak, I don’t see how they survive a year of really bad luck in the insurance business. Competitive moats come in many forms. Will AI tear some of them down?
While trying to work through my thoughts on the issue, I realized that Berkshire Hathaway could be the world’s top applied AI company, should they choose to be. That might seem surprising since Berkshire is not a tech company, but hear me out. Here are four reasons I think Berkshire could be strong in AI.
1. The conglomerate structure is rare in business these days, but holds advantages for a corporate level AI team. There must be all kinds of interesting data sets across Berkshire’s businesses, some that could benefit from the fact that the others exist. From a recruiting perspective, the financial strength of Berkshire plus the diversity of AI challenges within all the businesses could be really interesting for AI engineers. It would be easy to move around and work on different things. And you could imagine an interesting internal-only conference for all the AI happening at Berkshire companies.
2. The financial resources of Berkshire mean they could invest very heavily in compute and storage for automation and learning, as a capital expenditure rather than an operational one. They could buy entire companies (NVIDIA?) if they were deemed to be strategic, or when taking them off the market put competitors at a disadvantage.
3. Berkshire’s culture of NOT being a tech company is actually an asset for applied AI. Berkshire wouldn’t need research scientists focusing on the next neural net topology and publishing papers, they just need strong data scientists working on real applied problems. Most companies are not adopting AI in the right places. It is being adopted where there is a forward thinking exec bringing into her business unit, or where it is highly obvious (things like logistics). But no one is financial statement focused. Companies aren’t looking at the venn diagram of where data set availability, income statement improvement, and stable AI technology intersect. Berkshire could do that. Looking through a traditional capital allocation lens, combined with a competitive moat lens, at AI deployment could have awesome impacts.
4. The pure scale of Berkshire would mean AI methods that have a marginal impact on a percentage basis could have significant impacts in nominal dollars.
I doubt this will happen. We already know some Berkshire businesses are deploying AI on their own, and so, it will probably be a unit by unit thing, outside of corporate control. But if Berkshire were to consolidate an AI group at the headquarters level, it could have a tremendous impact.