Buffett's Heavy Bet on Japan's Five Major Trading Houses: Moats Face AI-Era Challenges
Taylor Wilson
Berkshire Hathaway holds roughly 10% of each of Japan's five major trading houses; their combined market cap has quintupled in seven years. New CEO Greg Abel says the stakes are meant to last "fifty years or forever" — but AI is now threatening the very information edge that makes the model work.
How big is Berkshire's Japan bet?
Berkshire owns about 10% each of Mitsubishi, Itochu, Mitsui, Sumitomo, and Marubeni. Since Buffett began building the position roughly seven years ago, the five firms' combined market cap has grown fivefold.
New CEO Greg Abel says Berkshire plans to hold these stakes "fifty years or forever" and will visit all five companies in Japan next month — a first.
This means → This is not a trade. It is one of Berkshire's rarest moves — an ultra-long-term bet on a single country's business model.
What do trading houses actually do — and why did Buffett pick them?
The five firms deploy roughly 5,000 staff worldwide, spanning supply chains from instant noodles to defense. Their core skill is spotting and filtering profit opportunities across global supply networks.
In plain terms = trading houses don't make things. Their job is knowing where money can be made, then connecting the resources and relationships to capture it.
Markets long discounted the model as too scattered — the so-called "conglomerate discount." Around 2020, the houses responded by actively culling losing businesses. Mitsubishi sold part of its Australian coking-coal portfolio in 2024, even though the unit had been a major profit source just two years earlier.
This reflects a deeper shift: as UBS analyst Harunobu Goro put it, "the trading houses have become portfolio managers."
How could AI undermine the moat?
The houses' deepest competitive advantage is a global information network built over decades — they often know first when a supply chain breaks or a market opens up.
This means → That is precisely what AI does best: process information at scale, find patterns, and render judgment. If AI can replace the "information middleman" role, the moat gets thinner.
Mitsui CEO Kenichi Hori was blunt: "If we lack independent judgment, AI will destroy our business."
How are the houses fighting back?
Sumitomo has already embedded AI in its investment-decision process. Its platform was trained on internal meeting records from 1,500 past investments and 4,000 post-loss reports to analyze why deals went wrong.
In plain terms = Sumitomo is feeding decades of expensive lessons into AI, so the machine can flag whether the next deal risks the same mistakes.
CEO Shingo Ueno's goal is explicit: "We want to be a trading house that profits from AI."
What is the make-or-break variable for Berkshire's bet?
The core question is singular: can the houses convert their unstructured, on-the-ground, first-hand knowledge into a barrier AI cannot replicate?
If yes, the trading-house model becomes more valuable in an era of frequent supply-chain disruption. If no, AI will steadily erode their role as information middlemen.
This means → Berkshire's "hold for fifty years" pledge is ultimately a wager that human judgment plus AI beats pure AI.
Content is for reference only, not financial advice.