The five Japanese trading houses — Berkshire's basket, six years on

As of 7 September 2026. Prices are Tokyo close, 3:30 PM JST. Every figure below is sourced; where I could not source something I say so rather than estimate it.

Why this memo is not an Oracle teaser

Oracle reads SEC XBRL — companyfacts, 10-K filings, Form 4 insider filings. Itochu, Mitsubishi, Mitsui, Sumitomo and Marubeni are not SEC registrants. They file with Japan's FSA through EDINET and disclose to the Tokyo Stock Exchange in the tanshin format. Oracle's pipeline has nothing to read for them, so none of its lenses, scores or provenance stamps apply here. This is hand-researched from company disclosure and public market data instead.

Two consequences worth keeping in mind:

• There is no Form 4 equivalent, so nothing like Oracle's INSIDER lens can be computed. Japanese disclosure of insider dealing works differently and is not a comparable dataset.
• Fiscal years end in March, and Japanese companies name them by the opening year. Itochu's own IR calls the year ended March 2026 "FY2025". To avoid that trap this memo always writes the year-end date in full.

The one number that answers the yen question

The yen's fall is not a footnote to this investment — for a dollar-based investor it is most of the story.

So an unhedged dollar investor loses 32.0% of any yen return on conversion. That is a multiplier, not a subtraction, and it bites hardest on the biggest winners:

Berkshire never took that hit. It funded the purchases by issuing yen-denominated bonds, so yen liabilities sit against yen assets and the currency largely cancels out. The carry is the point: in 2025 the basket paid Berkshire about $812 million in dividends against roughly $135 million of interest on the yen debt — yen borrowing costing under 1% a year against a basket yielding around 4%. Greg Abel said on 2 September 2026 that Berkshire still intends to raise yen debt "as appropriate" despite Japanese yields at 30-year highs.

This is why DXJ is the right benchmark and an ADR is the wrong one. Berkshire is running a currency-hedged position, achieved with debt rather than forwards. DXJ hedges by selling yen forwards monthly. They are the same trick.

Against the USD-hedged WisdomTree index

DXJ — WisdomTree Japan Hedged Equity Fund. $179.97 (4 Sep 2026), $7.28bn in assets, 0.48% expense ratio, 431 holdings, dividend-weighted with an exporter tilt — it only admits Japanese companies earning less than 80% of revenue inside Japan.

The hedge, measured against its own unhedged peer group:

Hedging the yen was worth roughly 150 percentage points over five years. Not stock selection — currency. DXJ's trailing one-year total return to 4 September 2026 was 45.1%, and its 52-week range is $125.92–$183.11, so it is close to its high.

The catch you should know about this comparison. DXJ holds the trading houses. Mitsubishi Corporation is a top-ten position at about 2.73%, and because the index is dividend-weighted with an exporter screen, the others qualify too — these are among Japan's largest dividend payers with the most overseas revenue. Benchmarking the five against DXJ is therefore partly benchmarking them against themselves. It is still the fairest available yardstick for the hedged experience, but it is not an independent one.

The five businesses

A sogo shosha is not a conglomerate in the Western sense and not really a trader any more. Each is a portfolio of operating stakes — mines, LNG trains, power plants, aircraft leases, supermarkets, salmon farms — plus the logistics and financing that moves the output. Profit comes from equity-method income in hundreds of subsidiaries, so the right way to read one is as a listed private equity fund with a trading arm attached, not as a commodity proxy.

The single most useful split between them is resources versus everything else. Mitsubishi and Mitsui earn heavily from energy and metals, which makes them cyclical and commodity-priced. Itochu is the least resource-dependent, with a consumer and domestic tilt — textiles, food, FamilyMart convenience stores. That difference explains almost every divergence in the tables below.

Year ended March 2026 (actual) → year ending March 2027 (company guidance)

Combined guidance is about ¥4.18 trillion of net profit for the year ending March 2027 — every one of the five guiding up, after a year in which three of them went backwards. Mitsubishi's +37% is the swing: it takes the profit crown back from Itochu, driven by US shale gas, LNG Canada and its strategic investments, having fallen 15.8% the year before.

Itochu's figures are on a 5-for-1 split effective 1 January 2026, and Sumitomo's on a 4-for-1 split effective 1 July 2026 (1.195bn shares became 4.780bn). Per-share comparisons to older data will be wrong unless adjusted.

What each is worth today (7 September 2026)

Read that last column. The resource-heavy names — Mitsui, Marubeni, Mitsubishi — are 17–24% below highs set within the last year, while consumer-tilted Itochu and Sumitomo sit within 4% of theirs. WisdomTree attributes the July 2026 break to Japan's momentum trade unwinding on narrow, AI-driven concentration. On forward guidance Mitsui at 15.6× and Marubeni at 13.5× are the cheapest of the five while guiding profit up 10% and 7%.

Note the headline yields understate what shareholders receive, because all five run large buybacks alongside dividends. Currently authorised: Itochu up to ¥300bn (to 29 Jan 2027), Mitsui ¥200bn (to 29 Jan 2027), Sumitomo ¥80bn (to 31 Mar 2027), Marubeni up to ¥60bn. On a total-yield basis a 2025 survey put the five between 4.3% and 7.8%.

Berkshire's position

Berkshire crossed 10% in Sumitomo (9.30% → 10.05%, 119.8m shares) and Marubeni (9.32% → 10.10%, 165.3m shares) on 7 May 2026, having already passed it in the other three. The original commitment capped Berkshire below 10%; each company individually granted permission to go beyond, which Buffett flagged in the February 2025 letter. This was Greg Abel's first material capital deployment as CEO, and he has said Berkshire intends to hold for decades.

Dividend growth since August 2020 differs sharply by name — Marubeni +120%, Mitsubishi +35%, about +70% collectively, against +12.8% for the S&P 500 over the same span.

Forecasts: what actually exists

There is no analyst consensus feed behind this memo, so what follows is company guidance — management's own published numbers — plus named analyst targets where a source gave one. That is a real forecast with an author, not a blended estimate.

Every one of the five built its current guidance on ¥150 per dollar. Spot is 155.88, so the yen is already about 4% weaker than assumed, which is a tailwind to the guidance above. Confirmed at ¥150 for Itochu, Mitsubishi and Mitsui; I could not source Sumitomo's or Marubeni's stated assumption, though Marubeni describes yen appreciation as a headwind it expects to offset.

Disclosed sensitivities:

• Mitsubishi: each ¥1 of yen weakness adds ¥5bn to net profit. Each $1 on Brent adds ¥2.4bn. Its plan assumes Brent $78 and Henry Hub $3.90/MMBtu, and carries a ¥30bn contingency for geopolitical disruption including a possible closure of the Strait of Hormuz.
• Itochu: a ¥5 revision was worth about ¥15bn last year, so roughly ¥3bn per yen.
• Mitsui: assumes ¥150–151, oil $78–80, US gas $3.50/MMBtu.

First-quarter results (April–June 2026, reported early August) beat consensus at all five, with progress rates around 30% against full-year guidance: Mitsubishi net profit +47% (¥298.5bn, 27% of the year), Mitsui +53%, Sumitomo +11% reported and +29% underlying (¥190.1bn), Itochu +4% (¥293.7bn) with core profit +38%. None raised full-year guidance, but Mitsubishi, Sumitomo and Marubeni have all signalled possible upward revisions at the half year — a notable clustering.

Named analyst targets, for what they are worth:

Longer-horizon company targets: Mitsui is guiding to ¥1.1tn profit and 12% ROE by the year ending March 2029, with a stated vision above ¥1.4tn and 13% ROE by 2030. Mitsubishi targets 12%+ ROE by March 2027 against 8.5% delivered. Itochu, which has abandoned multi-year plans in favour of a single-year plan each April, guides to about 15% ROE — it earned 14.6% and is the only one of the five Morningstar awards an economic moat.

The tension worth understanding

The thing that made this investment work for a dollar investor is now the thing most at risk.

Japan's 10-year government bond yield touched 3.0% in early September 2026, its first time since 1996, having more than tripled in two years. The Bank of Japan meets 17–18 September with markets pricing an 80–90% chance of a hike to 1.25% — 0.75 points in nine months. In late July the yen hit 163.73, its weakest in about four decades, and Japan and the US intervened jointly for the first time since 2011, spending an estimated $85bn in two days to push it back to around 159. It is 155.88 now.

That sets up a genuine conflict rather than a simple trade:

• A stronger yen helps an unhedged dollar holder of the Tokyo shares, and hurts DXJ, which gives up the depreciation tailwind that produced most of its 150-point advantage.
• A stronger yen hurts the companies' reported profits, because every one of them is guiding on ¥150 and earns heavily abroad. Itochu explicitly names yen appreciation amid rising domestic rates as a downside risk to both equity-method income and overseas revenue.
• Berkshire is indifferent to the level and exposed to the rate. Its hedge is yen debt, so currency moves wash out — but higher Japanese yields raise the cost of rolling that debt, which is precisely what made the sub-1% carry so attractive. Abel's 2 September comment that high yields are "not a challenge for the trading houses right now" was about the companies, not about Berkshire's own funding, and he said Berkshire will keep borrowing in yen anyway.

Forecasters do not agree on direction: Goldman Sachs sees USD/JPY falling to 140–145, Crédit Agricole sees a rebound to 163 by December, and bank consensus sits near 158 for December 2026. That spread — 140 to 163 — is about 15% of the exchange rate, which on the arithmetic at the top of this memo matters more to a dollar return than any of the company guidance does.

What I could not source

Stated plainly, because these gaps change what conclusions are available:

• Per-company total returns since August 2020. I could not find split-adjusted price history for the individual names, and both Itochu and Sumitomo have split since. The only sourced figure is a range of +227% to +551% across the five in yen, and that is as of October 2025, not today. Applying today's exchange rate to a ten-month-old price range would be a fabricated number, so the FX table above is deliberately presented as arithmetic on a hypothetical yen return rather than as measured per-name dollar performance.
• Sumitomo's and Marubeni's stated USD/JPY assumptions.
• Sumitomo's prior-year dividend per share, where sources disagree (¥140 versus ¥150 pre-split). The forward figure of ¥40 post-split is solid and consistent with the observed 2.20% yield.
• Price-to-book for any of the five, which for asset-heavy holding companies is arguably the more informative multiple than P/E.
• Berkshire's exact current percentage in each of the five beyond "above 10%", and the exact stake in Mitsubishi and Mitsui.
• Live market data. IB Gateway resolved every contract — 8001.T, 8058.T, ITOCY, MITSY, SSUMY, MARUY, DXJ, USD.JPY — but returned Error 10197, "no market data during competing live session", so nothing here comes from our own feed. Prices are from public quote pages, timestamped above.

Sources

• Company IR: ITOCHU results outlook · Mitsubishi FY results and forecast · Mitsubishi Q1 presentation · ITOCHU Q1 summary · Sumitomo stock data
• Berkshire: Nikkei Asia on crossing 10% · CNBC, positions top $30bn · CNBC, Abel on Japanese yields · Bloomberg, stocks gain on Berkshire commitment · Hedge Fund Alpha
• Index: WisdomTree DXJ · Morningstar DXJ performance · DXJ holdings
• Prices: 8001 · 8058 · 8031 · 8053 · 8002
• FX and rates: Federal Reserve G.5, August 2020 · Euronews, 10-year yield hits 3% · Investing.com USD/JPY
• Splits and guidance detail: Sumitomo 4-for-1 split · Mitsui FY2027 guidance · Mitsui buyback and return policy · Morningstar sector view