SoftBank at 8,541: A 17% NAV Discount the Market Hasn’t Closed

Start with the gap. Even after the stock doubled in three months, SoftBank still trades at roughly a 17% discount to net asset value as last marked. That is the number that defines the setup, and it is the wrong number for a holding company whose two largest listed exposures are repricing upward in real time. The market gave the shares an 111% move from 4,046 to 8,541. It has not yet given them the discount compression that usually follows when monetization catalysts move from speculation to schedule.

The fiscal year that closed in March did most of the heavy lifting on the fundamental case. Revenue rose 7.7% to 7.79 trillion yen. Net income reached roughly 5 trillion yen, a record, carried by about 46 billion dollars in Vision Fund gains, the bulk of which traced to the mark on the OpenAI stake. This is the distinction that matters. The top line expanded while the portfolio marked higher. A holding company can manufacture a good year through marks alone. SoftBank did not have to.

Arm is the engine the look-through earnings run on. With SoftBank holding close to 90% of it, Arm’s fiscal fourth quarter — 23% revenue growth to 4.9 billion dollars — flows straight into the parent’s NAV per share. When Arm shares add double digits in a session on AI demand and Nvidia’s print, that is not sentiment leaking into the quote. That is the listed stake being marked to market, and a controlling stake at that.

So track the discount, not the level.

Here is the mechanism a holding-company investor should hold in front of everything else: NAV discount is the entire thesis, and its direction is what pays. A 17% discount on a portfolio dominated by Arm and OpenAI is the market saying it does not fully trust the marks, or it does not trust that the marks can be turned into cash without a haircut. Monetization collapses that doubt. SoftBank has IPO preparation underway for SB Energy and the robotics spin-off Roze, multi-currency bond issuance funding the AI infrastructure push, and an OpenAI position whose own listing is now openly discussed. Each of those is a mechanism to convert a paper mark into a realized one. Each realized mark narrows the discount the next holder is willing to demand. The buyback yield near 1.86%, against 4.8 trillion yen repurchased since FY2019, is the company doing the same work from the other side.

What gets missed in setups like this is the order of operations. Look back at how the market has historically treated conglomerates sitting on a single dominant listed asset before the spin-off arrives: the discount stays wide right up until the monetization is dated, then compresses fast once the cash path is visible. The error earlier observers make is reading the wide discount as a permanent feature of the structure rather than a temporary price on uncertainty. The pyramid structure of cross-shareholdings and listed stakes does not earn a perpetual penalty. It earns a penalty until the assets at the top start coming down as cash.

The backdrop reinforces rather than competes. Under the Takaichi government, Japan has named AI, semiconductors, and advanced communications as strategic sectors, layering subsidies, tax incentives, and regulatory easing through METI and the Industrial Competitiveness Enhancement Act. The US-Japan tech alignment channels capital and policy toward trusted infrastructure, and SoftBank’s announced data-center commitments in France and its US power-for-compute arrangements sit directly inside that flow. A softer Yen lowers the real cost of servicing the offshore debt funding all of it. None of this is the thesis. It is the environment that lets the monetization run faster than a hostile one would allow.

Domestically, the legacy telecom business is the ballast that the more speculative spin-off math is built on. It throws off the operating cash flow that funds capital recycling without forcing the company to sell into a weak credit window, which is exactly the flexibility that lets an IPO be timed for value rather than for liquidity.

Now the part that can break it. The 17% discount is a function of where Arm and OpenAI are marked, and both marks are AI-cyclical. Arm trades at a multiple that prices years of growth forward. If AI infrastructure spend slows and Arm’s revenue growth decelerates from the low-20s toward the single digits, the NAV the discount applies to shrinks, and a narrowing discount on a falling NAV is no gain at all. Gearing cuts the same way it helps. The bond issuance funding the infrastructure build inflates volatility on both sides of NAV, so a drawdown in the listed stakes hits the equity harder than the headline portfolio move suggests. And the OpenAI mark is the least liquid of the large positions. A private valuation that re-rates downward does not announce itself through a daily quote the way Arm does. It arrives at the next funding round.

That gives the clean line. If Arm revenue growth falls below 10% year-over-year and the OpenAI position is marked down at its next round, the NAV anchor under this 17% discount erodes and the re-rating case breaks. Short of that, the stock at 8,541 is being asked to carry a discount the catalysts are built to close, with a top line that grew while the portfolio marked higher rather than instead of it. The average analyst target of 6,829 sits below the current quote, which tells you the consensus is anchored to the discount as a constant. The whole argument here is that it is not one.

Figures reflect the most recent available data and may differ slightly from live market prices. · © Mathstock