SoftBank Says Vision Fund 2 is $21.8 Billion Ahead. One Company is Doing All the Work

On SoftBank’s earnings slide, Vision Fund 2 has finally crossed the line from cautionary tale to apparent comeback.

The fund has invested a cumulative $122.6 billion. Its investments have produced a cumulative return of $144.4 billion through fair value, sale proceeds, interest and dividends. The difference is $21.8 billion.

That should be one of the most remarkable recoveries in venture-capital history. Vision Fund 2 survived the technology crash, the collapse of startup valuations and the damage inflicted by investments made near the top of the funding boom. It now appears to be comfortably ahead.

Then comes the number printed quietly inside the same SoftBank chart.

OpenAI accounts for $99.6 billion of Vision Fund 2’s cumulative return. Its investment cost at September-end was $54.6 billion.

Remove those two numbers and the fund’s remaining record changes completely. Vision Fund 2 excluding OpenAI has cost approximately $68 billion and generated a cumulative return of about $44.8 billion. The shortfall is $23.2 billion.

For every dollar invested outside OpenAI, the fund has returned approximately 66 cents on SoftBank’s own reported basis.

The $45 billion paper gain on OpenAI has not merely added sparkle to Vision Fund 2. It has lifted the entire fund from a large cumulative deficit to a reported surplus.

How does SoftBank get to the $21.8 billion gain?

SoftBank’s cumulative investment-return chart for September 30 reports $122.6 billion of investment cost and $144.4 billion of cumulative return for Vision Fund 2. The return figure combines the fair value of investments still held with sale proceeds from investments already exited, plus interest and dividends.

Subtracting cost from return produces the headline $21.8 billion surplus.

The calculation is gross. SoftBank says the return is presented before deducting third-party interests, taxes and expenses. It is not a measure of cash profit available to shareholders, nor is it the same as the return an outside investor would have received.

There is a second disclosure in the Vision Funds investor briefing. That performance snapshot reports $123.5 billion in acquisition cost, $21.1 billion in cumulative gross investment gains and $144.7 billion in total fair value.

The two sets of figures are close but not identical because SoftBank presents them on different bases and with different classifications. Our ex-OpenAI calculation uses the $122.6 billion and $144.4 billion figures from the SBG-basis chart because that same chart separately identifies OpenAI’s $54.6 billion cost and $99.6 billion value.

This is not a number SoftBank publishes. It is a subtraction from four numbers that it does.

How did one investment reverse the fund’s record?

SoftBank’s OpenAI investment was worth $99.6 billion at September-end, producing a cumulative unrealised gain of approximately $45 billion on the $54.6 billion then invested.

That gain is more than twice Vision Fund 2’s entire reported surplus.

The relationship can be expressed simply. The rest of the fund is approximately $23.2 billion below cost. OpenAI is approximately $45 billion above cost. Combine the two and Vision Fund 2 finishes $21.8 billion ahead.

SoftBank completed another $10 billion OpenAI investment on October 1, taking its cumulative commitment to $64.6 billion for roughly 13 per cent ownership. That final tranche is not included in the September-end calculation.

The reported $45 billion gain was also unchanged during the latest half-year. SoftBank made additional investments in April and July, but it did not increase OpenAI’s valuation from the previous financial year-end. This means OpenAI did not rescue the latest quarter. Its existing cumulative mark is rescuing the fund’s historical scorecard.

That distinction matters because the headline can otherwise suggest that OpenAI produced a fresh $45 billion windfall in three months. It did not.

Is the $45 billion gain real money?

It is real under fair-value accounting. It is not yet realised cash.

OpenAI remains a private company. SoftBank values the shares using the price and terms of investment transactions, along with the information available to it. The holding can therefore contribute to reported investment gains and net asset value before SoftBank sells a share.

The value becomes cash only through a sale, a secondary transaction, a public listing or another monetisation arrangement.

SoftBank has already found a way to obtain liquidity without selling. In August, a Vision Fund 2 subsidiary drew a $10 billion loan under an agreement that references the value of its OpenAI preferred shares. The shares are not formally pledged, but a significant decline in their value can trigger additional cash collateral or mandatory repayment.

The loan demonstrates why a private valuation matters even before an exit. It can influence borrowing capacity and financing conditions. It also raises the stakes if the valuation moves down.

SoftBank itself warns that there is no assurance unrealised investments will eventually be sold at the values used in its presentation. Actual proceeds will depend on company performance, market conditions, transaction costs and the timing of any sale.

What happened to the rest of Vision Fund 2?

Vision Fund 2 was announced in July 2019 with an expected capital pool of approximately $108 billion and memoranda of understanding from companies and financial institutions including Apple, Microsoft and Foxconn. SoftBank initially said it intended to contribute $38 billion.

The grand syndicate did not materialise as planned. SoftBank began deploying its own capital while trying to rebuild investor confidence after the first Vision Fund’s WeWork debacle. By late 2020, Masayoshi Son acknowledged that SoftBank was effectively the sole investor in the second fund at that stage.

The timing proved brutal.

Vision Fund 2 expanded rapidly during the era of cheap capital and soaring private valuations. When interest rates rose and technology stocks collapsed, those marks reversed. SoftBank’s Vision Fund operation recorded a $26.2 billion quarterly loss in early 2022 and another $23.1 billion loss in the following quarter across its investment funds.

The portfolio contained genuine winners, but the losses were spread across a large collection of public and private companies acquired at aggressive prices. The current results show that recovery outside OpenAI remains incomplete.

SoftBank’s latest investor briefing reports that private investments represented 89 per cent of the unrealised fair value across Vision Fund 1, Vision Fund 2 and the Latin America funds. Private marks are not fictitious, but they are less continuously tested than listed share prices. They may be supported by a new funding round, comparable-company movements or internal valuation models rather than a liquid market.

That makes the apparent recovery dependent on both company performance and the ability to exit at something close to the reported values.

Does OpenAI make Vision Fund 2 successful?

In one sense, absolutely.

Venture capital is built around power-law returns. A small number of extraordinary winners are expected to compensate for numerous losses. Funds do not need every company to succeed. They need one or two investments capable of returning the portfolio.

SoftBank can therefore argue that OpenAI is not distorting the model. It is validating it. Son placed the largest bet behind what SoftBank believes will become a generational company, and the increase in its value has repaired billions of dollars of damage elsewhere.

But the scale creates an unusual version of the power law.

OpenAI represents about 69 per cent of Vision Fund 2’s cumulative return value and approximately 68 per cent of the $147 billion fair value SoftBank attributes to the combined funds’ late-stage portfolio. A single private holding is therefore doing more than providing the fund’s biggest win. It is dominating its reported performance and its prospective listing pipeline.

The fund’s results have become less diversified at the moment they have turned positive.

Why does the difference between paper value and cash matter?

Vision Fund 2’s $144.4 billion cumulative return should not be confused with distributions.

The separate fund snapshot says Vision Fund 2 had distributed $23.7 billion by September-end. Those distributions can include proceeds from investment sales, financing transactions and other specified payments to limited partners. They are far below the reported total fair value.

Most of the fund’s apparent value therefore remains tied to investments that have not been fully monetised. OpenAI is the largest by a considerable distance.

This does not invalidate the gain. It changes the question investors should ask. The central issue is no longer whether SoftBank can identify a transformative company. It plainly has. The issue is whether it can eventually convert enough of that private value into cash without damaging the price that currently supports the fund’s recovery.

An OpenAI listing could create liquidity and a visible market price. It could also expose the holding to daily volatility. A secondary sale could realise part of the gain but reduce SoftBank’s participation in future upside. Borrowing against the value retains ownership but adds interest costs and repayment conditions.

Every monetisation route carries a trade-off.

Has Vision Fund 2 completed its comeback?

Its arithmetic is positive. Its recovery is concentrated. Its cash realisation remains limited compared with the value on the slide.

The $21.8 billion surplus is therefore both accurate within SoftBank’s disclosed framework and incomplete as a description of the fund’s health.

OpenAI has demonstrated what the Vision Fund model was designed to capture: a rare company capable of becoming more valuable than dozens of weaker investments combined. At the same time, the fund’s ex-OpenAI shortfall shows how much capital was lost or remains impaired across the wider portfolio.

SoftBank does not need all its bets to work. That is the nature of venture investing. It does, however, need one private-company valuation to keep doing an extraordinary amount of work.

Vision Fund 2 is $21.8 billion ahead on SoftBank’s cumulative-return slide. OpenAI is $45 billion of the reason.