The Malta Independent 4 October 2026, Sunday
View E-Paper

Can we learn from Softbank’s AI adventures?

George M Mangion Sunday, 4 October 2026, 08:00 Last update: about 1 day ago

SoftBank Group was founded in 1981 by Masayoshi Son and is headquartered in Tokyo. It has moved repeatedly into new strategic arenas, from software distribution to telecom, internet platforms, and large-scale AI technology. Currently, its management team sits in London and is best understood as a hybrid of three things: a founder-led investment holding company, a Japanese telecom and internet platform through its controlling stake in SoftBank Corp, and the majority owner of Arm, one of the most important semiconductor intellectual-property companies in the world.

The Vision Funds are SoftBank Group's late-stage technology investment vehicles, built on an "AI revolution". They sit beside SoftBank's older operating businesses and its direct stakes (Alibaba historically, Arm now). There are three related investment pools: Vision Fund 1 (SVF1), Vision Fund 2 (SVF2), and the smaller LatAm Funds. After buying Arm in 2016, SoftBank repositioned itself as a strategic investment company. Capital came from internal sources plus large outside limited partners, most famously Saudi Arabia's Public Investment Fund and Abu Dhabi's Mubadala, with smaller tickets from names such as Apple. The fund was designed to write very large payments into late stage "unicorns" in AI, mobility, consumer internet, and enterprise software. The style was concentrated, high-conviction, and often valuation-inflating - SoftBank would pay up to become the defining late-stage backer. Early paper marks looked strong as private rounds and a wave of listings re-rated the portfolio.

In 2019, SoftBank announced a sequel, SVF2, to the previous SVF1, and a Latin America vehicle. SVF2 was originally marketed as another mega-fund with outside LPs, which included Apple, Microsoft, Foxconn, Japanese banks, and others. After "WeWork" and a cooling in LP appetite, most of that external capital did not close. SVF2 became, in practice, almost entirely SoftBank-funded, with smaller investments and more companies than SVF1.

In 2023, SVF1 sold its Arm stake to a SoftBank subsidiary ahead of Arm's Nasdaq IPO. That created a large segment gain inside Vision Fund, but Arm marks no longer flow through Vision Fund P/L because Arm is now consolidated. Arm's later re-rating became a SoftBank net asset value story more than a Vision Fund mark-to-market story. To date, SoftBank Group Corp. has 73,677 consolidated employees worldwide.

Next, we shall visit the history of SoftBank Vision Fund. These funds are late-stage technology investment vehicles, built around Masayoshi Son's bet on an "AI revolution". They sit beside SoftBank's older operating businesses and its direct stakes. Regrettably, "WeWork" was an emblematic problem, with a failed IPO, a collapsed valuation, then a governance crisis, and later bankruptcy. Combined with the 2020 market crash, it forced write-downs, staff cuts at the adviser level, and a more cautious posture. SoftBank talked about shifting from "defense" back to "offense". Then rates rose, growth stocks sold off, and private marks caught up. SoftBank slowed new investing, yet focused on liquidity and NAV, and prepared Arm for a public listing. Arm succeeded in its realisation, having a narrower AI focus.

From 2023 onward, the funds emphasised new monetisation and distributions from SVF1 with selective new bets in SVF2 / LatAm, increasingly framed around AI, including later-stage AI platforms. The Vision Funds did not invent SoftBank's taste for scale - Alibaba and Arm came first - but they pushed to industrialise it. They changed late-stage venture by making $1bn+ rounds normal, then showed the cost of that model when liquidity vanished.

SoftBank Group narrowed its full-year loss and returned the Vision Funds segment to an investment gain. The improvement came mainly from public-market recoveries and an internal Arm transaction - not a broad wave of cash-generating exits.

Group results: Sales rose modestly. Pre-tax results swung back into profit, while the Group still posted a net loss for the year - yet much smaller than the prior year. Vision Funds moved from a very large prior-year loss into an overall investment gain - the first annual gain in three years. That segment figure includes a large intra-group gain from SVF1 selling its Arm stake to a SoftBank subsidiary. After those subsidiary-related gains are accounted for, Vision Funds still showed a modest loss in the consolidated income statement.

Since inception through March 2024, SVF1 was in a cumulative gross gain and SVF2 in a cumulative gross loss. This resulted from its listed technology holdings and selected private marks, which recovered versus the prior year's deep markdowns. Thus, recovery smiled at SoftBank, with realised gains rising sharply, including exits and the Arm transfer. It is true that it was Arm's listing and later share-price strength that lifted SoftBank's NAV, but ongoing Arm marks do not flow through Vision Fund or consolidated investment P/L since Arm is a consolidated subsidiary. The large Vision Fund segment gain for the year includes the intra-group sale of SVF1's Arm stake, which is eliminated on consolidation.

Thus, Vision Funds improved sharply at the segment level and helped cut the Group's annual net loss, but results stayed valuation-driven, uneven, and volatile. The year was a recovery in marks and internal transactions more than a broad exit cycle.

The largest single-year swing came from marking up the OpenAI stake, mostly inside Vision Fund 2. In the year ended March, SVF2 booked multi-trillion-yen investment gains dominated by OpenAI. SoftBank committed tens of billions more, partly via bridge loans and later bond issues, and treated OpenAI as the centrepiece of its "all-in on AI" thesis. That mark-to-market gain is what flipped the Vision Fund segment from large losses to large profits.

A Vision Fund 3 has been rumoured for years; yet so far, it has not become another $100 billion product in the way Fund 1 was. Thus ends the cheerful story of Masayoshi Son, and practitioners are encouraged, together with FinanceMalta, to attract some of these funds at home.

 

George M. Mangion is a senior partner at PKF Malta 


  • don't miss