Vision Fund I, The Strangest LP Capital Structure in VC History, $40B Preferred at a 7% Coupon

By 2016-2017, global venture capital had entered a phase where Uber, Airbnb, Didi, and Xiaomi rounds individually exceeded the size of most existing VC funds. Sequoia, a16z, and Accel were running flagship funds in the $1-3B range; no single LP could anchor a $10B+ round. Sovereign wealth funds, Saudi PIF, Abu Dhabi Mubadala, Singapore GIC, Norway's NBIM, began stepping in directly as LPs, shifting the gravity of PE and VC capital from US pensions and endowments toward oil-driven sovereigns and Asian state pools. The Vision Fund was both the apex and the explosion of that shift.

The Vision Fund's arrival created a distinct Vision Fund premium layer in global late-stage valuations, rounds led solo by the Vision Fund chronically traded above their later IPO or secondary marks. That gap is widely cited as one of the structural causes of the 2019-2022 mega-VC cycle unwind. PIF and Mubadala's LP participation became a centerpiece of Saudi Vision 2030 and UAE Economic Vision 2030's non-oil capital deployment strategy. After Vision Fund 2 failed to raise external commitments in 2019, both sovereigns effectively stepped back from anchoring global mega-VC vehicles.

SoftBank Group was founded in 1981 in Fukuoka by Masayoshi Son. By the 1990s it had built Yahoo! Japan, by 2000 it had made its now-legendary $20M investment in Alibaba (worth more than $50B at the 2014 IPO), and by 2013 it had acquired Sprint. The 2016 acquisition of Arm Holdings for $31B set the stage for the 2017 Vision Fund launch, positioning SoftBank Group as the world's largest technology private investment platform. At the time of the fund's first close, SoftBank Group's FY2016 (year ended March 2016) revenue was roughly ¥9.16 trillion with operating profit of ¥1.03 trillion and total assets of ~¥24.5 trillion. Alibaba, Sprint, and Arm were core balance-sheet assets; the parent itself committed $28B directly to the fund.

Vision Fund I was set up as a Cayman LP with SB Investment Advisers (UK) Ltd, FCA-authorised in the UK, as the operating GP. Fund term was 12 years (2017-2029, with potential extension). The structural innovation was the tranching of LP capital into Class A common (~60%) and Class B preferred (~40%). Class B paid a 7% annual cash coupon for the full 12 years regardless of fund performance, functionally a fixed-income claim against fund NAV with SoftBank Group as the implicit credit backstop. Class A took no coupon and captured carry and upside only. SoftBank Group committed its entire $28B as Class A common, voluntarily occupying the first-loss layer. PIF and Mubadala took a blend of preferred and common, protecting their downside through the coupon while retaining some upside.

Vision Fund I was a fund formation rather than a typical M&A transaction, so the advisor lineup differs from standard precedents. LP-side legal and tax advisors negotiated commitment terms, while GP-side counsel structured the Cayman LP and UK GP vehicle. Financial advisory was largely run in-house at SoftBank (CFO office plus continuity advisors from the Arm acquisition). Several mandates remain unconfirmed in public disclosures.

Note: Advisor mappings are based on public disclosures and market observation. Private fund LP negotiations are typically confidential and some assignments remain unconfirmed.

Unit: ¥100M | J-GAAP/IFRS consolidated | Source: SoftBank Group annual reports. FY2018 operating profit jump reflects Vision Fund valuation gains (Uber, WeWork, etc. mark-ups).

In this deal, valuation refers not to an acquisition price but to the pricing of the LP capital structure itself. The key economics are the 40/60 preferred/common split and the 7% coupon. From a PIF or Mubadala perspective, the preferred tranche is a debt-like claim on the fund: 12 years of 7% coupon plus principal, with SoftBank Group's balance sheet effectively standing behind the credit. The common tranche carries no coupon and depends entirely on residual upside. SoftBank Group committed all of its $28B to the common tranche, voluntarily backstopping the preferred coupon obligation against its parent balance sheet. The transaction is, in substance, SoftBank Group issuing a $40B emerging-market tech-fund bond at a 7% coupon and laying an equity layer on top of it.

Note: Preferred/common split percentages and per-LP tranche allocations are estimated from market reporting and partial disclosures by SoftBank and individual LPs. Exact figures are visible only in fund settlement statements.

Nine years later the verdict is clear. The fund did generate a cumulative gross gain of roughly $22B, driven by Coupang, DoorDash, Slack, the Arm IPO, and ByteDance markups (PitchBook 2025), but gross IRR has stabilised around 6-7%, almost exactly at the level of the preferred coupon. In other words, residual returns to the common tranche (SoftBank Group's $28B and the ~$10.7B of common held by PIF and Mubadala) sit close to zero. Cumulative writedowns on WeWork exceeded $11B; Wag and Brain Corp collapsed; OYO's valuation was severely impaired. Vision Fund 2, launched in 2019, failed to attract external commitments, PIF and Mubadala declined to participate, and SoftBank funded approximately $56B directly from the parent balance sheet, ultimately recognising cumulative losses well above $20B. The Vision Fund segment posted its first annual profit in three years in FY2024 (year ended March 2024, ¥724.3B), driven primarily by the Arm IPO of September 2023 and recovering portfolio marks. In February 2025, Masayoshi Son publicly apologised at an investor conference for failing to deliver returns to PIF. Meanwhile, Son has been repositioning SoftBank around an AI revolution thesis built on large OpenAI investments, deeper use of Arm, and the Stargate data center joint venture. External fundraising for a Vision Fund 3 is on indefinite hold.