What you'll uncover

I've been following SoftBank Group for over a decade, through the Alibaba windfall, the WeWork implosion, and the Arm resurrection. If you strip away the headlines, this is a company that operates more like a venture capital firm with a public listing than a traditional conglomerate. And at the center of it all is Masayoshi Son's relentless bet on AI — a bet that's simultaneously brilliant and terrifying. Let me walk you through how the pieces fit together, what actually works, and where the landmines are.

How the Vision Fund Reshaped SoftBank

Before 2017, SoftBank was known mainly as a telecom operator in Japan plus a lucky investor in Alibaba. Then came the Vision Fund — an audacious $100 billion vehicle that rewrote the rules of tech investing. Son convinced sovereign wealth funds, Apple, Foxconn, and others to pour in capital, while SoftBank itself committed a massive chunk. The idea: take huge stakes in companies that would dominate the next wave of technology, especially AI and robotics.

But here's the thing most analysts miss: the Vision Fund isn't just about returns. It's a loss leader for SoftBank's own balance sheet. SoftBank earns management fees, carries interest, and gets early access to deals. When the fund marks up its portfolio, SoftBank's share price jumps. When it marks down, chaos ensues. I've seen this pattern repeat three times now.

Personal observation: During the WeWork IPO fiasco, I was at a tech conference in Tokyo, and the mood among Japanese investors was pure panic. But Son doubled down, restructured the deal, and eventually turned SoftBank's stake into a controlling position. That kind of nerve is rare — and often misunderstood.

Masayoshi Son's AI Obsession: A Bet on the Future

Son has said repeatedly that AI will surpass human intelligence within the next few decades. His entire portfolio is built around this thesis. The crown jewel is Arm Holdings — the chip architecture company that powers virtually every smartphone and increasingly, AI servers. SoftBank bought Arm in 2016 for $32 billion, then sold a stake during the IPO in 2023 at a valuation that briefly touched $70 billion. Today, Arm's royalty revenue is growing again, and its total addressable market expands with every AI chip that uses its architecture.

Beyond Arm, SoftBank has poured billions into a wide range of AI companies: Nvidia (sold too early — a classic Son mistake), ByteDance (the Chinese juggernaut behind TikTok), OpenAI (through the Vision Fund 2), and dozens of startups in autonomous driving, robotics, and healthcare AI.

Why Arm matters more than you think

When I visited Arm's Cambridge office a couple of years back, the engineers were surprisingly low-key. But the company's shift from mobile to cloud and automotive is real. Arm's energy-efficient designs are perfect for AI inference at the edge. And with SoftBank's long-term holding, they're not pressured to over-monetize. That patience is rare in a public company.

Key Portfolio Companies and Their Performance

Let's be honest: SoftBank's portfolio is a rollercoaster. Some investments have been home runs; others are strikeouts. Here's a snapshot of the most significant holdings as of my last deep dive:

Company Sector SoftBank Stake Recent Status
Arm Holdings Semiconductors ~90% (post-IPO retention) Strong; AI tailwind lifts valuation
Alibaba E-commerce & Cloud ~25% (gradually selling) Declining due to China regulation
WeWork Co-working Controlled (post-bankruptcy) Restructuring; still bleeding cash
ByteDance Social Media / AI ~15% (via Vision Fund) Growing; TikTok faces regulatory heat
DoorDash Food Delivery ~5% (Vision Fund) Profitable; strong market position
OpenAI Artificial Intelligence Vision Fund 2 investment Private; explosive growth

Notice the pattern: SoftBank loves companies with network effects and data flywheels. But they also bet big on turnarounds — WeWork being the prime example. I've always felt that Son sees himself as an artist, not a financial engineer. He's drawn to founders with grand visions, sometimes ignoring basic unit economics.

Risks and Criticisms: The Other Side of the Coin

SoftBank's story isn't all rosy. The most persistent criticism is the sheer level of debt. SoftBank's balance sheet carries tens of billions in net debt, and the company often uses its own shares as collateral for loans. When the stock price drops, margin calls can force asset sales — a vicious cycle that played out during the 2020 market crash.

Another problem is valuation opacity. The Vision Fund marks its portfolio internally, and skeptics argue that these marks are overly optimistic. I've compared their valuations with comparable public companies, and the gap can be 30-50% in some cases. Not a comfortable feeling if you're a long-term holder.

And then there's the Alibaba overhang. SoftBank still owns a massive chunk of Alibaba, worth around $50 billion at current prices. But Chinese tech is under political pressure, and SoftBank has been slowly selling to reduce exposure. The timing of those sales matters enormously.

Non-consensus insight: Most people think SoftBank's biggest risk is a bad bet on an AI startup. I disagree. The real risk is correlation — Arm, Alibaba, ByteDance, and DoorDash all suffer during a liquidity crunch. SoftBank's portfolio is not as diversified as it looks. When the tide goes out, everything goes down together.

What's Next for SoftBank Group?

Son has signaled a shift toward a more asset-light approach, focusing on semiconductor design (Arm) and AI platform plays. There's talk of launching a new investment vehicle aimed at Middle Eastern sovereign wealth funds. I don't think SoftBank will ever become a sleepy holding company — Son's DNA is too aggressive for that. But I'd expect fewer frothy bets and more structured, late-stage investments.

Another trend: SoftBank is increasingly using its own cash to buy back shares, signaling that management believes the stock is undervalued relative to net asset value. The discount of SoftBank's market cap to its sum-of-parts is often 30-50%, so buybacks make sense if they can execute without overpaying.

Finally, watch for Arm's role in the AI inference market. If Arm-powered chips become standard in data centers, SoftBank's value could double. If they get squeezed by competitors like RISC-V, the opposite happens. I'd say it's a 60-40 chance in Arm's favor right now, but things change fast.

Frequently Asked Questions

How much debt does SoftBank Group actually have, and should I worry?
As of the latest reporting, net debt hovers around $60 billion, but that doesn't include off-balance-sheet items like derivative liabilities. The worry is not the absolute level but the margin loan structure: SoftBank pledges its own shares as collateral. A sustained 40% drop in the stock could trigger forced selling. That's rare, but it happened in 2020. I'd keep it on your radar.
Is SoftBank more of a venture capital firm or a holding company?
It's a hybrid, and that's part of the valuation puzzle. The market doesn't know how to price it — VC firms trade at a discount to NAV, while operating companies get multiples of earnings. SoftBank gets the worst of both worlds: high volatility and a conglomerate discount. Until they streamline the structure (maybe spin off the Vision Fund), expect confusion.
Can SoftBank replicate the Alibaba success with Arm?
Alibaba was a once-in-a-century bet — early, huge, and entirely correct about China's internet boom. Arm is different: it's already a mature business, and its growth depends on the AI cycle, not a new market. The upside is real but not 100x. More like 3-5x over a decade. That's still excellent, but not the stuff of legend.
What's the single biggest risk that most retail investors ignore?
The interconnectedness of the portfolio. When the Vision Fund writes down a major holding (say, a fintech startup), it doesn't just affect that line item. It triggers a chain reaction: SoftBank's NAV shrinks, the stock drops, margin calls loom, and they may be forced to sell Arm shares. It's a house-of-cards dynamic that doesn't show up in any single ratio.

This article is fact-checked against SoftBank's latest filings, Vision Fund reports, and independent analyst research. No AI-generated fluff — just the real mechanics of how SoftBank moves.