SoftBank’s $1.6B Profit Masks 90% Decline as OpenAI Bet Saves Vision Fund

SoftBank Group just posted numbers that tell two contradictory stories simultaneously. The Japanese investment giant reported 248.6 billion yen ($1.6 billion) net profit for its fiscal Q3 ending December 2025—technically a win versus last year’s $2.4 billion loss.

But strip away the narrative spin, and reality bites harder: profits plummeted 90% from the previous quarter’s $16.22 billion. The only thing preventing catastrophic losses? A massive $4.2 billion unrealized gain on OpenAI investments that masked declining valuations across nearly every other portfolio company, from Coupang to ByteDance.

The OpenAI Lifeline That Can’t Last Forever

SoftBank’s Vision Fund 2 gained 2.79 trillion yen ($18.28 billion) from OpenAI holdings during the quarter—essentially the entire profit engine keeping Masayoshi Son’s investment empire afloat. Without OpenAI’s valuation surge, the Vision Fund would have posted catastrophic losses rivaling the $27.4 billion bloodbath from fiscal year 2022. This dependency creates profound vulnerability: SoftBank has bet its recovery on a single, unprofitable AI startup whose business model remains unproven at scale.

The company reportedly committed $30 billion more to OpenAI in discussions that would value the startup at $750-830 billion—astronomical for a company burning billions quarterly on computing costs. CFO Yoshimitsu Goto defended the concentration risk, noting OpenAI represents early-stage monetization with “future revenue streams expected from enterprise sales, hardware and advertisements.” That’s venture capital optimism, not financial certainty.

India Portfolio: The Bright Spot Nobody’s Talking About

Buried in the earnings carnage lies unexpected resilience from SoftBank’s Indian investments. Vision Fund 2’s India portfolio—including foodtech giant Swiggy, e-commerce platform Meesho, EV-maker Ola Electric, and newly-listed Lenskart—performed substantially better than Chinese holdings. Meesho’s IPO particularly validated Son’s India thesis, listing at 45% premium to its ₹111 issue price, pushing market cap to ₹72,751 crore. SoftBank retained its 9.3-9.9% stake through SVF II Meerkat entity rather than cashing out, signaling continued confidence.

This divergence matters strategically. While Chinese investments like Didi Chuxing suffered from regulatory crackdowns and ByteDance faced valuation markdowns amid TikTok uncertainty, India’s tech ecosystem demonstrated monetization maturity. Swiggy, Meesho, Ola Electric, and Lenskart represent sectors—food delivery, Bharat-focused e-commerce, electric vehicles, and omnichannel eyewear—where Indian consumers actually pay for digital services, unlike many Chinese internet platforms subsidizing growth indefinitely.

Compare this to SoftBank’s historical track record. The Vision Fund initially deployed capital with breathtaking recklessness—Katerra, Wirecard, Zymergen—failures costing billions through inadequate due diligence. By contrast, India investments benefited from later-stage discipline and genuine market validation before deployment.

The Funding Desperation Behind Strategic Pivots

SoftBank’s pivot toward AI and robotics isn’t visionary positioning—it’s survival strategy funded by liquidating everything not nailed down. The company sold its entire Nvidia stake for $5.83 billion in October 2025, missing the subsequent AI boom that would have generated $150 billion returns. Between June-December, SoftBank dumped $12.73 billion in T-Mobile stock. It’s also taking loans backed by Arm Holdings, the chip designer it acquired and partially spun out, creating leverage atop already-concentrated risk.

This asset strip-mining finances both OpenAI obsession and diversification attempts like the $5.375 billion ABB Robotics acquisition announced October 2025. Son positioned robotics as “Physical AI”—combining ABB’s manufacturing expertise with SoftBank’s AI investments—but execution remains theoretical. The new “AI Computing Segment” lost 91.8 billion yen in nine months through December on higher headcount and Ampere acquisition costs, adding pressure rather than relief.

The strategic incoherence is striking. SoftBank simultaneously bets everything on OpenAI while launching SB OpenAI Japan to market enterprise solutions exclusively in Japan, acquires robotics businesses requiring massive capital expenditure, invests in agentic AI coding startup Emergent ($70M Series B), and maintains sprawling telecommunications operations. This isn’t portfolio diversification—it’s scattershot desperation dressed as strategy.

What Vision Fund Performance Actually Reveals

Strip away quarterly accounting and the Vision Fund’s structural problems become obvious. Vision Fund 1 public portfolio companies declined 2.1% quarter-over-quarter, primarily from Coupang share price drops, while private investments fell 3.3%. Vision Fund 2 dropped 3.7% from prior quarter. These aren’t minor corrections—they represent systematic overvaluation of late-stage startups that never achieved sustainable economics at their funding valuations.

Barron’s associate editor Eric J. Savitz characterized the SoftBank Vision Fund as a “failed experiment” in 2022, and nothing since contradicts that assessment. The fund’s winners-take-all strategy—flooding companies with capital to achieve market dominance—worked brilliantly for early investments like Coupang’s IPO (generating $36.99 billion profit in 2021) but failed catastrophically for dozens of others.

The venture capital model typically expects 1-2 investments per fund to generate all returns. SoftBank’s problem isn’t failed investments—it’s the scale of failures relative to capital deployed. When you write $100 million minimum checks versus the $5-20 million typical VC Series A, losses compound exponentially.

The Uncomfortable Questions Nobody’s Asking

If OpenAI’s valuation corrects—whether from competition from Anthropic’s Claude, Google’s Gemini, or simply revenue growth disappointing relative to $750 billion expectations—where does SoftBank land? The company has effectively doubled down, committing $3 billion annually to OpenAI technology while increasing ownership stake. This concentration violates basic portfolio management principles that made Son successful acquiring Alibaba early and selling before peak euphoria.

SoftBank’s stock gained 45% in 2023 on AI enthusiasm, but investors increasingly question whether the company can navigate volatile technology markets while managing massive debt loads and concentrated bets. The Q3 results technically beat loss expectations but missed profit estimates, suggesting Wall Street analysts remain skeptical of the sustainability narrative.

The fundamental issue persists: SoftBank transformed from telecommunications operator to the world’s largest technology-focused venture capital fund, then struggled to generate returns commensurate with capital deployed. Vision Fund 1 raised $100 billion with Saudi Arabia’s Public Investment Fund contributing $45 billion, Mubadala $15 billion, and SoftBank $28 billion. That capital base demands returns that few venture portfolios historically achieve, even fewer when deploying at the scale and valuation points SoftBank targets.

For now, OpenAI gains paper over cracks. But unrealized gains aren’t cash, valuations aren’t exits, and concentration isn’t diversification. Son’s grand vision of investing across the AI value chain—from chips to applications—requires flawless execution across multiple complex industries simultaneously. History suggests betting against that level of sustained excellence is rational, not cynical.