SoftBank Is Betting $64.6 Billion on OpenAI - Can It Pay Off?

SoftBank is no longer merely investing in the artificial intelligence boom. It is borrowing at high interest rates to place one of the largest concentrated technology bets in corporate history. If its final planned tranche closes, the Japanese group expects to have invested $64.6 billion in OpenAI for an ownership interest of about 13%. The investment could become Masayoshi Son's next Alibaba-sized success. It could also leave SoftBank owning a valuable company without earning a sufficiently high return on the price and financing risk it accepted.

Let's break down how much SoftBank is putting into OpenAI, why the latest bond sale changes the risk and what OpenAI's revenue, valuation and cash needs imply for investors. We will also examine the wider AI funding chain and the consequences for US technology companies and Indian investors.

What is the latest SoftBank-OpenAI deal?

On September 21, 2026, SoftBank launched a proposed sale of $10 billion of US dollar-denominated senior unsecured notes and €1 billion of euro-denominated notes. Reuters reported that the dollar bonds were divided across 3.5-year, 5.5-year and 7.5-year maturities while the euro bonds had four-year and six-year maturities.

The primary purpose is to fund the third $10 billion tranche of SoftBank's follow-on investment in OpenAI. That payment is scheduled for October 1, 2026. Part of the proceeds may also be used for general corporate purposes. Pricing was expected on September 24 and settlement on September 29, which means the final coupons and proceeds were not known as of September 22.

If completed at the proposed size, Reuters said the offering would be the largest Asia-Pacific and Japan non-financial corporate bond deal on record. Fitch assigned the proposed notes a BB+ rating. This matters because SoftBank is not issuing low-cost investment-grade debt to finance a mature cash-generating asset. It is using speculative-grade borrowing to increase its exposure to a private company whose future cash requirements remain unusually large.

The date distinction is important. SoftBank had launched the financing by September 22 but the bonds had not yet been priced. Reported initial yield discussions of roughly 9% to 10% therefore represented market indications rather than final borrowing costs.

How did SoftBank's OpenAI investment reach $64.6 billion?

SoftBank said in February 2026 that it had already invested an aggregate $34.6 billion in OpenAI through Vision Fund 2 since September 2024. It then committed another $30 billion in three equal tranches at a $730 billion pre-money valuation.

The first $10 billion tranche closed on April 1 and the second closed on July 1. The third is planned for October 1. Once all three are completed, SoftBank expects its cumulative investment to reach $64.6 billion and its ownership interest to be approximately 13% on the fully diluted basis described by the company.

The $64.6 billion figure is therefore not a single cheque and it was not all funded at the same valuation. It combines earlier investments with the three 2026 follow-on payments. This distinction matters when calculating SoftBank's overall return because its original shares have already appreciated significantly while the latest capital is being invested at a much higher valuation.

Why is SoftBank willing to make such a concentrated AI bet?

OpenAI has already reached a commercial scale that few private technology companies have achieved. In March 2026, the company said it was generating $2 billion in monthly revenue, had more than 900 million weekly active ChatGPT users and had crossed 50 million subscribers. It also said enterprise products contributed more than 40% of revenue and its application programming interfaces were processing more than 15 billion tokens per minute.

OpenAI also closed $122 billion of committed capital at an $852 billion post-money valuation in March. SoftBank co-led that round alongside other global investors. Amazon, Nvidia, Microsoft and several large financial institutions participated.

Those numbers explain Son's conviction. OpenAI is not being valued only as a chatbot company. The bullish case assumes that it can become a broad layer of computing infrastructure across consumer search, coding, enterprise software, agents, commerce and scientific work. If that happens, today's revenue could represent only a small part of the eventual opportunity.

The strategic fit is also wider than SoftBank's equity stake. SoftBank controls most of Arm Holdings, owns AI infrastructure and energy assets and is expanding into robotics and data centres. OpenAI can potentially create demand across this network. In simple terms, SoftBank is trying to own part of the intelligence layer, part of the chip architecture and part of the physical infrastructure that delivers AI.

The real gamble is valuation, not whether AI will grow

It is easy to frame the SoftBank-OpenAI deal as a simple question of whether AI succeeds or fails. That is the wrong test. AI adoption can continue growing quickly and SoftBank can still earn a disappointing return if it paid too much or if future funding rounds dilute its stake.

OpenAI's official March numbers implied a $24 billion annual revenue run rate. Against the $852 billion post-money valuation, that is roughly 35.5 times annualised revenue. The comparison is not a formal price-to-sales multiple because the valuation includes fresh cash from the funding round and OpenAI is private. It is still a useful indicator of how much future growth is already embedded in the price.

The Financial Times reported in September that OpenAI expected 2026 revenue of about $36 billion and 2030 revenue of $350 billion. These are internal forecasts reported by the publication and not audited guidance. Achieving that path would require revenue to rise at approximately 76.6% a year for four years.

These ratios do not prove that OpenAI is overvalued. Early-stage platform businesses can grow into apparently extreme valuations. They do show that investors are paying today for several years of exceptional execution. Slower adoption, lower pricing or heavier competition can hurt the investment even if OpenAI remains the category leader.

What return does SoftBank actually need from OpenAI?

A simple ownership model shows the hurdle clearly. If SoftBank ultimately owns 13% after investing $64.6 billion, the stake's look-through value equals 13% of OpenAI's future equity value. The table below assumes no additional dilution and ignores taxes, financing costs and any change in ownership.

The calculation reveals the most important point in the deal. SoftBank needs OpenAI to be worth about $1 trillion in five years merely to earn a gross annualised return of around 15% under these simplified assumptions. A 20% return requires a valuation of roughly $1.24 trillion while a 25% return requires around $1.52 trillion.

The actual hurdle is higher. SoftBank must pay interest on the debt used to finance part of the investment. OpenAI may issue more shares to employees or raise more capital. Taxes and transaction costs also reduce realised returns. A future OpenAI valuation above $1 trillion might sound spectacular but it would not automatically make this an extraordinary investment from SoftBank's entry point.

Why OpenAI's cash burn changes the equation

Revenue growth and cash generation are not the same thing. OpenAI must pay for data centres, chips, power, model training, inference and highly specialised employees before much of its revenue can become free cash flow.

The Financial Times reported on September 19 that OpenAI expected cumulative negative free cash flow of almost $280 billion between 2026 and 2030. It also reported projected spending of about $856 billion over that period. These figures are forecasts and can change materially but they show why OpenAI keeps raising capital despite its rapid revenue growth.

One way to understand the economics is to think of OpenAI as building an airline and designing its aircraft at the same time. Every new route can bring more passengers and revenue but the company must keep buying aircraft, building airports and improving the engines. Scale is valuable only if revenue eventually rises faster than the cost of serving each user.

OpenAI argues that more compute produces better models, better products and more revenue while hardware and algorithmic improvements lower the cost per unit of intelligence. That is the positive flywheel. The risk is that competitors create capable models at lower prices and force OpenAI to pass cost savings to customers instead of retaining them as profit.

This makes the reported $280 billion cash-burn estimate more than an OpenAI issue. It is a funding-duration test. OpenAI needs capital markets to remain willing to fund the gap between fast-growing revenue and even faster infrastructure spending. SoftBank needs OpenAI's valuation to compound while it is helping finance that gap.

SoftBank can afford the bet today, but the balance sheet is becoming more sensitive

SoftBank's latest reported balance sheet was stronger than the size of the OpenAI commitment might suggest. At June 30, 2026, the group reported adjusted stand-alone holdings worth ¥83.11 trillion and adjusted stand-alone net interest-bearing debt of ¥10.81 trillion. That produced a loan-to-value ratio of 13%, below its policy ceiling of 25% under normal market conditions.

Its net asset value rose to ¥72.30 trillion from ¥40.06 trillion at the end of March. OpenAI and Arm were major contributors to the wider increase in asset value. A low LTV gives SoftBank room to borrow because the market value of its holdings is much larger than adjusted net debt.

The same data also exposes the vulnerability. A large share of SoftBank's asset value depends on market prices and private-company valuations. If Arm shares fall or OpenAI is marked down, the denominator in the LTV calculation shrinks. Debt does not fall simply because asset values decline.

SoftBank's finance cost rose 98.9% year on year to ¥328.7 billion in the June quarter. Interest expense at the parent increased to ¥265.8 billion, partly because of borrowing backed by Arm shares and bridge loans used for OpenAI. Net income attributable to shareholders fell 17.7% to ¥347.3 billion even though net sales rose 10.9% to ¥2.02 trillion.

SoftBank is not facing an immediate solvency problem based on the latest reported LTV and liquidity. The risk is reflexive. Rising AI valuations increase SoftBank's NAV and borrowing capacity which allows it to invest more in AI. Falling valuations can reverse that process and force asset sales or more expensive refinancing at an uncomfortable time.

The AI funding loop now reaches chips, clouds, data centres and power

The OpenAI transaction is part of a much larger capital chain. OpenAI raises equity from companies including SoftBank and Nvidia. It then spends heavily with chip suppliers and cloud partners such as Nvidia, Microsoft, Oracle, Amazon and other infrastructure providers. Some of those partners also invest in OpenAI or commit capital to connected data-centre projects.

This is not proof that AI demand is artificial. OpenAI's user and revenue growth demonstrate real adoption. It does mean that investors should separate three types of activity: external customer revenue, capital raised from investors and contractual infrastructure commitments. Combining all three into one headline can make the economic flywheel look stronger than the cash ultimately generated by end customers.

SoftBank and OpenAI each invested $500 million in SB Energy in January 2026. OpenAI also signed a 1.2-gigawatt data-centre lease for an initial site in Texas. This gives SoftBank exposure to both sides of the transaction: it owns part of the AI company demanding compute and part of the infrastructure ecosystem supplying it.

That structure creates upside if utilisation is high. It also concentrates risk if OpenAI slows spending. The health of the AI ecosystem therefore depends increasingly on whether consumer subscriptions, enterprise software and developer usage can support the capital promised for chips, clouds, data centres and electricity.

Investors looking for listed exposure to this chain can compare AI stocks and technology stocks but the risks are not identical. Chip companies earn when infrastructure is purchased. Cloud companies need contracts to turn into high-utilisation revenue. SoftBank needs the equity value of its portfolio to rise faster than its financing cost.

What does the SoftBank-OpenAI bet mean for Indian investors?

Indian investors cannot generally purchase OpenAI shares like a listed US stock. The exposure available in public markets is indirect and each route represents a different part of the AI value chain.

SoftBank's OpenAI stake can affect the value of SoftBank Group while Arm provides a separately listed chip-architecture exposure. Nvidia supplies a large part of the computing stack. Microsoft and Oracle supply cloud and infrastructure capacity. These companies may benefit from OpenAI's growth but they also carry different valuation, customer-concentration and capital-spending risks.

India is also becoming an important demand and infrastructure market for OpenAI. In February 2026, OpenAI said India had more than 100 million weekly ChatGPT users. It announced a partnership with Tata Group that begins with 100 megawatts of local AI-ready data-centre capacity and may scale to one gigawatt over time.

The Indian opportunity is real but user growth does not automatically equal high revenue. Local pricing, enterprise adoption, data-residency requirements and competition from domestic and open-source models will determine monetisation. For Indian investors the useful question is not simply which company is connected to OpenAI. It is which company can turn AI demand into durable free cash flow without taking disproportionate balance-sheet risk.

What could make SoftBank's OpenAI investment work?

The bullish outcome requires several things to happen together.

First, OpenAI must convert its enormous user base into higher recurring revenue. Enterprise adoption matters because business contracts can be larger and more predictable than consumer subscriptions.

Second, the cost of inference must fall faster than pricing. If each query becomes cheaper to serve while users move toward higher-value workflows, gross margins can improve even as usage expands.

Third, OpenAI must maintain a strong competitive position across models, agents and distribution. A technology lead is valuable only if customers remain willing to pay for it.

Fourth, capital spending must eventually generate operating leverage. A data centre is attractive when it remains highly utilised. It becomes a burden when capacity is built ahead of durable demand.

Finally, SoftBank must protect its ownership percentage and refinancing flexibility. The headline OpenAI valuation can rise while SoftBank's return disappoints if repeated funding rounds dilute its stake or if interest costs absorb too much of the gain.

What could go wrong with the AI funding gamble?

The first risk is valuation compression. A private company can grow revenue rapidly and still receive a lower multiple when interest rates rise or investor enthusiasm cools.

The second is funding dependence. OpenAI's reported cash needs are large enough that access to capital remains a strategic requirement rather than a convenience.

The third is competition. Anthropic, Google, Meta and Chinese AI developers can pressure pricing and shorten the useful life of expensive model investments.

The fourth is infrastructure execution. Delays in securing land, power, chips and regulatory approvals can push cash outflows forward before revenue arrives.

The fifth is concentration at SoftBank. OpenAI and Arm now have an outsized influence on its NAV. That makes the group more coherent as an AI investment company but less diversified when the AI cycle turns.

The sixth is accounting volatility. SoftBank measures OpenAI shares at fair value through profit or loss. A higher private valuation can create a large accounting gain without generating cash while a markdown can reduce reported profit even if OpenAI's operating business continues to grow.

What should investors track after the SoftBank bond sale?

The bond pricing on September 24 is the first near-term checkpoint. The more important evidence will arrive over several years through OpenAI's revenue quality, gross margins, cash consumption and need for fresh equity.

Is SoftBank's $64.6 billion OpenAI bet worth the risk?

SoftBank's investment is not irrational. OpenAI has an exceptional user base, a fast-growing commercial business and a credible chance to become one of the central platforms of the AI economy. SoftBank also entered before the latest reported valuation discussions and had already recorded a $45 billion cumulative unrealised gain on its OpenAI investment by June 2026.

However, the attractive company and the attractive investment are not the same thing. At SoftBank's expected $64.6 billion cost and roughly 13% ownership, OpenAI needs to approach a $1 trillion valuation within five years to produce a simplified gross return of about 15% a year. The hurdle rises once debt costs, dilution and taxes are considered.

Our view is that the most likely mistake is not assuming that AI will disappear. It is assuming that industry leadership guarantees superior investment returns at any entry price. SoftBank is using leverage to accelerate exposure to a business that may need hundreds of billions of dollars before it becomes sustainably cash generative. That combination can create extraordinary upside but it leaves little room for merely good execution.

The $64.6 billion bet will ultimately be judged by three numbers rather than the excitement around AI: OpenAI's free cash flow, SoftBank's ownership after future funding rounds and the gap between SoftBank's equity return and its cost of capital. Everything else is narrative until those numbers begin to prove the case.