Markets love AI. Creditors, much less so. Why SOFTBANK is struggling to raise $6 billion

When SoftBank began accumulating stakes in OpenAI and backing its associated infrastructure projects, few observers imagined that the main obstacle would be neither technological, regulatory nor competitive. Yet the slowdown in talks to raise at least $6 billion through a loan secured against its OpenAI stake offers a reminder of a more fundamental reality: the artificial intelligence revolution is entering a phase in which financing is becoming as important as innovation.

According to Bloomberg, SoftBank has suspended discussions with several potential lenders over the proposed financing. A few weeks earlier, the Japanese group had already reduced its initial target from $10 billion to $6 billion in response to lenders’ reluctance. This caution is surprising given that OpenAI has just filed for an initial public offering, a step that might have been expected to reassure the markets.

The contrast is all the more striking because SoftBank has once again become Japan’s largest listed company by market capitalisation, overtaking Toyota. Its stake in Arm has benefited fully from the surge in AI-related stocks. The group is also sitting on substantial unrealised gains from OpenAI. Yet despite this favourable environment, converting those assets into several billion dollars of debt is proving more difficult than expected.

An OpenAI stake remains difficult to value

The heart of the problem lies in the nature of the asset being used as collateral. A margin loan allows an investor to borrow against a pledged financial asset. The amount provided by lenders generally corresponds to a fraction of the collateral’s estimated value.

This mechanism is common for listed shares. Banks can continuously observe their market price, assess their liquidity and adjust their risk exposure.

OpenAI, however, meets none of these criteria. Despite a valuation of several hundred billion dollars, the company remains privately held. Its price is known only through successive funding rounds or limited secondary-market transactions. Lenders must therefore rely on estimates rather than a liquid and transparent market.

The difficulty is compounded by the company’s unusual structure. OpenAI resembles neither a conventional publicly listed technology company nor a traditional financial asset. Its governance mechanisms, the economic rights granted to investors and the organisation’s specific constraints make valuation more complex for lenders.

SoftBank’s OpenAI stake is therefore probably one of the most coveted assets in the global technology market, but also one of the most difficult to use as collateral for a bank loan.

The IPO does not eliminate uncertainty

The announcement of OpenAI’s IPO filing has nevertheless changed the nature of the discussions. Until now, banks faced an uncertain time horizon. Lenders can now anticipate the emergence of a market price, a public float and the ability to sell shares gradually.

But this does not appear to resolve every concern. Banks must still assess the valuation that will ultimately be set at the IPO, the depth of the market after the listing and the stability of the share price in the months that follow. The excesses of previous technology cycles have left a lasting impression on credit teams.

The experience of WeWork and other overvalued IPOs is a reminder that a company’s private valuation does not always translate into realisable value on public markets. More than a few lenders have learned that lesson the hard way.

Competitive risk is becoming more visible

Lenders’ hesitation also reflects the changing competitive landscape of artificial intelligence. Just two years ago, OpenAI appeared to be the almost undisputed winner of the generative-model race. The picture has since become more complex.

Anthropic has established itself as a credible competitor among large companies. Google continues to invest heavily in Gemini. Meta is accelerating its development efforts. xAI continues to gain ground. Several specialised companies are emerging across vertical markets.

A venture capital investor may view this competition as a driver of innovation. For a lender, however, it primarily represents an additional risk factor.

The future value of SoftBank’s OpenAI stake depends directly on the company’s ability to preserve its technological lead, maintain market share and convert its growth into sustainable cash flows. All of these variables are difficult to project over a decade.

The real issue lies elsewhere

The slowdown in the financing comes at a particularly significant moment for SoftBank. The group must contend with the maturity of a $40 billion bridge loan used to support its OpenAI investments. That debt falls due in March 2027.

The transaction under consideration is therefore not simply an opportunistic refinancing. It forms part of a broader strategy to secure the financial resources required to pursue Masayoshi Son’s ambitions in artificial intelligence.

SoftBank is involved in discussions surrounding Stargate in the United States. The group also plans to invest up to €75 billion in data-centre infrastructure in France. Through Arm, it remains exposed to the entire high-performance computing value chain.

Every new initiative mechanically increases its financing requirements.

Artificial intelligence becomes a balance-sheet industry

Having initially been financed primarily by venture capital, the artificial intelligence industry is entering a new phase in which its capital requirements increasingly resemble those of infrastructure rather than software.

Building a multi-gigawatt data-centre campus requires tens of billions of dollars in investment. Deploying hundreds of thousands of specialised processors, securing access to energy and developing the associated networks now requires amounts comparable to those invested in electricity, telecommunications and transportation.

AI companies are therefore entering a world historically dominated by banks, insurers, infrastructure funds and pension funds.

The next AI battle will be fought in the financial markets

The slowdown in SoftBank’s proposed loan secured against its OpenAI stake does not call into question either the company’s prospects or those of the artificial intelligence market. It nevertheless provides an important signal.

The coming decade could produce a different category of winners from those observed so far. Alongside OpenAI, Anthropic and NVIDIA, the true beneficiaries of the AI revolution could also include the institutions capable of structuring and financing the trillions of dollars in investment required for its deployment.