OpenAI raises 110 billion, a megatower in a closed economy, and without Microsoft

OpenAI has just completed a financing round of up to $110 billion, at a valuation of $730 billion. The operation goes beyond the standards of venture capital and is more akin to an industrial structuring prior to an IPO planned for the end of the year.

However, one element catches our attention: Microsoft, historical partner and main individual shareholder, is not participating in this megatour.

An absence that reconfigures the balance

Since 2019, Microsoft has been OpenAI’s central strategic ally. Azure hosts most of its infrastructure, while the publisher benefits from an exclusive license on the models integrated into its products. This relationship, long perceived as almost organic, is not legally called into question and OpenAI specifies that the agreement concluded with Amazon affects neither the licenses nor Microsoft’s exclusive access to its intellectual property.

But the absence of Satya Nadella’s group in this round modifies the perception of the balance of power. OpenAI is no longer part of a bilateral logic, and now organizes a multi-partner ecosystem, where Nvidia, SoftBank and Amazon take a decisive place.

OpenAI seeks to reduce its structural dependence on a single cloud provider, while Microsoft, already heavily exposed, seems to consider that its strategic positioning is consolidated without additional participation.

Industrial investors rather than funds

Nvidia and SoftBank each commit up to $30 billion. Amazon is investing 15 billion immediately, with an additional 35 billion conditional either on the IPO or on reaching a major technological threshold. To these amounts are added around 10 billion from sovereign funds and institutional investors.

The majority of contributions come from industrial players, and not from traditional venture capital funds. It is the infrastructure providers who now finance the development of the models.

The heart of the system: a 100 billion contract

And the most structuring aspect does not lie in the valuation but in the contractual commitments. OpenAI plans to spend an additional $100 billion on cloud services and chips from Amazon over eight years, complementing an existing $38 billion contract. At the same time, the company maintains massive commitments to purchase computing power, now valued at around $600 billion by 2030.

In other words, a substantial part of the funds raised returns to investors in the form of infrastructure spending.

This architecture creates a circular mechanism:

• the injected capital finances the acquisition of computing capabilities
• these capabilities enable training and deployment of models
• revenues generated support new infrastructure investments

Financing and production become interdependent.

From software startup to infrastructure operator

The numbers give the measure of the transformation. OpenAI had about $13 billion in revenue last year. Forecasts call for 30 billion this year, then more than 60 billion in 2027. The group does not, however, plan to achieve positive free cash flow before 2030. With nearly 40 billion dollars of cash on the balance sheet after the operation, the company provides itself with a substantial financial cushion to absorb prolonged losses.

But its economic profile is changing. The cost structure is dominated by multi-year computing commitments, the construction and operation of data centers, and the massive purchase of semiconductors.

OpenAI is closer to a digital infrastructure operator than a traditional software publisher.

A closed, capital-intensive economy

This model has a unique characteristic: investors are also the main suppliers. Nvidia secures demand for its GPUs. Amazon guarantees the occupancy of its cloud capacities. SoftBank is positioning itself on an infrastructure capable of irrigating a large ecosystem.

Generative AI is thus becoming a sector where capital and computing are inseparable. Access to models depends on access to chips; access to chips depends on funding.

This pattern promotes concentration. Players capable of mobilizing hundreds of billions of dollars have a structural advantage over entrants (see Richard Menneveux’s editorial on this subject, AI will enrich a minority and ruin everything else)

Strategic repositioning

On the commercial side, OpenAI says it wants to increase the share of its revenues from businesses to 50% by the end of the year, compared to around 40% today. The B2B market offers higher contract visibility and more stable monetization than the consumer segment.

This growth must take place in a lively competitive context. Google develops its own models within an integrated infrastructure. Anthropic has positioned itself primarily on professional uses. xAI continues a parallel trajectory.

The ability to secure computing over time constitutes a differentiating factor as determining as the quality of the models themselves.

A model that only holds if…

OpenAI’s circular financing only holds if revenue growth sustainably absorbs capital intensity. As long as demand grows faster than infrastructure commitments, the loop works: capital raised, compute purchased, models deployed, revenue reinvested.

To hold up over time, the model assumes four conditions:

• sufficiently predictable recurring and contractual revenue
• a technological advantage to preserve margins
• a continuous improvement in the marginal cost of computing
• diversification of partners in order to limit critical dependencies.

It becomes fragile if this balance is disrupted. AI is thus entering a phase where algorithmic performance is no longer enough, and where financial and industrial robustness becomes the decisive variable.