OpenAI has secured its long-awaited and contentious restructuring, clearing the way for an eventual public float of the $US500 billion ($760 billion) leader of the scramble to develop and cash in on artificial intelligence and an eventual parting of the ways with long-time backer, Microsoft.

Carving what was a not-for-profit structure into two – a continuing not-for-profit that would have a 26 per cent shareholding in a for-profit commercial entity – will create more funding options for OpenAI, which has an insatiable appetite for capital.

It will also free it from the shackles of its partnership with Microsoft, which invested in the start-up in 2019 and which has turned a $US13.75 billion investment over time into a $US135 billion shareholding in the world’s hottest AI company.

OpenAI’s Sam Altman has spent the best part of 18 months working on a restructuring that would achieve both those ambitions.

With financial commitments to buy computing power from third parties that now amounts to more than $US1 trillion, OpenAI’s need for funding has far outstripped Microsoft’s capacity to provide it.

Moreover, Microsoft has its own AI ambitions, helped by a deal with OpenAI that gave it exclusive access to the group’s intellectual property and products, but which made it a competitor and generated tensions and conflicts of interest.

Under the deal struck on Tuesday, Microsoft will retain those exclusive rights until 2032, or until an independent panel determines that OpenAI’s technology has achieved artificial general intelligence (AGI), or human levels of intelligence.

Previously, Microsoft didn’t own equity in OpenAI but had a revenue share and a profit share capped at 100 times its investment. It will continue to be entitled to 20 per cent of OpenAI’s revenues until (or if) the company achieves AGI, but has also struck a deal under which OpenAI will buy an extra $US250 billion of cloud computing capacity from Microsoft’s Azure platform.

In effect, Microsoft has eschewed the potentially massive pay-off if OpenAI can achieve AGI in return for what private equity firms would call a “10-bagger” return on its investment and a $US50 billion or so new revenue stream from selling more cloud services to OpenAI. Its share price rose 2 per cent in response, taking its market capitalisation above $US4 trillion.

OpenAI creates a structure that enables it to give existing investors more conventional returns on their investments, more easily and cheaply raise new debt and equity and introduce competition to Microsoft for cloud capacity and chips, lowering its own costs.

If it were to launch an initial public offering, it would, on the valuations provided by its most recent equity raisings, be valued at more than $US500 billion.

Moving to a more conventional governance model isn’t without its challenges, not the least of which is Elon Musk (a co-founder of OpenAI and, via his xAI, a competitor). Musk has taken legal action that accuses OpenAI of breaching its contractual duties by incorporating a for-profit entity within what was supposed to be a not-for-profit structure.

The board of the not-for-profit that will hold 26 per cent of OpenAI Group PBC will also provide all its directors, effectively enabling the OpenAI Foundation to maintain control.

It was the launch of OpenAI’s ChatGPT chatbot that ignited the boom in AI investment that has underpinned the surge in the US sharemarket, and which has made OpenAI, with Nvidia, one of the key pillars of that boom.

The challenge for OpenAI is that its ambitions are so extreme that Microsoft could never fund them. Indeed, it needs a “whole-of-sector” funding support to try to realise them.

It has raised conventional equity and debt but, along with Nvidia, has been heavily involved in some very innovative financing structures where Nvidia and others have agreed to supply it with chips and cloud capacity – and effectively provided the funding for the purchases.

OpenAI has committed, for instance, to buying $US100 billion of Nvidia’s chips, in stages. In return, Nvidia has agreed to invest $US100 billion in OpenAI, in stages.

It’s those sorts of deals that have generated references to “circular financing”, “financial engineering” and “bubbles”, and parallels have been drawn to the vendor financing that was a feature of the dotcom and telco boom of the late 1990s, and its bust.

With OpenAI committing to buy about 26 gigawatts of computing power and each gigawatt estimated to cost about $US50 billion, it has no choice but to be creative in its fundraising.

Nvidia, as the major supplier of the chips vital to AI, is at the epicentre of the AI boom. But unlike most of the other participants, it is exceptionally profitable even as it becomes more dependent on the ability of OpenAI and its aspiring rivals because of the tangled web of financial exposures it has been creating to help them finance their demand for its chips, which represent about 70 per cent of the cost of their computing power.

Nvidia’s Jensen Huang said on Tuesday that he didn’t believe there was a bubble in AI, saying that his company was using many AI models and services and was happy to pay for them. There is an increasing level of interdependence, however, between the chipmakers, cloud providers and the AI start-ups.

There’s also an increasing level of debt. An element of OpenAI’s desire to shift to a for-profit structure has been its ability to raise more debt, more cheaply. OpenAI isn’t an investment-grade company at this point.

The challenge for OpenAI is that its ambitions are so extreme that Microsoft could never fund them, indeed it needs a “whole-of-sector” funding support to try to realise them.

OpenAI is investing, and committing to invest, far more than the revenue it is generating, and despite being on target to generate about $US13 billion of revenue this year, it is burning cash and increasing its losses. This makes it completely reliant on raising new capital to continue. Other AI aspirants are in the same boat.

Altman has said profits aren’t among his top-10 concerns, although OpenAI has been adding services constantly this year to boost its revenue, suggesting the payoff for investors would come as the technology evolves.

He has positioned OpenAI as a web browser, in competition with Google for search, as a software company competing with the likes of Microsoft and is tinkering with the launch of smart devices – all of which are established and very profitable sectors that OpenAI’s technologies could steal share from – although the sheer scale of investment required to get OpenAI to any sort of maturity and proximity to profitability, if it can get to that point, is mind-boggling.

The Business Briefing newsletter delivers major stories, exclusive coverage and expert opinion. Sign up to get it every weekday morning.