If OpenAI is going to float this year, it has to get serious about its business model. The wow factor around the US company – the poster child of an AI industry boom that has stoked fears of a stock market bubble – has been long established, but when will the profits come? The party can’t go on for ever.

The developer of ChatGPT is one of the biggest startups in the world and is now valued at $850bn (£645bn). Meanwhile, it is reportedly spending $600bn on infrastructure (the amount it invests in datacentres and chips to power its AI models) by 2030. At least this is a reduction on an initial estimate of $1.4tn.

Despite the slimmed-down spending plans, the startup is nowhere near being profitable. In fact, if things stay as they are, it will burn through half a trillion dollars by the end of the decade. Boosters may point out that Uber, for example, spent billions before turning a profit – but that was $30bn, not $600bn.

OpenAI, led by Sam Altman, its chief executive, appears to be making decisions fast, as a market reckoning of sorts approaches with a mooted flotation towards the end of this year. Three areas of its business have been jettisoned in the past month; one more has been proven to offer lacklustre promise at best.

In early March, OpenAI pulled back from Instant Checkout, a plan in which consumers would shop for goods directly inside ChatGPT. This was after a five-month trial in which the company appears to have found that building a successful commerce platform is harder than it looks. “Like many of OpenAI’s initial launches, it felt more like a public demo of what the tech could do than a very sustained effort to set up a commerce business,” said Niamh Burns, an analyst at Enders.

Then, last week, it ditched Sora, its video-generation platform, and with it a $1bn deal in which Disney was going to license OpenAI-generated content to “unlock new possibilities in imaginative storytelling”. This was strategic for OpenAI, because Sora was a money pit. It was awkward for Disney, which reportedly learned that the platform would be axed an hour before the public did.

Finally, last week, it also pulled the plug on erotic chatbots, a repeatedly delayed plan announced last year to “treat adult users like adults” and let them have sexy conversations with ChatGPT. “This would have been a ridiculously risky launch,” said Burns, especially with mounting scrutiny around online safety. “It would have been a complete nightmare from both a product safety and PR perspective.”

Optimistically, this all represents a company trimming the fat before an initial public offering (IPO), in a competitive market where Anthropic, the maker of the Claude chatbot, appears to be winning more and more loyalists among business customers. OpenAI “is under serious pressure to show strategic discipline”, said Burns. “It has cast the net too wide.”

Adrian Cox, a managing director at Deutsche Bank Research Institute, said OpenAI was making the right moves if, as reported, it was gearing up for a flotation valuing the business at $1tn. This compares with its annualised revenue – a projection calculation based on its short-term performance – of $25bn, which the company reportedly attained in early March.

“If OpenAI is moving to an IPO and seeking a wider pool of investors, those investors are going to want to see real evidence of strong, sustainable revenue growth over the years to come,” Cox said. “By focusing its business model in this way, OpenAI is probably aiming for that growth in the best way possible.”

He added that OpenAI appeared to have stopped battling rivals with an “everything” business model and was now narrowing its focus.

“There had been concern about the lack of obvious ways to monetise what is, by far and away, the leading consumer AI brand,” Cox said. “It now appears to be making hard choices that allow it to better monetise its business in the future. Many investors may say this is the best news they’ve heard from OpenAI in months.”

And the signature product of OpenAI, indeed of the entire AI boom, remains popular. ChatGPT now has more than 900 million weekly active users and more than 50 million paying subscribers. OpenAI makes its revenues from these subscriptions – which account for 75% of its income – and offering businesses its corporate versions of ChatGPT, while allowing companies and startups to build their own products with its AI models.

But there’s a feeling among analysts that it could have found rigour earlier, especially as it burns through billions of dollars every month on experiments that end up being little more than that. A Forbes columnist labelled OpenAI “the most distracted company in technology” after Instant Checkout fell through.

Burns said: “We have seen so many consumer product launches, promising to disrupt the browser, online commerce, content creation, search … Actually focusing your strategy and executing on a product that people will want to use and, crucially, be willing to pay for in some real form, is the harder challenge.”

Last week, OpenAI announced what appeared to be a win amid the chaos: a trial of advertising in ChatGPT made $100m in annualised revenue, which means it made about $12m in six weeks. Perhaps this is a route to profitability; ChatGPT, after all, knows a great deal about its users and can presumably uniquely target ads.

Even that, as with all the other things the company had trialled, would probably take a lot more effort to get right, said Burns. “It could very quickly begin to feel creepy and risk a user backlash and privacy concerns.”

On the other hand, ads in ChatGPT won’t drive much business if they remain just “a glorified banner ad below the answers” without targeting, she said.

Nikhil Lai, an analyst at Forrester, said the ad trial went “better than expected”, but this did not mean OpenAI was anywhere close to being able to monetise advertising.

Lai said it would probably be a “couple years before OpenAI can get there, if they ever get there”, adding: “They’d have to do a lot and they’d have to change a lot.”

The maker of the world’s most-hyped technology has to find a way to turn a profit from it and limit an unsustainable cash burn. Investors await the answer.

An OpenAI spokesperson said that the infrastructure to run AI, or “compute”, was in short supply so it was prioritising investments.

“With user demand outstripping supply, compute is the critical resource when it comes to AI,” the spokesperson said. “Along with locking in our long-term compute needs via our infrastructure strategy, we are also ruthlessly prioritising the allocation of that compute across where it drives the most long-term economic value: advancing frontier research, growing our 900m-plus global base of users, and powering enterprise use cases.

“As we continue to secure more and more large-scale compute, this disciplined focus on where we apply that compute allows us to grow, innovate faster and deliver more efficiently to enterprises and developers.”