REUTERS, London

If there’s one thing Sam Altman knows, it’s how to grab attention. The boss of ChatGPT developer OpenAI can be seen on viral clips in which his likeness, generated by the company’s buzzy new Sora video model, steals computer chips. He regularly makes waves with announcements on X, which is owned by his arch-rival Elon Musk. He’s even unveiled grand investment plans at the White House. He’s good at keeping attention, too, with the company boasting 800 million weekly active users and a billion on the horizon. The real test for Altman and his competitors, though, is how well they can win over a more skeptical audience: big corporations.

Right now, OpenAI makes most of its money from regular people. The company’s annualized rate of revenue has reached $13 billion, and may hit $20 billion by the end of this year. About 70 percent of that comes from consumers paying for ChatGPT subscriptions. Over 5 percent of the chatbot’s users pay, though that’s just enough to cover the immediate cost of the computing power generating responses for everyone, if not OpenAI’s other enormous expenses.

The remaining 30 percent of revenue comes from enterprise offerings. Some of that is via usage-based pricing for accessing OpenAI’s large language models directly through what is known as an application programming interface (API). Customers are billed per unit of input or output, known as a token. Prices for these tokens start high when a cutting-edge model is new, but quickly go into free-fall: between GPT-4’s initial release and the refined GPT-4o Mini model released 18 months later, costs fell by 99 percent.

It’s therefore a constant war of attrition to keep big companies hooked on the latest and greatest — and most expensive — models. One way to get a little more stability, though, is to sign enterprise clients like Morgan Stanley and T-Mobile, to bigger, more comprehensive contracts. After all, they can be willing to do so when OpenAI can demonstrate a clear return on investment, say by helping a call center resolve tickets more quickly, or by speeding up programmers’ code reviews.

If they can be convinced, corporations are a gigantic prize to play for. Gartner forecasts that enterprise software spending will reach $2 trillion globally by 2029, of which a mere $76 billion will go to generative AI. If chatbots really are as disruptive as their boosters suggest, they could snare far more of that budget.

To boot, consumers are fickle, switching constantly between new and shiny products. Look, for instance, at Alphabet-developed Gemini, which has doubled market share to 14 percent over the last year as ChatGPT’s share has slipped, according to Similarweb data. Corporate buyers can be much stickier. Once a system is integrated and covered by a multimillion-dollar contract, switching becomes a hassle, making every enterprise deal more lucrative over time.

Consumer ubiquity can help pry open enterprise wallets. Engineers discovering ChatGPT through personal use may push for adoption of the familiar tool inside their organizations. When companies eventually sign up, they pay for added security, compliance features, and administrative functions, often at much higher price points.

That cost — and the dependence on OpenAI’s computing resources, which Altman frequently points out are overstretched — can become a sticking point, though.