AI’s path to profits is being driven by consumers

MARKETS naturally see through the lens of businesses. When tech stocks took a dive last month on concerns of an “AI winter”, investors were egged on by a study showing 95 per cent of corporate artificial intelligence (AI) pilot programmes failed to deliver gains in productivity or profit, making all this expensive AI start to look a little useless.

Consumers would beg to differ.

While businesses grapple with how best to plug generative AI tools into their systems – as they’ve naturally done with every other tech wave in history, from personal computers to smartphones to social media – individuals have been embracing the technology.

The phenomenon is easy to overlook when you only measure success with quantifiable metrics like time and money, and when the cost of running data centres is still so high. But some of AI’s biggest winners so far are companies who are chasing squishier value propositions such as entertainment and camaraderie. 

A study by Harvard Business Review this year found that the three most popular use cases for generative AI were therapy and companionship, organising life and “finding purpose”.

Character.ai’s roughly 20 million monthly active users are on its platform for role-playing with AI-generated characters. It is on the cusp of breaking even on the cost of running its AI models – its so-called “inference costs” – which is an impressive milestone for any AI startup.

A year ago, when the company was still being led by its founder, former Google researcher Noam Shazeer, a large proportion of its capital spending was on the cost of renting data centres and AI chips, in large part because Shazeer had grand ambitions to build AI models that were smarter than humans – so-called artificial general intelligence.

But he returned to Google this year for an eye-watering US$2.7 billion, and under new management, Character.ai has focused on becoming an entertainment platform. Instead of only building its own AI models, it uses those of other companies like OpenAI. Its compute costs have now halved.

Revenue from its US$9.99-a-month subscriptions has grown to nearly match the US$4 million it spends each month on inference, or the cost of generating AI content for its users. The app currently shows small banner ads, but will likely introduce advertising in a more sophisticated way in the next year or so, turning on a potentially enormous revenue spigot.

OpenAI will eventually do the same.

In May this year, it hired Fidji Simo, an executive who helped to transform the delivery app Instacart into an advertising powerhouse.

Last week, the company announced it was spending US$1.1 billion on Statsig, a product-analytics company that helps consumer apps such as Notion, SoundCloud and Linktree test new features, helping them decide which to keep or ditch. Expect OpenAI to use that new testing ability to introduce advertising to its platform.

OpenAI’s purchase of Statsig reminds me of Facebook’s 2013 acquisition of Onavo, a mobile-data analytics company that enabled the gathering of intelligence on mobile-usage trends and potential competitors.

It was this that ultimately steered Mark Zuckerberg toward his US$19 billion purchase of WhatsApp. Statsig could also end up working quietly behind the scenes to guide OpenAI towards monetising its all-important consumer base.

Despite OpenAI’s enterprise ambitions, it is ultimately a consumer business.

In less than three years, ChatGPT has amassed more than 700 million weekly users; some 99 per cent of them are individual consumers using free and paid versions of ChatGPT. Roughly five million are business or enterprise customers of the company, which says it will make US$13 billion in revenue over the next year.

For all the talk of corporate disappointment with AI, the most important story may be happening at the individual level. Millions of consumers are already spending their own time and money on tools that entertain, organise and even console them.

This shift points to an industry in which monetisation will come increasingly from subscriptions and advertising – just as it did with gaming companies and social media businesses – and where the firms best positioned to thrive are those willing to build for feelings as much as for efficiency, and who can capitalise on the model infrastructure being built by the likes of OpenAI and Google.

If Wall Street still sees only an “AI winter”, it is missing the spring that is already underway in personal AI.