The rise of ChatGPT late last year sparked a surge of interest in new generative artificial intelligence technologies, and nowadays almost every big software provider has jumped on the bandwagon.

Yet despite the massive enthusiasm for generative AI from technology providers and customers alike, no one has figured out yet how to make money from these powerful new products. On the contrary, companies such as Microsoft Corp., Google LLC and OpenAI LP are all believed to be losing enormous amounts of money by offering generative AI capabilities to their users, as the Wall Street Journal reported today.

For instance, one of the earliest generative AI products to launch was GitHub Copilot, a service that’s used by programmers to create, fix and translate software code. Microsoft, which owns GitHub, has claimed it has more than 1.5 million users and writes around half of the code they generate.

But even with its widespread adoption, it has become a huge money pit, one person familiar with the numbers told the Journal. GitHub charges users $10 per month to use Copilot but loses about $20 per month per customer on average.

Generative AI is an expensive technology because the models can take years to train and fine-tune, and even after that’s done, they require enormous resources  just to run on a day-to-day basis. “They require massive compute power,” Jean-Manuel Izaret, head of the marketing sales and pricing practice at Boston Consulting Group, told the Journal. “They require massive intelligence.”

There is also an element of overkill in many use cases. For instance, ChatGPT is powered by OpenAI LP’s GPT-4 model, which is said to be one of the world’s most powerful models of all. Yet many ChatGPT enterprise subscribers use it to perform extremely limited tasks. As the Journal put it, using GPT-4 to summarize an email is like delivering a pizza in a Lamborghini.

Higher prices, cheaper models

To stop bleeding money, many companies are looking to develop less powerful models to perform simpler business tasks, while others are simply planning to raise their costs. Microsoft, for example, is going to charge an additional fee of about $30 per month for an AI-infused version of its Office 365 software suite. Currently, the cheapest version of Office 365 is around $10 per month. The AI features will be able to compose emails and PowerPoint presentations, generate Excel spreadsheets automatically and perform many other tasks.

Similarly, Google is planning to ask users to cough up $30 a month for the generative AI features within its own productivity software, where its lowest subscription tier currently costs just $6.

At the same time, reports elsewhere have stated Microsoft is looking to develop less powerful models to eliminate the overkill problem. For example, it’s building smaller, inexpensive AI models destined for Bing, which will be dedicated to web search only. Some of those models will reportedly be based on open-source AI from companies such as Meta Platforms Inc.

Adobe Inc. is going with a different tactic. It has created a system of credits for its AI image generation tool Firefly. Should a customer use up its allotted monthly credits, the Firefly service will be slowed down, in what the company said is a way to discourage overuse.
“We are trying to provide great value but also protect ourselves on the cost side,” Adobe CEO Shantanu Narayen told the Journal.

Customers and investors still have questions

The challenge Microsoft, Google and others who are thinking about charging more for their AI services is that they must perform a delicate balancing act, as not everyone believes the software is worth paying for. “A lot of the customers I’ve talked to are unhappy about the cost that they are seeing for running some of these models,” said Amazon Web Services Inc.’s CEO Adam Selipsky.

For now, the issues of monetizing generative AI have not dissuaded investors, who have collectively thrown billions of dollars at the most promising startups this year. The Journal said OpenAI is now discussing the possibility of a share sale that would take its value to more than $90 billion, three times what it was worth at the beginning of the year.

However, many in the industry believe that it’s only a matter of time before investor enthusiasm wanes. When that happens, a lot of people are going to take a long, hard look at the costs of running AI and ways of using the technology profitably.

“Next year, I think, is the year that the slush fund for generative AI goes away,” said May Habib, CEO of Writer Inc., which raised $100 million in funding just last month.

Image: vecstock/Freepik