Every time I see Google or Microsoft in particular roll out one of their AI features on a mass scale, like Microsoft has recently done for Windows - integrating Bing Chat into Windows 11, with a dedicated button on the task bar - the first question that comes to my mind is how much they're losing on this.
Running LLMs and other 'AI' or 'machine learning' infrastructure is incredibly expensive at the moment, OpenAI's own API isn't that cheap for GPT, and I'm guessing Microsoft are mostly throwing their own Azure capacity at the problem until either it becomes cheap enough that they're no longer losing so much money - or they've successfully built their moat - and they can start introducing subscription fees or increasing them.
The WSJ article highlights some of these issues, with suggestions that Github Copilot is was costing Microsoft at least $20 per user, per month. A loss of at least $10 per person.
Individuals pay $10 a month for the AI assistant. In the first few months of this year, the company was losing on average more than $20 a month per user, according to a person familiar with the figures, who said some users were costing the company as much as $80 a month.
With heavy users going as high as $80. This isn't confirmed, but it also wouldn't surprise me.
It'll be interesting to see how the price changes on Github copilot. Many software engineers at the company I work for use Github copilot every day, it's a fairly seamless part of their workflow already. I think many engineers may be willing to tolerate even a 100% price increase if it came to it.
I don't write as much code these days, but it has been quite useful to me when trying to pick up and contribute in languages I'm either new at or a little rusty in, so as much as I'd obviously like it to remain at $10 - unfortunately I believe Microsoft/Github are likely to hike this price up over the next 12 months - and many of us will pay up.
This is without even touching on organisation/enterprise level agreements, in which Microsoft are in a pretty good position to start squeezing companies who signed up for lower priced AI services.
Update - Nat Friedman, former CEO of Github, refutes these claims