The AI value proposition ain’t what it used to be

Pour one out-preferably a 15 year cask-aged single malt-for the CEO homies that were itching to reduce their head count only to find out it’s prohibitively expensive when AI hyper scalers charge full freight for second rate services. Sorry, overwhelmingly, boys. Looks like your stuck with a carbon based workforce for the time being.

The AI industry leaders like OpenAI and Anthropic have raised fabulous amounts of money over the past five years, by some accounts over a trillion dollars and counting. They also enjoy historic burn rates that see that cash treated to a Viking funeral the split second it comes in the door, so that begs the question, where does all the money go?

AI companies are presumably running lean in terms of human staff, so it’s not employee overhead. The vast majority of capital expenditure or capex goes to data centers, R&D and the purposefully nebulous category of “compute” which swallows big, wet chunks of cash like a blue whale going through a bait ball. For a few years the strategy of feeding a constant torrent of legal tender into the AI woodchipper was great for end users because the AI service providers and their mountains of venture capital were essentially subsidizing the true cost of the service. The same thing happened with services like Uber. Your ride from LaGuardia never cost the eight dollars you paid. It always cost five times that, but or the first half decade venture capital paid the difference. Once riders were asked to pay full cost usage dropped and the company pivoted to delivery. Now that the AI bubble has become visible from space many early AI investors are anxious to get their money out, hence the rush to IPO, revenue growth and the corresponding huge increases in cost to consumers for tokens, compute, data and the rest.

The end result is a product that doesn’t do half the stuff it claims to do for the bargain price of nine times what you used to pay. There’s no shortage of companies that went hard on early AI adoption urging, and even incentivizing, staff to gobble up tokens like Ms. Pac-Man that are now limiting usage as the gap between price and cost is snapping shut. C-suite executives had images of empty cubicles staffed by AI agents working diligently to create enough shareholder value to fuel a never ending, upward spiraling executive compensation package, but alas, they can’t clean out the break room fridge just yet, or maybe ever.