AI costs

Let’s switch gears and talk about AI costs. The impetus for this topic is a quote I heard in a video:

“…I saw one person who spent $5,000 worth of API calls on a $39 a month program. I think that this problem is across the industry.” Ed Z

Ed Zitron on Prof G Markets podcast on 5/14/2026, “AI Skeptic: This Business Makes No Sense”

That was a stunning fact thrown out there and got me thinking about the profitability of these AI firms, especially in light of their desire to do IPOs (OpenAI, Anthropic, and SpaceX).

Nuance on that quote

Upon doing some research, I learned that there is a nuance on that quote. That $5,000 is most likely a retail price quote, not an AI company expense quote. AI companies themselves are likely to pay $200-$300 a month due to owning the infrastructure. Still, the companies are losing money.

That end user of the $5,000 retail price is most likely to be a software developer, super user or AI agent. It is not a regular Joe consuming compute at that level.

Types of retail pricing

Just to put out there, here’s what I know of the different types of pricing for consuming these AI models:

• Subscription prices: generally $20/month, $200/month, free.
• API calls: a flat price to route the “messages” back and forth
• Token pricing: price on a per token packet, which theoretically sounds more expensive.

Some regular news of AI costs out there

Here’s some news about those costs of using AI:

• Microsoft cancelled the Claude Code licenses after 6 months after encouraging the developers to use it in their work. Apparently, the uptake was so successful that it ended up being more costly. Now everyone is forced to use GitHub CoPilot CLI.
• Uber used up its annual AI budget in four months.
• Nvidia says “cost of compute is far beyond cost of employees.”

Looking toward the future

So, I started to think about what the pricing for regular consumers will be after those companies undergo IPO. In my mind, the cost structure will become visible to everyone, and we will learn how much money those companies are hemorrhaging. Currently, venture capitalists are funding the expenses in hopes of hitting payday once the IPO is done.

If the companies are actually losing money badly, shareholders will demand changes to get to profitability: somehow reduce costs or increase prices.

I’m thinking it is possible that the days of free usage or $20/month subscription are numbered. The only companies that will be able to afford this technology will be the larger ones or the ones raking in money hand over fists. Average user might lose out.

Or, we all get so used to the productive use or the magic of the technology that we will be willing to fork over the money.

What about those using agents?

Here’s where the cost increase could become extremely painful: AI agents consume tokens far beyond what human can consume. They operate 24/7, at super speed, and they can re-do tasks over and over, if need be, to find a solution.

Costs can add up really quickly.

What can one do?

There are 2 ideas I had to wrangle the costs to a reasonable level:

1. Differentiate tasks that can be automated through programming from those that can’t due to a fuzzy nature. In finance, a lot of the work is of a deterministic nature so I would rather use a program to do the work. I would use AI to aid me in developing programs to automate tasks. For fuzzy stuff, such as scenario planning or customer sentiments, I would use AI.
2. Avoid using agents until you understand what kind of costs you will be incurring. Use AI only for high value stuff or for things humans absolutely cannot do. For example. AI will be far more capable of sifting through a large database than a human with a spreadsheet. Then, once you understand the costs and can track it, you may be able to decide which activities would be worth bringing in AI.

Closing

So, the AI revolution could come to a screeching halt, temporarily, as shareholders force AI companies to improve profitability. The pricing structure might very well change, causing extreme pain.

I don’t know when that day will come but it feels like it will be this year.