We have all been reading and watching the discussion around the AI bubble. The day is coming when the world stops being impressed by what AI can do and starts asking what return it actually delivers.

As someone who has spent the last year praising the capabilities of AI, I think it is only fair to talk about the other side of the equation as well.

The costs are rising.

This week was the first time in at least a year that I hit a token threshold and was told I had used my allotted capacity. I have been building like crazy, so part of me felt like I should have expected it. What I discovered, however, was both a gift and a curse.

I had hit my limit.

Inline image

But I was only paying half of what I thought I was paying.

For the last year, I thought I was on the $200 plan. Turns out I was on the $100 plan. Upgrading to the plan I thought I already had, for roughly five times the capacity, was a no-brainer.

Then I stopped and thought:

I ran out. Really?

So I started researching. Was it me, or had something changed?

The answer was both.

A lot has changed since I signed up a year ago. AI usage has exploded. Demand has exploded. Infrastructure has struggled to keep pace. Meanwhile, tariffs, wars, higher interest rates, and investor pressure have shifted the conversation from growth at all costs to one simple question:

Where is the ROI?

The same companies that were racing to build AI infrastructure are now being asked to justify every dollar spent. The same enterprises that bet heavily on AI to improve efficiency are now trying to quantify cost versus reward.

Is AI really cheaper than humans for every task?

Increasingly, many organizations are finding the answer is more complicated than they expected.

That led me to a conclusion.

The low-cost, high-return AI environment that small businesses enjoy today may not last forever.

At some point, the bill comes due.

Not because AI doesn't work. It clearly does.

Not because the technology is overhyped. I use it every day and know firsthand how powerful it can be.

But because computing power, energy, infrastructure, and capital all have costs. If we don't see major breakthroughs in efficiency and energy production, AI could become the great neutralizer—another technology that slowly becomes concentrated among those with the money, power, and influence to afford it at scale.

So what should small businesses do?

Prepare now.

Take advantage of today's pricing while it exists. Build systems, workflows, and data platforms that create long-term value beyond the AI itself. Focus on owning your processes, your data, and your institutional knowledge.

Most importantly, think beyond the AI tool and focus on the AI strategy.

My bet is simple: the businesses that win won't necessarily be the ones spending the most on AI. They'll be the ones that use today's affordable access to build assets they own tomorrow.

When free becomes paid, and efficiency comes with a price tag, those businesses will already be ready.