The Cost of AI: When the Light Switches On

I recently listened to a fascinating conversation on a tech podcast discussing the “AI bubble,” and it pushed me to think more critically about where AI truly is right now and where we may actually be headed. To be clear, I have listened to a variety of opinions related to AI: what it is and what it is not. However, this conversation cut through to me.

One thing that stood out was the way the conversation illustrated the AI ecosystem into three major segments:

• AI labs where AI models are continuously trained,

• AI startups fueled by venture capital, and

• AI infrastructure, where we see the commodification of data centers, land, and energy.

There was also a mention as to whether AI will be “the next electricity.” Personally, I’m not completely convinced. Not because I think AI is fake or useless, but because there is more to be discussed when it comes to AI.

In specific use cases, AI can be an incredibly powerful tool. Right now, most people experience AI as an enhancement integrated into existing systems:

• smarter search,

• workflow automation,

• coding assistance,

• recommendation engines,

• summarization, and

• customer support

To me, AI currently feels more like an extension of the internet and modern software than something entirely separate from it.

The “All-You-Can-Eat” Phase of AI

One of the more interesting points from the podcast was the idea that we are in an “all-you-can-eat” phase of AI. Right now, organizations are deploying AI systems at relatively low cost compared to the enormous infrastructure required to support them.

But eventually, venture capital will let go, and things will become real. At some point, the switch will flip.

How does venture capital work? Well, pretend you and your friends want to open the best-ever coffee shop. You have the recipes and the vision, but you do not have enough funds to buy state-of-the-art espresso machines, renovate the space, or market your coffee shop. So, you seek out a venture capitalist, a wealthy investor who gives you money to help grow the coffee shop quickly.

In exchange for their investment, you give them a percentage of ownership in the business. But they do not actually want to make coffee themselves. They are betting that your coffee shop could eventually grow into a huge brand with regional or global locations. If that happens, their ownership becomes worth far more in the future because another company could buy your coffee shop, or it could become large enough to go public one day. Getting back to real life, many AI companies are still being backed by venture capital.

And when that happens, we will see the true cost of AI. Maybe the long-term cost becomes minimal because the technology scales efficiently and becomes widely distributed. Or maybe the real costs are much larger than people currently realize.

And I don’t just mean financially.

The Hidden Costs of AI

The current phase of this AI boom is tied to a race for infrastructure:

• data centers,

• land acquisition,

• energy consumption,

• water usage,

• chip manufacturing, and

• large-scale compute expansion.

But what is the environmental cost to be paid for this venture? If the future of AI requires endless expansion of energy-intensive infrastructure, then environmental impact cannot be treated as an afterthought. And beyond environmental concerns, there are also economic and social costs:

• labor disruption,

• shifting productivity expectations,

• bias embedded in systems, and

• lack of regulation and standards.

That uncertainty matters.

At the End of the Day

My hope is that we center people and environmental systems, and not artificial 'minds' taking over humanity. Our real challenge may not be whether AI becomes “the next electricity.” It may be whether our society can thoughtfully balance the benefits of AI against the environmental, economic, and social costs that come with building it.