"Technology is best when it brings people together."

—Matt Mullenweg, CEO of Automattic

Watts Up with AI?

Artificial intelligence may be transforming industries, but it is also reshaping the U.S. electric grid.

You may not even use any Ai at all, but the infrastructure supporting these tools is still affecting your monthly power bill.

Electricity Prices Are Climbing

In May 2025, the average residential rate in the U.S. hit 17.47 cents per kilowatt-hour, a 6.5% increase from the year before.

Some states saw far sharper jumps. For example, Maine reported a 36.3% increase, while electricity prices in Connecticut rose 18.4%, according to the U.S. Energy Information Administration.

The Grid Is Under Pressure

LLMs and AI tools require enormous computational power. These models are hosted in hyperscale data centers, many of which consume hundreds of megawatts of electricity.

Though OpenAI has not released any official data on ChatGPT 5’s energy usage, researchers at the University of Rhode Island estimate that a single response— about the length of a solid paragraph or a customer-facing email—uses roughly 18 watt-hours of electricity. That is enough to power a 60-watt lightbulb for 18 minutes

With ChatGPT processing an estimated 2.5B requests per day, the total energy use adds up to the daily electricity demand of more than a million U.S. homes.

Plus, Microsoft, Meta, and Google have invested billions in more data infrastructure which spells trouble for utilities already racing to keep up.

What Should We Do?

You can’t stop AI data centers from being built. You also can’t fix the grid in a snap. But you can prepare your business for the reality that electricity is getting more expensive and less predictable.

Start by tracking your energy use the same way you track lead volume or labor hours (if you don’t know where the spikes are, you can’t do anything about them). And, consider investing in lighting upgrades, time-based automation, and smart climate control to keep your costs as low as possible.

Ask more from your vendors. If your CRM, marketing stack, or service scheduling tools run on AI, find out how they manage infrastructure. Companies that build for scale but ignore cost and energy efficiency are passing that burden to someone else—usually you.

You do not need to be anti-AI. You do need to be pro-visibility. The more your business understands where energy is going and why, the more control you keep as this next wave of infrastructure rolls in.

Last week, OpenAI rolled out GPT-5, calling it their most powerful model to date. Faster, sharper, and supposedly “more human,” it’s now available to everyone including free-tier users.

The headline improvements? More reasoning power, longer memory, and fewer hallucinations. Think of it less like a model and more like a multitool—with built-in judgment.

One standout feature is with vibe coding. Just describe an app, and GPT-5 builds it. In demos, it created two fully functional language learning tools in seconds from the same prompt. They weren’t flawless, but they worked.

GPT-5 is also better at knowing its limits. It can now recognize when a task can’t be completed and explain why, rather than guessing or stonewalling. Instead of shutting down risky prompts, it gives safe, high-level answers that stay within the guardrails.

So far, reviews are mixed. Some users say the model feels slower, less accurate, and less friendly than GPT-4o. Others are frustrated by message caps and the removal of model switching. But under the hood, this is still OpenAI’s most capable model yet.

We think this meme sums it up pretty well:

ChatGPT might be dressing a little more business formal these days, but that doesn’t mean you can’t still have some fun with it.

This week’s prompt is a creative one just for kicks:

The latest Dealer Playbook hits hard on the reality of AI in retail auto. Our very own Michael Cirillo unpacks why some roles will disappear, and why others will thrive.

From Google’s AI-first search results to the staggering energy buildouts driving next-gen models, the message is clear: learn fast, experiment often, and make AI your competitive edge.

Watch the full episode and start creating the future version of your job before someone else does.

• AI is now part of NADA’s playbook for teaching competitive dealership strategy. 🦾

• The Boston Public Library is teaming up with OpenAI and Harvard Law School to make its collection of historically significant documents more accessible. 📚

• After 34 years of static and screeches, AOL is pulling the plug on its dial-up internet service. 🪦