> reducing demand by 1GW is not the same as generating an extra 1GW.

Reducing demand is better, because you don't even need to pay the cost of generation.

We have this in California. PG&E attached a device to my central AC to shut it off during peak demand, and I get a discount in return.

Where I am we had a company called Pooled Energy they tried to do the same thing with swimming pools they went bust fairly quickly

https://www.aer.gov.au/news/articles/communications/aer-ensures-continued-supply-former-pooled-energy-and-weston-energy-customers

AI means waste 1GW of power in some big ass datacenter 24/365 in order to try to gain 1GW 20% of the time.

Yeah. Nope.

It does not need AI to modify slightly the temperature of thermostats to optimize energy consumption.

Regular prediction and interconnection will do.

And things like heating and cooling have peak times for a reason. Being told "everyone else is using their heating so you have to stay cold" won't go down very well. All you can do is defer things which are not time critical, like the laundry or dish washing etc.

> This is really just about how they want to own and control your thermostat instead of you.

The easiest way to make sure all my heating and cooling decisions are made in my families best interest is to make them myself. That way there is absolutely no confusion.

Software has been renamed "AI". Didn't you get the memo?

I get where you're coming from - it can feel like utilities are just grabbing control of your thermostat when the grid gets overloaded. That’s essentially what traditional Demand Response (DR) is: when there’s a risk of a brownout or blackout, the utility sends out a signal to participating devices (like your thermostat) to cut back on power usage. It’s a straightforward mechanism to prevent the grid from failing, and it’s valuable enough that it’s traded on the energy markets, both for the promise to reduce demand if needed and for the actual reduction itself.

But this new approach goes beyond just nudging the thermostat up when demand peaks. It’s about leveraging predictive algorithms and machine learning to intelligently manage energy consumption. Instead of a simple override, the system can now learn the thermal profile of each home. By monitoring how fast the house heats or cools, considering outside temperature, insulation, and the efficiency of the HVAC system, it builds a unique thermal model for your home. It’s like the system gets to know your house’s behavior, including how long it can coast without needing active heating or cooling.

Armed with this information, the system can make smarter decisions. Instead of just cranking up the thermostat when things get tight, it might pre-cool your home when energy is cheap or demand is low. By doing this, it takes advantage of the house’s thermal inertia, letting the temperature drift up naturally when demand spikes later in the day. It’s not simply raising the set point; it’s anticipating the demand curve and optimizing your energy use in advance. This way, you stay comfortable while the system reduces the overall load during peak hours.

This isn’t a one-size-fits-all approach either. The machine learning aspect comes into play as the system continuously learns from new data. It tracks how your home responds to changes, how quickly temperatures shift, and even how you react to different temperature settings. Over time, it adjusts its strategy based on this feedback, getting better at finding the sweet spot between energy savings and comfort. It’s a dynamic process that adapts to your preferences, rather than imposing a blanket rule.

It’s also worth mentioning that modern DR programs are typically opt-in, with incentives for participation. You get to choose whether you want to take part, and in exchange, you often receive rebates or lower rates. It’s less about taking away control and more about offering a smarter, more flexible way to manage energy that benefits both you and the grid. If you’re open to the idea, these systems can help save money, reduce strain on the infrastructure, and even prevent outages.

So, while the term "AI" might seem like buzzword marketing, there’s actually substance behind it. This isn’t just a utility flipping a switch on your thermostat. It’s about using predictive algorithms and adaptive learning to orchestrate energy use in a way that balances comfort, cost, and grid stability. It’s a step up from the old DR methods, aiming to optimize rather than simply cut back.

Source: I design DR systems for smart thermostats

We clearly need an AI capable of ordering popcorn at the right time.

It's the Hype, Baby.

Hype makes subventions rain, and gives some startups a growth push.

2/ Yes, this is dumb terminology throwing around the "AI" buzzword and marketing wank to try to fool people into believing additional generation capacity exists when it doesn't.