Your hot-water tank is a battery. GridShare Japan’s CEO on letting AI run the home

I realized I don’t actually know much about electricity. Not properly, anyway.

We recently wrote about AI data centers and the power they need. But electricity isn’t just for data centers. It runs your lights, your fridge, your hot water. It’s the most everyday piece of infrastructure we have.

So I started with renewable energy: solar and wind. Japan aims to get 40–50% of its electricity from renewables by fiscal 2040, according to its national energy plan. How well is the country using the renewables it already has?

The answer surprised me. On sunny spring afternoons, Japan makes more solar power than it can use. To keep the grid from overloading, utilities ask plants to stop generating, a step called output curtailment. In fiscal 2024 alone, about 1.6 billion kWh was switched off this way. That’s roughly what 400,000 Japanese households use in a year (our calculation). In March 2026 it happened in the Tokyo area for the first time, which means every region of Japan now does it.

The money tells the same story. Homes with rooftop solar can sell their power at a fixed price for 10 years. In the first year of the program, that price was ¥48 per kWh. After 10 years, it drops to ¥8.5 from TEPCO Energy Partner, Tokyo’s main utility, or ¥7 from Kyushu Electric. Power bought at night costs several times that.

Households sell their midday power cheap, then buy power back at night for much more. Even at home, Japan isn’t using its renewables well.

Homes do have ways to store energy: a battery, and the EcoCute, a Japanese heat-pump water heater that keeps hot water in a big tank. GridShare Japan, part of the trading house Itochu’s group, uses AI to run those devices and cut the waste. It connects to about 40,000 batteries and EcoCutes from different makers and decides, home by home, when to store energy and when to use it.

And then, we got to sit down with the company’s president, Masashi Nishio, himself!

Here’s the short answer up front. GridShare Japan has two sides.

• For homes: an AI autopilot
  The AI runs the battery or EcoCute a family already owns, every day, for a monthly fee.
• For utilities: pooled power from tens of thousands of homes
  It runs many homes’ devices together, helping balance the grid when power runs short or is in surplus.

Use even the hot-water tank as a battery, pool tens of thousands of homes, and you get something a utility can lean on, like a small power plant. Let’s take it one piece at a time.

WHO IS

Masashi NishioMasashi Nishio

President & CEO, GridShare Japan

“We're connected to 40,000 devices, and we run 40,000 custom control plans, one for each customer”── Masashi Nishio, in this interview

1. 1999From optimal control to ItochuStudied systems science at the Tokyo Institute of Technology’s graduate school, working on models that steer systems to their best state. At Itochu, worked in logistics, investment, corporate turnarounds, M&A and more
2. 2011–2018An MBA, then a British AIEarned an MBA at the University of Michigan. Led the alliance with UK startup Moixa and founded GridShare Japan in 2018
3. NOWPresident & CEO, GridShare JapanRan TRENDE, a peer-to-peer power trading company, before taking the top job in April 2025

Born in 1974, Nishio graduated from Chiba University’s engineering faculty, studied systems science at the Tokyo Institute of Technology, and joined Itochu in 1999. He brought a British battery AI to Japan and founded GridShare Japan in 2018. In 2022 he led the partnership with US-based Lunar Energy, which had acquired the AI’s developer. From June 2023 to March 2025 he ran TRENDE, a company that lets households trade power with each other, and in April 2025 he became president of GridShare Japan.

There are two main sides.

The first is households. Families with solar panels and a battery or an EcoCute sign up for our AI service on a monthly contract. The AI learns how each home uses power and runs the devices in the way that saves the most. For the customer, the power bill goes down with no effort.

The second is utilities. That one is about running many homes’ devices together. I’ll come back to it in a bit.

Exactly. If you have a supported device, you just sign up. Nothing extra gets installed in the house. Right now we support batteries from Itochu, Omron and Kyocera, and Daikin’s EcoCutes.

From there the AI takes care of everything, so customers don’t have to lift a finger.

That’s our biggest strength.

Makers are good at running their own equipment. But switch to another brand or another kind of device and they don’t know it. Battery people know batteries, but not EcoCutes, and the other way round.

We can run devices from different makers together, from the customer’s point of view. And if you ran a battery and an EcoCute yourself, you’d have two different remotes. We get rid of all that and run the whole house for the best result. These days, you don’t even need a remote. The AI does it.

Customers can see the savings every day in the app. On average, with both a battery and an EcoCute, it comes to ¥4,635 a month, about ¥50,000 (roughly $330) a year.

We interview customers regularly, and we hear things like “my bill dropped 40 to 50%,” “it’s about ¥15,000 lower since April,” and “I leave everything to the AI and it’s fine.” That’s why hardly anyone cancels. Once people join, they basically stay.

VOICE

After installing it, our power bill dropped sharply, to about half of last year’s. In April, for example, it was about ¥15,000 lower than a year earlier. I can really feel the savings.

VOICE

Since we added the AI-controlled battery, our power bill is about 40% lower than before. We’re very happy with it.

VOICE

I leave the battery entirely to the AI. It hasn’t caused any inconvenience in daily life, and it runs without problems.

Half the bill. ¥15,000 off in a single month. I had to read those customer comments twice. And all they did was let the AI handle it.

Even after we optimize each home, there’s spare capacity left. We pool it and run it when a utility asks. That’s demand response.

When a utility says, “We’re short of power, please discharge the batteries,” or “There’s too much, please charge now,” we run the devices that way, with the customer’s consent. Sometimes that costs the customer a little, but the utility gives them points, so they come out ahead overall. We do this with Tohoku Electric, TEPCO and Chubu Electric Power Miraiz, and we’re talking with other utilities.

Yes. It works out to something like a small power plant.

Beyond that, we’re preparing to trade customers’ power in the market. Owning these devices lowers your bill, earns you points, and eventually lets you trade. The more of these devices you own, the more you gain, so more people want them. That’s the ultimate goal.

That’s exactly the idea.

Households get lower bills and peace of mind. Stored power also helps in a blackout, so it doubles as disaster preparedness. Utilities get flexibility to balance the grid, and we pass that value back to customers. Our goal is to raise the value of the home and of the power system at the same time. If a device saves its owner money, people choose that device, so it’s good news for makers too.

Systems engineering. You describe how something works in equations, then control it so it runs as well as possible. I also did a bit of research on what we’d now call AI.

But I wanted to use it in business rather than stay in research, so I joined Itochu. My first posting was logistics. I’d spent years on theory, and I wanted to see how physical things actually move.

I joined in 1999, right in the dot-com bubble. I got the chance to work at a distribution center in the US, for one of the early companies selling sporting goods online.

Day to day, I checked whether the stock on the shelves matched the stock in the system, alongside the local staff. I also worked on the warehouse design. People held stock options and worked late into the night. That was my first taste of startup energy.

Next I worked on acquisitions and venture investments. Investing is almost the same as running a business. You learn to judge which fields can work.

In 2011 the company sent me to do an MBA. The hot topic then was sustainability. Lithium-ion batteries and solar were taking off, and EVs were just hitting the road. Back in Japan, I started working on home battery sales. Most of what we do today, we were already debating in that MBA class.

There was. It came down to a program called FIT. Do you know it?

FIT stands for feed-in tariff. If you put solar on your roof, the utility buys the power at a fixed price for 10 years for a household. Back then the price was over ¥40 per kWh, and electricity was cheaper than now, so it was an incredible deal. The more you sold, the more you made. That’s why rooftop solar spread so fast in Japan.

But when the 10 years end, everything changes. The price drops to somewhere between ¥7 and ¥11.

Buying power cost about ¥30 back then and about ¥40 now. If you sell at ¥8 and buy at ¥30, you’re far better off using it yourself. So you buy a battery, store the midday power and use it at night.

But nobody can check the weather, think about the family’s plans and flip the battery between charging and discharging every day. So let AI learn the tariff, the solar output and the household’s habits, and do it for you.

I also didn’t want a battery to be a one-time sale. I wanted it to keep saving customers money after they bought it, like a subscription. That’s when I came across Moixa, a startup in the UK.

Moixa ran home batteries with AI in the UK, at a scale of a few hundred homes. Itochu formed a capital alliance with it and got the rights to roll out the AI in Japan. That’s how GridShare Japan began. In 2018 we started selling Itochu’s Smart Star batteries with the AI built in.

No, that part was hard. We spent six months to a year adapting it. Everything was in English, and they didn’t know how Japan’s power system or prices worked.

Japanese utilities, for example, offer all kinds of time-of-use plans. We taught them those things one by one and built a Japanese version. Seven years on, they have Japanese engineers there too.

Exactly. In Japan we connect to about 40,000 devices. At the time, no other AI was running that many scattered devices at that scale.

That track record caught the eye of Lunar Energy in the US, which bought Moixa. Today the same AI runs about 100,000 batteries in the US and about 140,000 batteries and EcoCutes worldwide. The software team is in the UK, and they ship it to the US and Japan.

That’s roughly right. The key is how much information goes into that decision.

There’s the weather. There’s how the family uses power, which differs between weekdays and weekends, and between school holidays and normal weeks. Bigger solar panels make more power. Every home has a different tariff. Batteries come in different sizes. The AI takes all of that in and runs the home in the way that saves the most.

So with 40,000 customers, we’re running 40,000 custom control plans.

We process a lot of data. Using each customer’s history, the AI forecasts ahead while solving, every day, the problem of how to run the home for the best result.

I actually studied this kind of thing in graduate school. It was all theory then, but I understood the logic. In a roundabout way, I’m now using what I learned as a student in business.

It catches up with a home in about a week. After that, the longer someone uses it, the more accurate it gets. A big jump at first, then steady gains.

Say tomorrow’s forecast is rain. A person would think, “It’s going to rain, so I’ll fill the battery tonight with cheap power.”

But solar panels make a fair amount of power even on cloudy or drizzly days. The AI always has a 48-hour forecast, so it can decide, “It’ll be cloudy, but we’ll still get this much. Buy only a little tonight and fill the rest from the sun.” It can make the best call, one that’s hard for a person to make.

The default modes built into the devices are designed to play it safe, so they don’t go that far. Doing this by hand every day? Practically impossible.

An EcoCute usually heats water late at night with cheap power and stores it in a tank. It makes the morning and evening hot water in one go.

Meanwhile, even homes with a battery have spare solar at midday. The battery is often full by around noon, and after that the power has nowhere to go. So move the heating from the middle of the night to that sunny window.

An EcoCute heats slowly, drawing about 1.5 kW for four to five hours. Once heated, the water stays hot until evening. You’re storing electricity as hot water.

Yes. Nobody else offers home-by-home EcoCute control, because it’s technically hard.

In a home with only an EcoCute, there’s no data on how much the rooftop solar is producing. A battery knows, but an EcoCute isn’t connected to the power meter. If you can’t tell how much power you’ll get, shifting the heating to midday is risky. Running out of hot water is a big deal.

So we use satellite data. Beyond the weather forecast, we look down at the clouds, even how thick they are, and forecast what that customer’s roof will produce. Then we shift the heating accordingly. So far, we’re the only ones in the industry doing it.

Usually you add a HEMS, a home energy management system, to measure the output, or customers have to type in when they can use power. A HEMS can cost ¥50,000 to ¥200,000.

We don’t need one, so customers can start any time, even years after buying the equipment. They just connect the EcoCute to home Wi-Fi and register it in the maker’s app. Once they sign up, we connect to it.

We have real customer data. Lay the AI’s forecast over what actually happened, and they almost match.

Take one home with both a battery and an EcoCute on a clear day, May 30 this year. While solar charged the battery, the EcoCute heated water twice around midday. The tank went from 40% to 100%, a 60% top-up that normally happens overnight. That day moved about 6 kWh to midday. At ¥30 per kWh for late-night power, that’s about ¥180 saved in one day.

Here’s a fun one: a customer in Matsudo, east of Tokyo, on June 18, a day of drizzle and cloud. The panels are rated at 6 kW but peaked at only 3.6 kW. Still, they were producing. A person would have heated the water the night before because rain was forecast. The AI read the weather the previous day, skipped the overnight heating and decided midday would be enough.

It heated the water first, then switched straight to charging the battery the moment it finished, so the two never competed for power. That day shifted 2 kWh, about ¥60.

It really does add up. Batteries save ¥2,115 a month on average. EcoCutes, based on trial data, save about ¥2,520. Together that’s ¥4,635 a month, roughly ¥50,000 (about $330) a year.

The clearest measure is how much of your solar power you use at home. With solar alone it’s about 30%; the other 70% gets sold cheaply. Add AI-run EcoCute and it’s about 50%. A battery gets you about 70%, and both together about 90%.

If I’m honest, I wish makers would standardize their specs. Who wouldn’t?

But adapting to each of them is part of our know-how. Our job is to take away the inconvenience, and clearing those technical hurdles one by one is the value we bring. Technically, it’s all doable.

Mostly development: connecting more kinds of devices. We want a world where whatever combination of brands a customer buys, they can join GridShare. Right now it’s batteries and EcoCutes, and we’ll keep expanding. Makers have their own policies and technical constraints, so we work through them one at a time.

The rest goes to letting people know what it does, so more of them use it. We also want our current customers to be happier, so we’re improving the app and giving points to people who answer our surveys.

It may be small, but step by step, I think we can help Japan decarbonize.

The EcoCute launched in 2001. Twenty-four years later, shipments in Japan have passed 10 million. Picture a giant hot-water bottle set beside home after home across the country. Almost all of them were built to heat at night.

What if that heating moved to sunny middays? A government study group estimated, as reported by ITmedia, that if half of all EcoCutes heated at midday, 5.2 billion kWh a year of demand would shift into daylight. That’s more than three times the roughly 1.6 billion kWh of solar and wind Japan switched off in fiscal 2024. (It’s an annual comparison; the timing won’t line up perfectly.)

Japan doesn’t need millions of new batteries to start. The tanks already in people’s homes can stand in for them. What’s been missing is a brain to run them and a way to reach them. In a country where only 2.7% of homes have a HEMS, GridShare reads output from the clouds and runs EcoCutes with no extra hardware. I see its work as waking up those 10 million sleeping tanks.

On a smartphone, the camera, maps and payments come from different companies, yet they all run on one device. That works because there’s an operating system in between.

Home energy isn’t there yet. The battery has one maker’s app, the EcoCute has another maker’s remote. When Nishio says “battery people don’t understand EcoCutes,” it’s as if every app needed its own phone.

GridShare adds a layer on top that sees the whole house. That’s how it can decide, on whole-home economics, whether to fill the battery or heat the water first. Once Japan’s DR-ready rules take effect, connecting will get easier. But the job of deciding for the whole house can only sit outside any single maker. As standards spread, I think that layer gets more valuable, not less.

Resale apps turned the clothes and books in our closets into things we could sell. Electricity is heading the same way.

In California, Sunrun has pooled batteries in more than 80,000 homes and can send out up to 425 MW when evening power runs short. In Germany, about 2.2 million batteries, mostly in homes, were running as of the end of July 2025, according to the German Solar Association.

And Japan? The Japan Electrical Manufacturers’ Association counts 1.136 million home batteries shipped. That’s about half of Germany’s count. Add 10 million EcoCutes, though, and the picture changes completely. And from fiscal 2026, home devices can take part in the power market.

What to watch from here depends on who you are.

• Housing, equipment and utility companies: Homes are shifting from “sell and done” to “connect and keep delivering value.” The winners will design the connection in when they sell the equipment.
• Engineers and data people: Reading output from the weather, learning each household’s habits, running tens of thousands of devices at once. Forecasting, optimization and fleet control are running at real scale inside Japanese homes right now.
• Students and future homeowners: Rooftop power is moving from something you sell, to something you use up, to something you pool and trade. How we value a home is starting to change.

The AI era, which eats so much electricity, may also be the era when home power starts holding up the grid. That’s the idea I most wanted to bring home from this interview.

The sun shines at one time; people use power at another. That simple gap leads to switched-off plants at noon and expensive power at night.

GridShare Japan closes the gap with equipment people already own. It runs batteries smartly, heats water at noon and pools tens of thousands of homes into something a utility can use. The more homes that let AI handle their power bill, the less of Japan’s sunny afternoons go to waste.