Subscribe now

When the Texas grid ran tight on a 105-degree afternoon, Octopus Energy used to throw a party. The retailer hosted happy hours, announced a few days ahead when forecasts showed a demand crunch, built to pull customers out of their houses during the event window. An empty house is an idle air conditioner, and a few thousand idle air conditioners add up to real relief on the ERCOT grid. Look at what everyone walks away with. Customers get a free cold drink and a fun evening. The grid gets capacity at the exact hours it needs it. And an energy company becomes the brand that bought you a beer.

Most demand response programs send a push notification that says “please use less electricity.” Octopus solved the same problem with a happy hour. That difference has a name, and I found it put best in a book I have been reading this month, Starting a Startup by James Sinclair, in a passage that earned a highlight, a dog-ear, and now a full article:

“Create experiences that make users smile, that make their lives easier, that they can’t help but share with others. Because you’re not just building a product, you’re crafting an experience, you’re solving problems, you’re making people’s lives better, easier, more enjoyable. That’s the power of delight.”

I think delight is one of the most underpriced assets in business. It is not a garnish you add at the end. It is a strategy with measurable returns. So, let’s look at some of the best delight ever shipped, what it did for the businesses that built it, the users who felt it, and occasionally the world outside both. Then I’ll close with recommendations for cultivating delight in whatever customer experience you’re designing.

Delight is earned, not decorated

Let’s start with a definition first, because the word gets misused. Confetti on a broken checkout flow is not delight. Every product climbs a hierarchy:

1. First it’s functional

2. Then it’s easy

3. Then it can be delightful

Delight is the moment a product exceeds the emotional expectation a user brought to it. Nobody expects an energy bill to be funny or a language app to feel like a friendship. The surprise, the smile, the laugh out loud is the opportunity.

Every example below follows the same pattern. A company did the foundational, but maybe boring work first, then spent real effort on moments most competitors considered optional or even pointless.

The compound interest of small moments

Mailchimp mapped the emotional geography of its product better than almost anyone. Sending an email to 100,000 people is a sweaty-palm moment with one typo being permanent. So, after you hit send, Freddie the chimp offers you a high five. Two seconds of animation, aimed at the single most anxious click in the product. That is the craft: delight placed where the emotion peaks, not scattered at random.

Duolingo turned the same thinking into a retention engine. The streak looks like a toy. The numbers behind it are serious. Learners who reach a seven-day streak are 2.4 times more likely to return the next day, over 5 million users have held streaks of a year or longer, and churn fell from 47% in 2020 to roughly 28% in core markets, per Sensor Tower and Duolingo’s own reporting. The streak, the passive-aggressive owl, and the celebration animations compound into one of the best retention curves in consumer software. If your goal is a high daily active user count, repeated engagement, and reduced churn, a streak feature may be for you.

Then Duolingo found a delight channel most teams never think to touch, the app icon itself. In October 2023 the icon melted. Duo drooped down the tile, eyes sagging, beak distorted, and millions of users opened the app to figure out why. A later update made him visibly ill, red-rimmed eyes and a snotty beak, because the owl was, in the company’s words, “quite literally sick of reminding everyone to do their lessons.” Duolingo was open about the goal: get learners to open the app. The icon is a tiny amount of real estate every user sees on their home screen daily, and Duolingo turned it into a living character whose mood depends on whether you showed up. Users who hold a 30-day streak earn access to custom icons as a reward. The owl became a meme, the meme became free marketing, and the icon became a retention lever hiding in plain sight.

Spotify redesigned their app icon as well to commemorate their 20^th anniversary in May 2026, turning their logo into a glittery disco ball. Many users took to social media to post about their distaste of the change, but that triggered one of Spotify’s goals, to increase social media engagement and free marketing. It interrupted people’s behavior and people opened up the app again as a high ROI marketing tactic.

The grand gesture

Some delight is quiet craft. Some is an act of humanity so unexpected is resonates with true emotional significance. Chewy is the case that usually comes to my mind as a pet owner and dog lover. When a customer’s pet dies, Chewy has refunded the unopened food, suggested donating it to a local shelter, and sent flowers with a handwritten card. The company works with over 1,000 local artists to send hand-painted pet portraits, sometimes for no occasion at all. In 2022 a customer named Anna Brose tweeted about that exact experience after losing her dog. The post drew over 600,000 likes, and the replies filled with people saying they were moving their pet spend to Chewy; Fortune covered the whole arc. One act of care became a customer acquisition event no ad budget could buy.

Delight in unlikely places

I build products in power and energy, an industry with no reputation for making anyone smile. That is exactly why my favorite delight examples come from my own back yard. The bar is on the floor, so the returns are enormous.

The Nest thermostat’s green Leaf might be the most elegant behavioral design in any home product. Set a temperature that saves energy and a small leaf appears on the dial. No lecture, no kilowatt-hours, no guilt. Millions of people nudged the thermostat a degree to earn a pixel plant. And the Leaf quietly moves its own targets as it learns your habits, the threshold to earn it shifts, so it keeps leading you toward energy savings. It’s a game you don’t realize you’re playing, and the prize is a lower electricity bill.

Octopus runs a similar play across its whole business, and the happy hours are one expression of a deeper culture. In the UK, the company writes emails on a first-name basis, cracks jokes in support chats, and spins a wheel of fortune that rewards customers for submitting meter readings. Its Saving Sessions program has paid customers over £1 million for shifting usage away from peak hours, turning grid flexibility into a rewards game. Bain’s 2023 State of CX UK report measured the cumulative result: an NPS 39 points above the industry average, the largest gap of any company in any sector. That loyalty helped Octopus become the UK’s largest retail energy supplier. When every competitor sells the same product at nearly the same price, being loved is the moat.

One more thing about the energy examples. Every Nest Leaf earned is a negawatt, or energy that would typically be consumed but is never generated. Every customer sipping a happy hour drink during a demand response event is load off the grid at its most stressed hour. Every customer who likes their utility is more willing to join a smart tariff, shift EV charging overnight, or enroll a battery in a virtual power plant. Aimed well, delight changes physical outcomes on the grid, and the retention curve comes along for the ride.

What delight actually buys

Three layers of return show up across all of these. The business gets retention, word of mouth that pushes acquisition cost toward zero, and pricing power plus forgiveness when things break. The user gets a product that treats them like a person having an emotional experience: anxiety, grief, pride, boredom. These experiences are rare enough to be memorable. And when delight points at behavior that helps people, like daily language learning, saved energy, or donated pet food, the compounding continues outside the P&L.

One honest caveat. The same mechanics can be aimed the wrong way. A streak can motivate, or it can manufacture anxiety. A charming interface can serve the user’s goal, or camouflage the company’s. My test is simple: does this moment serve what the user is trying to do, or only what we want them to do? Delight that fails that test is manipulation with better art direction.

Cultivating delight in your own product

So how do you build this into whatever customer experience you’re designing? Two prerequisites, then three practices.

The prerequisites: earn the right first, because no amount of confetti celebrations fixes a flow that fails and frustrates a user along the way. And give your team explicit permission for craft, because nobody writes witty copy for a landing page when every sprint is a fire drill. Chewy’s agents send flowers without asking a manager. That is a staffing and empowerment decision, made long before it becomes a viral tweet.

Practice one: solve the need your customer realizes they have in the moment, before they express it. This is the highest form of delight because it proves you understand their life, and it converts a pain point into a loyalty point. Chewy refunds the food before the grieving customer asks. Octopus invites you to a happy hour before you’ve consciously thought “I’d rather be anywhere but this hot house.” The Nest Leaf answers “am I doing this right?” before the question forms. Study the moments where your customer is about to need something, then give them the solution early. Anticipation, more than surprise, is what people describe when they say a product feels like magic.

Practice two: do things that make users smile or laugh, on purpose and in the right place. Map the emotional peaks of your journey, the send button, the first bill, the bad news, the 404 page, and aim there, because delight placed at a peak is worth ten placed at random. Freddie’s high five lands at the scariest click in Mailchimp. Duolingo’s melting icon lands on the home screen at the exact moment a streak is about to die. A joke in a support chat lands when the customer expects a script. Humor is a risk, and that is precisely why it works. A brand willing to be funny reads as a brand actually run by real people.

Practice three: use design to deliver a curated experience rather than an exhaustive one. Curation means making choices for the user instead of handing them thirty settings to choose from. Nest compressed an entire home-efficiency report into one green leaf. Duolingo compressed motivation into a single streak number with a flame next to it. In each case the design team decided what the user should feel and cut everything that competed with it. If your product’s best moment requires the user to configure it, you haven’t designed it yet.

Then measure the echo and virality. Screenshots, unprompted posts, a customer showing their phone to a friend and prompting a referral. Those are delight’s native metrics and they lead to the retention and referral numbers your CFO cares about.

Sinclair’s quote still holds weight, parapharsed: create experiences that make users smile, make their lives easier, and often those experiences will lead to them sharing that experience with their network. Every product is a stream of moments in someone’s day, and each moment is a choice: make it just functional or make it worth a smile? The companies above chose the smile and collected loyalty, growth, and in a few cases a slightly better grid.

That’s the power of delight.

Subscribe now

Last week, Renew Home, Sunrun, and Tesla, announced an agreement to make available 16.8 gigawatts of virtual power plant capacity directly to utilities, hyperscalers, the PJM’s reliability backstop process, and first-come first serve interested parties. They are calling it a “Capacity-as-a-Solution framework”. A data center that could otherwise spend several years in the PJM interconnection queue could theoretically have capacity under contract before its next board meeting.

To put 16.8 gigawatts in terms most people can understand, 1 megawatt can power roughly 800 homes (considering regional differences in electricity usage patterns and average home size). On that basis, 16.8 gigawatts represent enough capacity to serve roughly 13.4 million homes, or roughly twice every household in New York City, Los Angeles, and Chicago combined. That comparison is about the massive scale of this partnership based virtual power plant. Caveat, virtual power plant resources typically dispatch for two to four hours at a time, not continuously to replace baseload. Nevertheless, the distributed energy assets are real, already installed, and growing at a healthy pace every day. Sunrun has been at least doubling their VPP megawatts under management at year over year recently. Nor do data centers need VPPs to replace their baseload power. The play is using VPP dispatch to maintain operations during grid stress events when utilities would otherwise ask data centers to curtail.

WHY MEGAWATTS MATTER

A megawatt (MW) is 1,000 kilowatts. Based on EIA data, 1 MW can serve roughly 800 average U.S. homes when running continuously. A gigawatt (GW) is 1,000 megawatts, or enough for about 800,000 homes. A single large nuclear reactor produces roughly 1 GW. The 16.8 GW in this announcement is the nameplate capacity of nearly 17 nuclear reactors, except instead of one central power plant, it is distributed across millions of homes, thermostats, and batteries in every major U.S. power market.

Speed to power has become the defining constraint in data center development, that’s why there are new well-funded startups like GridCARE, who recently raised a $64 million Series A in May 2026 with a core mission is to solve the speed to power issue. Hyperscalers are committing capital faster than generation developers can build the infrastructure to serve it. In Northern Virginia, the densest data center market in the world, Dominion Energy has a backlog of large-load interconnection requests it cannot serve on existing infrastructure. PJM released a backstop procurement framework in April to procure 14.9 gigawatts of new power for large loads by 2031, roughly the output of 15 nuclear reactors, an acknowledgment that the grid does not have headroom for the data center pipeline.

The combined 16.8 gigawatts come from two fundamentally different products. Roughly 54%, about 9 gigawatts, is Renew Home’s demand response network: more than 8.6 million homes of 11 million enrolled smart thermostats and connected devices that can reduce consumption within seconds of a dispatch signal. These are load curtailment assets.

The remaining 7.8 gigawatts comes from Sunrun and Tesla’s battery systems, which can actively export electricity onto the grid during high-stress periods or on a daily passive schedule. If we’re talking a Tesla Powerwall 3, which the majority of the batteries are, then every single of these 1.2 million batteries can move about 11.5 kW each back on to the grid simultaneously.

The capacity is geographically distributed across every major power market in the country, according to vppcapacity.com, which contains an interactive map to drill into each state plus D.C and Puerto Rico. California holds the largest concentration at 4,660 MW, enough to serve roughly 3.7 million homes simultaneously, more residential households than the entire state of Colorado. Texas sits at 1,730 MW, roughly the residential electricity load of the Dallas-Fort Worth metro area, and the second-largest data center market in the country. Puerto Rico ranks third at 1,540 MW, nearly all of it battery storage deployed after Hurricane Maria for residential resilience. That capacity operates on an isolated grid and serves a different purpose than the mainland programs discussed here, it’s included in the dataset but likely is not part of the BYOC framework available to continental data center operators. Virginia shows 313 MW available today, equivalent to the residential load of roughly 250,000 homes, more than twice the size of Richmond, Virginia’s entire customer base, with a committed path to 500 MW by 2030.

WHAT 300 MW MEANS FOR A DATA CENTER

A single large hyperscale data center campus typically draws between 100 and 500 MW of power. The 313 MW Renew Home, Sunrun, and Tesla have available in Virginia today is, on nameplate, roughly equivalent to one full large-campus power draw. That comparison illustrates the immediate commercial opportunity in the state, but also the gap: Virginia alone has many active data center campuses.

Virginia is where the commercial case is sharpest. Loudoun County alone hosts more than 130 data centers, and Dominion Energy’s queue of large-load requests has become one of the most-watched constraints in the power industry. The state passed a first-of-a-kind law this year directing Dominion and Appalachian Power to quantify and reduce grid waste through distributed energy resources.

The commercial structure has a precedent. Voltus introduced it as “bring your own capacity” last fall. The model is a hyperscaler contracts VPP capacity from distributed resources near its data center, the utility calls on those resources during high demand events, and the data center avoids curtailment to its own GPU workloads and operations. Voltus signed its first major deal under this framework with Google a month ago in June, a 100-megawatt contract in PJM, roughly the load of a mid-size data center campus. Renew Home, Sunrun, and Tesla are announcing open for business at 168 times that scale. I view this as one of the most consequential market signals VPPs have ever generated.

Now that hyperscalers are beginning to treat DER portfolios as part of their capacity procurement stack rather than a supplemental curtailment program, the economics of the whole system shift. The value of the megawatts from VPP portfolios may change dramatically. Traditionally VPP’s earn revenue from participating in utility, state-sponsored, or ISO programs that help the grid in times of stress. Think CAISO’s DSGS and ELRP programs, ERCOT’s ADER, National Grid’s ConnectedSolutions program, or any of the ISO’s energy or capacity markets. Those were the default places to deploy managed megawatts historically, but I believe this will shift pretty quickly as more and more supply agreements are reached with data center operators.

In Voltus’s BYOC Google press release they stated “Through this initial deal, Google and Voltus aim to create an industry-leading scalable blueprint for how data center capacity needs can be met affordably and reliably through smarter utilization of the grid. By proving this can be a valuable tool for ambitious data center load growth, the partnership establishes a repeatable path for other large energy users to follow – accelerating clean capacity deployment, bringing direct economic benefits to local communities, and contributing to a robust, reliable grid for everyone.” I believe this framework is exactly what Renew Home, Sunrun, and Tesla are trying to follow and personally I’m very excited to see where this leads for the future of VPPs.

The residential DER build-out that VPP practitioners assembled over the last decade, one battery and one enrolled thermostat at a time, just became the largest power offer in the country. The announcement tells us that the supply exists, and I believe the power-hungry market participants will respond quickly to take advantage of that supply.

Thanks for reading! I also combed through and pulled together the state by state, plus District of Columbia and Puerto Rico data so you don’t have to.

Sources

Household electricity consumption: EIA FAQ — eia.gov/tools/faqs

State-by-state VPP capacity data: explorer.vppcapacity.com

Other Coverage: Latitude Media · Heatmap News · Canary Media · Electrek · Sunrun Press Release

For a decade, the residential energy story started with solar. Batteries were the add-on, the upsell, the extra line item for customers willing to stretch the budget. That framing made sense at the time.

It doesn’t anymore.

What 2025 and 2026 are showing us is the inversion. Batteries installed at scale, doing the real work, they’re storing energy, dispatching to the grid during peak demand, protecting homes when the power goes out, and increasingly, paying homeowners back for all of it. The solar charges them, and then the magic happens.

GoodLeap just crossed 50,000 residential home battery installs across our loan and TPO financing products.

We deployed 68.4 MW of GoodGrid VPP eligible storage in all of 2025. By the end of April 2026, we were already at 63.5 MW, roughly 93% of last year’s full-year volume in just four months. That is not a normal growth curve. That’s a signal.

A Third Lever for the Grid

Resource planning has always had two levers for responding to peak demand: build more generation or pay large customers to curtail. The first is slow. Both are expensive and can take years before a single megawatt shows up when the grid needs it.

A residential VPP fleet at this scale adds a third lever, one that grows by enrollment instead of years of construction. A single home battery can dispatch 7 to 15 kWh during a grid event. Aggregate that across tens of thousands of homes and you have hundreds of megawatts of fast, behind-the-meter capacity available during a heat wave, a cold snap, or a summer evening peak. No new substation. No new peaker plant. No new transmission line needed.

The capacity is already in the ground. The question for utility planners is whether their modeling reflects that reality and most resource planning today does not account for this reliable capacity.

What Scale Actually Unlocks

At 100 batteries, you have a pilot. At 50,000, you have infrastructure, and the programs that weren’t economic to design at small scale become straightforward. Targeted dispatch in constrained feeders. Coordinated capacity market participation. Responding to demand fast enough and large enough to move the needle during the hours that matter most.

There’s a resilience dimension here too. A megawatt inside a neighborhood behaves differently than one at a distant generation facility when a wildfire shutoff hits or a winter storm takes down transmission. The electrons do not have to travel through the infrastructure that just failed. They are already there.

The Deployment Curve Is the Real Story

Going from 68.4 MW for a full year to 63.5 MW in four months doesn’t happen because one company executed well. It happens because the underlying economics shifted in a way that unlocked a new category of customers entirely.

That shift, and what it means for where residential storage goes from here, is worth its own piece. The short version: changes to the ITC made leased and PPA battery assets the fastest-growing segment in residential storage, and every one of those assets is an enrolled, dispatchable grid resource from day one.

Affordable access didn’t just expand the market. It expanded the grid.

No utility issued an RFP for a virtual power plant of tens of thousands of batteries. No grid operator designed it top-down. It was built one home at a time. The grid infrastructure of the next decade is going to look entirely different than the grid of the last century.

The grid needed scale to take residential storage seriously. Next up: we’ll look at the homeowners who built it and the changing market frameworks.

Big Tech is moving from large power customer to grid shaper, and flexible energy through virtual power plants (VPPs) are the battleground. If utilities and traditional VPP aggregators don’t modernize how they recruit, measure, and pay for distributed flexibility, customers will pick tech-driven options; and market power over demand-side resources will shift with them.

The stakes are immediate: data-center load is exploding, interconnection queues are clogged, and the fastest scalable capacity is on the customer side of the meter.

Why this is a race

The numbers are changing fast. Wood Mackenzie reports a 33% jump in VPP deployments year-over-year, with North America’s VPP capacity reaching ~37.5 Gigawatts in 2025 and a surge of new off-takers tied to data center demand. Yet capacity growth is lagging deployments, a sign that market rules and utility enrollment caps, not technology nor consumer demand, are the binding constraints.

Translation: demand for flexible capacity is real, however, program design and accreditation for these flexible resources need to catch up.

At the same time, the U.S. Department of Energy says scaling VPPs to 80–160 GW by 2030 could cover 10–20% of peak load and save ~$10 billion annually, a scale only achievable if customer enrollment and measurement become as seamless as any consumer app. That is Big Tech’s home turf.

Big Tech’s opening moves

Meta just created a power-marketing arm, Atem Energy, and filed at FERC to sell energy, capacity and ancillary services in wholesale markets, directly aligning its data-center growth with energy market participation. It is a clear signal: hyperscalers won’t wait passively for capacity. They will procure, build, and even sell it.

Google has already helped assemble residential flexibility at scale. In 2024, Google’s Nest Renew combined with OhmConnect to form Renew Home, which claims several gigawatts of controllable load today and a goal of 50 GW by 2030.

Amazon brings distribution more than programs, with hundreds of millions of Alexa-enabled devices and the Sidewalk low-bandwidth network that connects sensors and appliances outside Wi-Fi’s range. Pair that extensive plumbing with simple “Energy Savings Incentives” enrollment for the Amazon Smart Thermostat and you have a ready on ramp to mass market demand response, even if Amazon isn’t the aggregator of record.

Seen together, these moves don’t yet make Big Tech the winner in VPP operations. They make Big Tech a powerful distribution layer that VPPs ride on: instant customer reach to massive existing audiences, simple UX, and the cloud and AI tooling to forecast and dispatch capacity at scale.

Data centers change the stakes

The grid crunch is not abstract. New AI-based facilities are multi-hundred-megawatt to gigawatt-scale, with timelines measured in quarters, not decades. Rocky Mountain Institute’s (RMI) latest brief argues that “many traditional solutions now face high costs and supply chain constraints, finding new ways to quickly, cost-effectively upgrade the electricity system will be essential to power the next stage of American economic growth.” Rapid data center expansion heralds that next stage of economic growth and VPPs are the fastest, cheapest way to enable this step into the future. RMI outlines contract models in which large loads sponsor VPP capacity to earn faster interconnections. That’s a blueprint for hyperscalers to finance CELI's Medium pagelocal flexibility and for utilities to safely count it toward capacity needs.

Emerging offers make this concrete. Voltus’s “bring-your-own capacity” model lets a data-center developer fund a VPP in the surrounding community and deliver those megawatts directly to the utility’s capacity stack, an out-of-market structure aimed to meet the needs of hyperscalers between now and 2030.

WattCarbon’s “Repowering California” proposal would issue certificates for verified, utility-accredited savings from distributed energy resources that data centers can use to demonstrate local headroom when negotiating interconnections. Different mechanics, same idea: pay to stand up flexibility where you need it, now.

What Utilities can do now

• Treat VPPs as capacity, not pilots. Publish locational capacity needs and standardize accreditation so third parties (including hyperscalers) can fund VPP Megawatts that “count” toward interconnection needs and resource adequacy. Instead of continuing to proceed in the small capped megawatts world of the past.

• Adopt “sponsor-a-VPP” tariffs. Create optional riders that let large loads finance VPPs and recover costs while protecting other customers, mirroring how power purchase agreements operate today.

• Open up data access…securely Implement Green Button Connect–style sharing to accelerate enrollment and measurement & verification, while maintaining distribution-system safeguards. (RMI)

• Be an aggregator-of-aggregators. Procure capacity from multiple VPP providers (and, where it pencils, from hyperscaler-sponsored programs) to diversify portfolios and lower procurement cost and risk.

What Big Tech must prove

Big Tech’s platform edge won’t matter unless it delivers additional, reliable, and verifiable capacity at the specific nodes where utilities need it:

• Additionality: They can’t just rebadge legacy thermostat demand response capacity. They must finance resources that provide multi-hour, seasonal headroom, not only emergency curtailments.

• Reliability: Offer telemetry and performance guarantees familiar to system planners; align dispatchability with local constraints, not just wholesale price spikes. RMI’s models aim squarely at this bar.

• Speed with governance: Coordinate with incumbent planners so distribution protection, outage response, and customer bill impacts aren’t afterthoughts.

The control point that decides the winner

This race won’t be decided by who owns the most thermostats or batteries; it will be decided by who owns the enrollment, data, and workflows that turn devices into accredited capacity where and when it’s needed down to a hyper local level. Today, the consumer funnel and UX favor Big Tech. The authority to make those megawatts “real” still sits with utilities, aggregators, and market operators.

If utilities lean in, publishing needs, standardizing accreditation, providing data access, and buying capacity from VPPs, the result is faster interconnections for data centers, lower system costs, and better monthly utility bills for residential customers. If they don’t, Big Tech and its partners will keep inventing workarounds and, piece by piece, redefine who controls the future of the grid. The race is on, the smart money will fund flexibility and insists it counts.

Subscribe now

Leave a comment