Once or twice a year, a company called Griddy makes headlines for the insanely high electricity bills incurred by consumers in ERCOT, the Electric Reliability Council of Texas. This traditionally happens during the hottest days of the Texan summer, when the world of wholesale energy traders and grid operators are biting their fingernails around near-miss blackouts — days which usually come and go with most consumers remaining blissfully unaware of the creaking and straining grid powering their air conditioners. During the week of Valentine’s Day 2021, the headlines were caused by an extreme cold snap instead of an extreme heat wave, but the market dynamics remain the same: too much demand and not enough supply, with prices held constant at the artificially imposed price cap of $9000 per megawatt hour (MWh) for days on end. This particular consequence of the ERCOT blackouts has predictably caused politicians and pundits from both sides of the aisle to scream foul, but the reality is that the “bad guys” in this situation are not so clear cut. As an entrepreneur and CTO in the wholesale energy space who’s worked with major ERCOT players (as well as in several other global RTOs) for over three years, I’d like to offer some nuance and context for consideration.
TL;DR: time of use rates are, on net, a Good Thing; markets are great, except when they fail; low risk isn’t the same as no risk; we’re dealing with a homicidal planet of our own creation; grids are incredibly complex, the people running them are generally competent professionals, and we should mostly be in awe that it all works in the first place.
This story actually starts about thirteen years ago, during the 2008 financial crisis. Section 306 of TARP (“Accelerated Depreciation for Smart Meters and Smart Grid Systems”) allowed utilities to specifically recover the cost of smart meters over a 10 year period, as opposed to the traditional 20 year recovery period for infrastructure expenditures. The mechanics of ratebasing and the utility business model is beyond the scope of this post, but just be aware that this created a significant incentive for utilities to spend hundreds of billions of dollars deploying fancy revenue-grade smart meters (at $3k-a-pop) on buildings all around the country. And this is exactly what they did over the past decade: we went from 27M meters in 2011 to 107M today, installed at premises all around the country.
What is a smart meter exactly? Otherwise known as Advanced Metering Infrastructure (AMI) by energy wonks, a smart meter is a device that logs premise-wide power usage at high-resolution intervals. It can range from five minute to hourly resolutions. This is in contrast to older analog “dial” meters that simply aggregate net volumetric power usage over a monthly interval, to be checked once a month (if they even bother!) by a manual human reading. In contrast, smart meters can be read remotely, en masse, via radio waves — allowing for much higher velocity data around how much energy is exactly consumed, by whom, and when. These data readings are then collated to become your monthly energy bill.
The promise of smart meters was to allow a whole new world of energy efficiency, demand response, advanced smart grid operations, and business model and pricing innovation. It’s this last value proposition that I want to focus on. What kind of business model and pricing innovations are we talking about, and how can fifteen minute vs. monthly data recordings impact anything about the core business of serving electricity to consumers?
Let’s zoom way out. You might be aware that electricity markets exist. They operate similarly to other commodity markets: you can buy and sell futures contracts to consume or produce a specified volume of product at a specified future date, to be settled in cash at the real time price when the contract expires. Instead of wheat, or corn, or a physical commodity, the product here is electricity. And similarly to any other commodity market, you have all kinds of market participants at play: merchants interested in securing product for resale at margin, speculators playing in volatility to score returns, market makers ensuring liquidity, producers selling primary physical product, and more. In the electricity market, these players are represented by electricity retailers, utilities, institutional energy traders, and independent generators. The market venues share the same general similarities: you’ve got real time markets, day-ahead auctions, wholesale markets, OTC desks, custom derivatives, risk hedging, and everything else you might read about in a typical Money Stuff column.
Of course, we’re not transacting in corn or wheat. We’re dealing with electricity, and electricity has the unique property that active supply must always equal demand, at millisecond resolutions. At all times, every electron that you use was generated in a power plant miles away and transmitted to you almost instantly. This requires a delicate dance, coordinated by the regional transmission organization (RTO), which constantly monitors the past, present, and anticipated electricity demand, coordinating generators to come online and provide power at a constant 60 Hz to the grid region.
This dance can result in significant price volatility. Electricity demand is very fickle! Think of your own home energy usage: at any moment, you may turn on a light or turn off an air conditioner. Millions of premises taking similar actions requires a lot of micro adjustments in the supply needed to power them in aggregate, and as a result, the price of energy fluctuates constantly as a function of available supply against demand. In ERCOT, this settlement period is fifteen minutes.
Given the massive volatility in real time energy markets, how could anyone get accurately billed for power usage before smart meters existed? The answer is that they could not. When the only data point you have for premise-level energy consumption is an aggregate value over a month, the solution was (and still is, in some markets!) to take an archetype “load profile” for the building, pre-ordain synthetic intervals for this imaginary building, and scale actual building usages up and down according to their load profiles. Armed with these synthetic interval values, the premises may be reconciled against real time prices and energy consumption can be settled.
While this gets you the ballpark estimate of costs incurred for power, this load profiling method is ultimately not accurate at all. There are only a handful of crude load profiles in existence (think residential, commercial, large industrial, etc) and actual energy usage patterns are highly idiosyncratic. They may look nothing like the assigned load profile whatsoever. Smart meters allow us to ditch the load profiles and know exactly what the energy usage was at any given time, and fully reconcile energy usage with true real time prices at the time it was consumed.
In both cases, the actual consumer doesn’t generally see the difference. Residential consumers typically pay flat-rate bills, regardless if they have a smart meter or an analog meter. Consumers are not typically exposed to real time prices. The exposure lies with the energy retailer, who charge consumers some amount of margin for the service of assuming real time price risk on consumers’ behalf, as well as socializing any potential downside that may be incurred with poor risk management. The retailer is billed by the market operator for consumer energy consumption; consumers pay retailers flat rates and get stable monthly bills; the retailer makes money in exchange for managing market risk (eating losses and pocketing gains); generators get rewarded on the other side for making large illiquid capital investments and providing marginal power; on top of it all, the grid operator uses price signals as one of many inputs to run an efficient and reliable grid. It’s really beautiful capitalism when it works! Of course, except for when it doesn’t.
What I described above is the traditional business model of energy retail. But it doesn’t have to work this way! Smart meters enable a model called Time of Use, whereby the retailer assumes no market risk, passes all real time rates directly to consumers, and charges a very small flat fee for the service. This is the business model of Griddy. The premise is simple: since energy retail is generally a profitable business, consumers must be overpaying in some way. After all, the retailer must make a margin between COGS and revenue, and the average consumer must be paying that margin. Many studies have shown that indeed, when averaged out yearly, consumers pay about 30% less if they just pay real time rates for power, with lower bills most of the time and higher bills some of the time.
If you pair a retail energy provider like Griddy with an advanced suite of home efficiency technology, like programmable or price-responsive thermostats, you can have a building that automatically tailors its energy usage to the market environment, consuming more when prices are low and less when they are high. If every building on the grid acted this way, it would have a material effect on systemwide energy usage. I believe we ultimately need some degree of real time price exposure for consumers if we are to induce these kinds of widespread energy efficiency investments, because it can and does encourage significant reductions in individual electricity consumption during times of peak grid demand, which is good for the grid and good for the planet.
But there’s a catch. And we experienced it this week in ERCOT. When you are exposed to any kind of risk, you must consider tail events: the risk of a three, four, or five sigma outlier event in the tail of the distribution of probabilities. An event like we had in ERCOT this week was not normal, but the risk of it occurring was non-zero. And when you pair a nominally superior steady state average with a non-zero probability of an adverse tail event, sometimes it blows up in your face. A consumer on a Time of Use plan like Griddy is explicitly making the following bargain: they agree to pay lower nominal prices in exchange for the low (but non-zero!) risk that they might occasionally pay outlier high prices. And outlier prices, to put it lightly, are exactly what some incredibly unfortunate consumers experienced this past week.
It’s important to note that most electricity consumers are not on a Time of Use plan. Griddy only had around 25k customers prior to the blackouts. In a region with 11.7 million electricity consumers, this is tiny fraction, and a retailer with only 25k customers is a small player overall. So, while the headlines are likely true for those specific Griddy customers, know that they do not represent the average Texan, and know also that these people made the explicit choice to bear market risk through Griddy. Most Texans do not do this. There are a plurality of traditional electricity retail providers in heavy competition for Texans to choose from. Those that chose Griddy, chose higher risk, which comes with high potential rewards but also high potential losses. Without proper home automation and adequate market knowledge, it has almost certainly been highly inadvisable to chose Griddy, but presumably, these consumers knew the risks and believed they were acceptable.
It’s also important to note that while most consumers didn’t get financially hit, many retailers (who do bear market risk), did. If any electricity retailer wasn’t properly hedged before the prices hit the cap, they were on the hook for their customers load, which they must now pay real time prices for. This dynamic resulted in a lot of odd behavior leading up to the event whereby companies who normally aim for as much growth as possible were begging their customers to switch providers. The rationale is simple: these retailers were anticipating a higher load obligation then they had previously hedged against, and the cost of re-acquiring new customers was lower then the anticipated costs of paying real time market prices at the cap for the electricity these customers were likely going to consume. I’ll take this opportunity to plug my company, Amperon, which partners with all types of market participants as a comprehensive energy market risk management solution. We use smart meter data to provide highly accurate demand forecasts and load analytics as a tool to defend against these types of events and ensure for a resilient and reliable grid. We’re active in seven global energy markets and growing fast, and we’re always hiring, so if this sounds exciting to you, please get in touch!
So what happens now? In the short term, the market will correct itself: Griddy will probably lose all their customers, the vanilla Time of Use model will fall out of favor, and consumers will go back to traditional energy retailers offering flat rates, with perhaps a bit of blunt innovation (think free nights and weekends). Separately, it’s likely Texas consumers will see across-the-board increases in rates, as retailers contend with the dual reality that Texas can peak during winter and summer, combined with the realization that natural gas might not be the reliable base load we all thought it was. This last point will certainly reverberate deeply across power markets worldwide as we all prepare for widespread deployment of renewables.
However, we should really think hard about how to add a bit of a safety net to allow Time of Use in the future, before abandoning it entirely. It has far too many benefits, and in my view, it’s the only way to incentivize a material amount of energy efficiency across the entire system. Consumers who aren’t exposed to price risk have no fundamental incentives to conserve energy in times of need. The problem is that any modifications a market operator might make to price dynamics has much more profound implications than the experience of a handful of consumers. We’re also talking about the long term investment incentives of generators, many of whom base entire investment propositions around capturing those several hours a year of $9k prices. And it’s not just about natural gas: those prices help make wind and solar farms extremely profitable as well, which we want to encourage.
I believe there needs to be reckoning, designed for the extreme cases that we saw this past week. This week, we saw four straight days of $9k prices. After the first few hours, it’s clear that those high prices weren’t doing their job. They didn’t incentivize additional generation to come online and produce power, because the generators were out of commission due to frozen gas lines and underinvestment in winterization. No amount of higher prices in this period could have helped Texas in its moment of need. And the remaining generators that did remain operational captured a grossly outsized amount of return with respect to the volume of power that they actually were able to generate.
It’s a good thing that these remaining generators managed to stay online, and they should be duly rewarded for the investments they probably made to ensure proper winterization. But, in my opinion, the amount of money they are due as a result of this market design ($50 billion!!!) is truly staggering and disproportionate to the value created. And that’s assuming they can even take real credit for staying online, as opposed to getting lucky the particular gas pipeline that fed the plant didn’t freeze. Additionally, these generators didn’t require maximum prices to stay incentivized throughout the ordeal. They would have produced power as long as the price of electricity was higher than the marginal cost of fuel. The fuel costs did soar — but for the most part, not high enough to warrant the sustained prices over the duration of the blackouts.
What’s the answer here? How do we solve for a reasonable amount of risk borne by individual consumers while not destroying the market incentives for reliable generation? Some folks will argue for a capacity market, a discussion which warrants a much deeper treatment. I personally don’t think a capacity market would have prevented this, and I believe we need supra-market legislation mandating higher reliability standards for generators, as well different formulas for real time price settlement beyond a simple balance of supply and demand, expressly conditional upon the absence of widespread blackout regimes. In an emergency order made on February 16th, the Public Utility Commission of Texas (PUCT) acted in the complete opposite direction, directing ERCOT to raise prices to the cap during the blackouts:
I think this decision needs to be severely revisited, given the higher prices did nothing, in my opinion, to hasten the end of the blackouts. In the future, maybe they will act as a good price signal for new generation to get built, wanting a piece of future action. I’d preemptively argue that this is far too high a price for future maybe-reliability, and I’d also argue that the original case for $9k prices almost certainly did not model a continuous four day period at the cap. Either way: these extended price cap hours did nothing marginally for this blackout, and the disaster only ended when temperatures rose high enough several days later for the water in the pipelines to thaw, allowing the natural gas plants to fuel up and become operational. Dumb weather luck is not exactly the best allocator of capital!
The point is that while Griddy customers should definitely be entitled to relief, Griddy is not evil. Nor is ERCOT. The premise behind Time of Use is that consumer exposure to real time prices can and should ultimately result in cheaper electricity for everyone. I’ll also remind the reader that ERCOT is perhaps the most transparent and functional grid operator in America. ERCOT has smart meter penetration rates of nearly 100% (in comparison to around 60% elsewhere), it leads the country by far in wind power production (hitting a record of 22GW at 60% of the power mix just this January), there is an incredible amount of data available to both consumers and market participants, it leads the way in initiatives like Smart Meter Texas (which allows third parties to access meter data in ways that leapfrog similar utility-level Green Button deployments), and there’s a wide arena of retail electricity innovators offering reliable and cheap power to all types of electricity consumers in Texas. In my opinion, Griddy has been a very valuable experiment, and it could only have been built in ERCOT.
Taking a broader view, this outage simply cannot be blamed on a single player. It was a systems level failure. For better or for worse, it probably wasn’t helped by how quintessentially Texan the whole thing is, with a heavy emphasis on deregulation and the deliberate way the grid doesn’t cross state lines to avoid FERC influence. Perhaps this is why ERCOT has thrived to date and achieved what it has, but in this case, it prevented power imports from other interconnections which could have lessened the impact of the electricity supply deficit, and perhaps also there are existing FERC regulations governing reliability standards for thermal generation that ERCOT could have abided by.
I will refrain from postulating further here until the deeper investigation comes out over the next few months, but the factors at play were certainly complex, interacting, and at the very end of the day — this was an extreme winter weather event inflicted upon a grid designed for summer peaks. We may debate the exact market and legal mechanisms to prevent it from happening again, but even if we figure those out, it won’t change the fact that we are ultimately in the first innings of a new climate era defined by extreme and unpredictable forces of nature. This time, it was a polar vortex in Texas. Last time, it was a continent-spanning wildfire in Australia. Next time, maybe it’ll be a Category 5 hurricane in Maine, or another dust bowl in the midwest, or a tornado in New York City. I’m not a meteorologist, but you get my drift: climate change is here, folks. This is what it looks like.
So, back to the grid. If we want to enable a true smart grid and build for resilience in the face of this future volatility, we need to allow for proportional returns to generators and market participants, without sacrificing grid reliability on the supply side. And on the demand side, we need to incentivize consumers to consume less power as a crucial weapon to stem the tide of climate change, allowing real time market exposure through Time of Use plans but without the possibility of the insane consumer bills that we saw this week, which can absolve the need for additional peaker generators in the first place. I hope legislators, the PUC, the ERCOT board, and all other RTOs around the world act thoughtfully and accordingly in simultaneous pursuit of these goals.
The opinions stated in this piece are my personal views, and do not necessarily represent the views of Amperon. Thank you to Sean Kelly, Kevin Stevens, James McGinniss, and Elliott Chorn for reviewing drafts of this post.