Context

• Hyperscalers need enormous amounts of power ASAP to power AI datacenters. While highly speculative, estimates range from ~30-300 GW by ~2030.

• Speed is paramount. Procuring huge amounts of energy fast is a strategic imperative for hyperscalers and is of national security importance as well.

• Grid expansion today in the US is slow. While emerging clean firm generation technologies may promise eventual integration with datacenters, these technologies are unlikely to scale fast enough to meet enormous near-term datacenter energy needs. Many believe the AI needs in the US will most likely be met by building colocated natural gas power plants.

• While climate is not the top priority of hyperscalers today, building more natural gas would further increase emissions, which the US is trying to reduce. It would be preferable if we could find ways to build more energy without emitting more, without sacrificing speed or meaningfully increasing cost.

Key questions

• Could off-grid solar microgrids in the US be big enough, fast enough, and cheap enough to be a compelling near-term alternative to building more natural gas power plants to meet near-term AI energy needs?

• If yes, what would that look like and how would one do it?

What we did

• Scale Microgrids ran 20-year powerflow models for thousands of site configurations supporting a 100-1000 MW 24/7 load and ran them them all through Lazard's LCOE model¹ to find the best performers. Paces then did a search for all the land in the US Southwest that could accommodate them (filtering for things like distance from natural gas pipelines, property owner type, minimum parcel size, etc.).

• Download the full LCOE workbook developed by Scale Microgrids to play with assumptions.

Key findings

• Off-grid solar microgrids are categorically faster than new grid interconnections (5+year queues) as well as off-grid colocated gas turbines (3+ year lead times).

  • Estimated time to operation for a large off-grid solar microgrid could be around 2 years (1-2 years for site acquisition and permitting plus 1-2 years for site buildout), though there’s no obvious reason why this couldn’t be done faster by very motivated and competent builders.

  • While solar scenarios modeled in this paper also include 125% capacity of gas engines, these are easier to acquire and faster to build than large dedicated, always-on gas turbines.

  • Rental generators—such as those being used by XAI prior to getting a grid connection—are potentially the fastest ‘absolute’ path to power but their limited availability would be quickly exhausted by large-scale adoption (in addition to their high costs at >$300/Mwh). These can be utilized before a final project is commissioned for either off-grid solar microgrids or colocated gas turbine approaches.

• Off-grid solar microgrids today are near cost parity with natural gas and cheaper than other clean alternatives. Opportunities for further cost reduction are significant.

  • A microgrid that supplies 44% of lifetime demand from solar and includes 125% natural gas backup is approximately the same cost as using large off-grid natural gas turbines ($93/MWh versus $86/MWh).

  • A system supplying 90% of lifetime demand from solar is cheaper than repowering Three Mile Island ($109/MWh versus $130/MWh).

  • Further power system design optimization enabled by off-grid solar could reduce cost by another 10%+.

• Off-grid solar microgrids are enormously scalable, with >1,200 GW of datacenter potential in the US southwest alone.

  • This is enough suitable land to cover ~4-40X all of the datacenter growth projected in the US through 2030 (~30 to 300 GW of new development by 2030).

  • The vast majority of this is in West Texas, which is mostly a function of gas pipeline density (if you relax this constraint by using diesel backup generators instead, you can build almost anywhere with good sun). And, >95% of the land is private, making it commercially viable today.

  • Paces has a database of exact parcels that meet these criteria with contact information for landowners. Further design optimizations could further reduce land requirements by ~25%.

• Between 0.4 billion tons (30 GW new datacenters) and 4.1 billion tons (300 GW new datacenters) of CO₂ emissions could be avoided between now and 2030 if every new AI datacenter was built using the 90% solar configuration. At approximately $50/ton CO₂ reduced, replacing gas-powered datacenters with off-grid solar microgrids represents a cost-effective emissions mitigation strategy.

Increasingly, this is what buying electricity in the US may feel like.

Your electricity bill has two elements: the cost of generating power and the cost of delivering it. People typically assume their bill is mostly or exclusively driven by the former. Far off power plants produce energy, they sell it into the wholesale market, and those costs eventually get passed on to consumers.

In reality, the cost of building and maintaining all the infrastructure to get power to your house (transmission lines, substations, distribution networks, transformers, etc.) is a significant portion of your bill. In fact, when working with large energy users across the country, I even see utility bills where a majority of costs are associated with delivery.

And one of the most slept on trends in the energy industry is that this portion is growing rapidly. According to the US Energy Information Administration,

“After adjusting for inflation, major utilities spent 2.6 cents per kilowatthour (kWh) on electricity delivery in 2010, using 2020 dollars. In comparison, spending on delivery was 65% higher in 2020 at 4.3 cents/kWh. Conversely, utility spending on power production decreased from 6.8 cents/kWh in 2010 (using 2020 dollars) to 4.6 cents/kWh in 2020.”

Put simply, generating power is getting cheaper while delivering it becomes more expensive. In this post I will show how the utility business model creates this situation, how it can meaningfully slow the energy transition, and why DERs must challenge the monopoly if we want to fix it.

Why’s this happening?

Generating power has been getting cheaper for the last 20 years. In the early 2000s we built a truly massive fleet of natural gas power plants, just in time for the fracking revolution to unlock enormous quantities of low cost fuel for them.

Also, wind and solar began to get super cheap, which led to rapid deployment. Even at relatively low market shares, wind and solar suppress wholesale energy prices because they have zero marginal costs. And thanks to incredible learning rates for these technologies, we can only expect this to continue.

But cheap gas, wind, and solar are not the full story — simultaneously, delivery costs are also increasing. Exactly why this is occurring is hard to succinctly answer. Experts will cite many factors, including replacing aging grid equipment, implementing new smart technologies, and building new transmission lines to access the best locations for wind and solar power.

However, there is another explanation of these costs that is more controversial among experts. Power delivery infrastructure is the last bastion of the electricity monopoly, and from it, utilities intend to grow.

For context, here’s a quick history of electricity regulation:

• Historically, the generation and delivery of electricity was a monopoly granted by the state. Utilities built everything needed for us to have electricity, and were allowed to earn a regulated and guaranteed return on all of that capex.

• During the 1990s, energy market liberalization swept across the nation. Under this new paradigm, utilities were removed from the power generation process. While building and maintaining the power delivery system was considered a natural monopoly for which the traditional regulatory model still made sense, generating power was now conceived as a competitive market, where many players would compete to drive down the commodity cost of electricity. Note that, this only spread to about half the country.

• In 2005 electricity demand growth stopped. After decades of consistent growth, energy efficiency and industrial offshoring kept electricity demand flat. This meant utilities couldn’t build more stuff for the purpose of delivering more power.

Since utilities couldn’t make money on power generation any more, and there was no new load to justify new delivery infrastructure they could earn a return on, the utility business pivoted to deploying more and more capital into delivering power for existing demand. This resulted in an increasing unit cost of electricity delivery.

To be fair, there are good reasons to deploy more capital into delivering power for existing load. Smart grid technologies like advanced metering infrastructure (aka “smart meters”) cost money. However, if used well, these technologies also should help reduce peak demand on the grid, meaning future load growth can be served with less new delivery infrastructure. Therefore, the utility’s natural incentive is to convince their regulators to allow for deployment of this smart technology, which is capex they can receive a return on, but then not fully utilize it. That way, any new demand which may materialize can still be served with even more capital investment.

In short, the utility’s incentive is to deploy capital inefficiently, and all of that effort has been almost exclusively focused on delivery infrastructure for decades.

Decarbonization is levered to cheap electricity

There are obvious downsides to electricity becoming more expensive. For industry, high prices hurt the bottom line. For residential consumers, one can be forced to choose between a credit card balance and keeping the heat on. But within the more specific context of electricity generation becoming cheaper while its delivery cost increases, the impact on electrification should be considered.

The term electrification is used to describe fuel-switching a fossil-fuel energy demand to electricity. The two most common examples of this are electric vehicles and heat pump electric heating. Both are considered by most in the energy sector to be essential parts of decarbonization. The idea is that if we switch as much energy use to electricity as possible, then unlike when using fossil fuels, we can clean up the electricity source.

One reason electrification is supposed to be economical (in addition to sustainable) is that making power through renewables and batteries is getting cheaper by the day. As the power system converts to these sources, electricity prices should come down, making electrification economics more attractive. Or so the theory goes.

But if delivery costs continue to increase at the current pace, they’ll eat any savings generated by large-scale solar and wind. This is a problem for decarbonization, when around 130 million Americans live in a state where replacing an old gas furnace with a heat pump would create higher energy costs than simply buying another gas furnace (ignoring the higher capex as well).

And this doesn’t just impact residential home electrification. Buried in the business model of virtually every new and exciting decarbonization pathway (direct air capture, H2 for steel and shipping, e-fuels for aviation, etc.) is the assumption of very low cost electricity inputs made possible by renewables. You can listen to this episode of The Interchange for a deep dive on the subject, but the quick version is if that cheap renewable energy input is to be accessed via the grid, then delivery costs once again become a problem.

If we want rapid climate action the cost of electricity needs to fall, but its delivery costs may get in the way.

Load Defection? More like load optimization.

Another outcome of rising delivery and declining generation costs is electricity users will try to avoid those delivery costs. This will be accomplished by building on-site power systems or migrating to locations adjacent to existing cheap energy generation.

The former is becoming commonplace. Homes and businesses are installing on-site solar and storage at a rapid pace, and while they may not realize it, they are doing so to avoid delivery costs. The economics of this choice are often superior to signing up for a grid-delivered power contract because each unit of energy generates value at the retail rate (inclusive of delivery costs) rather than wholesale rate. And as the march of technology learning rates continues, the quantity of locations where this choice makes sense will grow.

Simultaneously, load will begin to migrate to places where large wind and solar projects already exist, in attempt to offtake power directly. This has started to happen with new industrial energy loads like hydrogen production and crypto-currency mining because they are extremely sensitive to the cost of power. Soon, other new-build industrial sites will do the same. And eventually, when this dynamic becomes compelling enough, businesses will even consider moving from their existing facilities to capture low cost power directly from large-scale energy sources. If you can avoid the distribution grid, you will.

In the electricity industry we refer to this as load defection, for which the primary concern is the so called utility death spiral. If load defection accelerates, delivery utilities will experience significant decline in energy sales, necessitating rate increases to pay for fixed infrastructure costs. This leads to more load defection, and the cycles continues. This is considered problematic because it will leave those without on-site power systems shouldering the cost of the delivery infrastructure we all (including on-site energy users) rely on.

This argument has been used to levy prejudicial fees on customers with on-site generation, and while this may seem logical, it ignores a crucial consideration; short-run versus long-run costs.

It is true that existing grid delivery assets are fixed costs. Reducing energy consumption from the utility doesn’t reduce the cost of that infrastructure. Once it exists, we’re stuck with it. However, reducing energy consumption (at the right times, more on that later), absolutely does decrease the need for future delivery infrastructure. If the delivery infrastructure needs to support “x” peak load today, and 1.5x at some point in the future, then a 0.25x reduction today means we only need to support 1.25x in the future. While it doesn’t translate immediately into delivery infrastructure savings, it most certainly will over time.

This matters because, while load hasn’t grown for 15+ years, it is about to explode. Earlier I mentioned electrification to show how increasing electricity delivery costs are a problem for decarbonization, but what I didn’t describe was the extent to which this will transform our electricity system. A study from the National Renewable Energy Laboratory (NREL) found high levels of electrification (but not even full electrification) will result in the power system requiring between 2 and 3.5 terawatts of generation capacity, relative to our 1.1 terawatts today.

Doubling or tripling our generating capacity will require a ton of new delivery infrastructure. It follows that reducing load on-site with distributed energy will absolutely reduce total delivery system costs. In turn, this means distributed energy will help keep delivery rates low for everyone, which is a very different outcome than the dystopian utility death spiral narrative some would lead you to believe.

Load defection is good for the grid, because it makes room in our existing delivery infrastructure to accommodate electrification. That results in a more optimized, cheaper power system for all, while making successful decarbonization more probable.

Virtual wires, the next frontier of liberalization

Earlier we established that the core monopoly utility incentive is to deploy capital inefficiently. With this in mind, utilities aren’t an ideal choice to own or operate the distributed energy technologies that will keep delivery rates low in the face of huge new load growth from electrification.

Instead, a variety of market participants should develop these solutions and be financially exposed to their performance. For example, what is the best way to deploy a smart thermostat to shift demand patterns? If the utility installs it, or even just controls it, chances are it will never be properly utilized. But when an end-user or their company of choice installs it, because they’re taking the risk on it working, you can be sure they’ll do everything they can to achieve the best outcome.

Consider that about a decade-ago the utility sector and its regulators began talking about Non-Wires Alternatives (this concept). Over that time, a few pilots have taken place, and much ink has been spilled about them, but no sincere effort has been made to turn this into a scalable approach for managing grid infrastructure costs. The common thread across all of these pilots was utility management of the programs.

The case for this is no different than that which led to wholesale power markets. Providing the service of avoided distribution infrastructure, or virtual wires, is not a natural monopoly, so it would be sub-optimal to treat it as one.

One key challenge to achieving this future will be designing the right price signals to properly compensate virtual wires for the value they offer. Options include incorporating these future charges into electricity bills and letting users install DERs to reduce them (rate design), implementing programs that compensate virtual wires services providers directly via the rate base (a DSO type market perhaps?), tariff programs that transact through the utilities (like NY’s VDER), and likely many other things. Whatever the best price signal is, we should strive to make it clear and standardized across the country. If every regulatory fiefdom takes a different approach, we’ll wind up with high transaction costs and providers that can’t scale.

A second challenge will be that utilities likely won’t appreciate this vision. The existence of virtual wires is conceptually (not legally) a challenge to their franchise right, the state-granted monopoly for building electricity delivery infrastructure. This will invariably lead to policy battles that force us to reckon with the fundamental nature of the monopoly regulatory model. And this battle probably wont be fair, given when backed into a corner utilities are known to fight in distasteful (astroturfing) or even illegal (bribery) ways.

Despite these challenges, managing the delivery cost problem and the system change implied by its solution will be a major theme as electrification grows. If we want a power system that is prepared for the challenges of the future and supports human prosperity, we must be successful in this endeavor.

In six months we’ve seen two black swan weather events cause an energy crisis. As we reckon with the systems engineering impact of extreme weather, energy resilience will undoubtedly emerge as an essential underpinning of our future grid. 

The first of these events was in August of 2020. A multi-state heat wave led to blackouts in California when the power system operator couldn’t secure adequate generation to meet immense air conditioning loads. Temperatures exceeded 121 degrees in LA County, the highest ever recorded. For reference, the most conservative value offered by the American Society of Heating and Cooling Engineers (ASHRAE) for designing air conditioning systems in the area is 91.4 degrees, which is supposed to cover 99.6% of hours in a year. The 50-year historical data set that recommendation is computed from has a absolute maximum value of 110.8, meaning what happened in August was, for all intents and purposes, essentially deemed so incredibly improbable it wasn’t worth considering. The 28.6 degree delta between the conservative design recommendation and the actual recorded August 2020 temperature is the difference between a nursing home’s air conditioning just barely keeping up and utterly failing. 

The second event is happening as we speak. A massive winter storm is delivering crushingly low temperatures across much of the US. Texas, most of which currently does not have power, is being hit the hardest, not necessarily in absolute temperature, but relative to their typical winters. Monday hit a high of 15 F in Dallas, whereas generally you’d expect around 55 F during this time of year. And Monday’s low of 5 F is extremely cold for a city with an ASHRAE 99.6% heating design temperature of 24.6 F. Even those who haven’t lost power and have functioning heating systems are still likely quite cold, while many homes are simultaneously facing electricity, heating, water/sewage, and telecommunications outages. And as of Tuesday, outages began to spread to neighboring grids such as the Southwest Power Pool and Midcontinent Independent System Operator.

While it's easy to see why an HVAC system wouldn’t hold up during either of these conditions, exactly why this extreme weather led to energy crises isn’t as intuitive. In each case there were a multitude of contributing factors ranging from engineering, to operations, and even communication issues, all of which are likely to be debated for some time. 

For example, right now in Texas many wind turbines are frozen and unable to spin because their developers didn’t purchase the cold weather package offered by manufacturers. Yet it turns out that this was a very small factor in the blackouts; many more conventional power plants also can’t cope with the cold, leading to a far greater contribution to the blackouts. Additionally, natural gas is being prioritized to heat people’s homes since it is so cold, which leaves the remaining natural gas power plants unable to procure the fuel they need to operate. The subsequent rolling blackouts required to keep at least some of the grid operating then in turn leave those with electric heating systems in the cold. Trying to keep people warm results in making other people cold. Confusing, right? 

As of Monday the Texas grid had over 30 GWs of expected power generation that didn’t deliver, mostly conventional plants. For scale, the entire country of Spain’s peak electricity load was 40 GWs in 2018. An entire country’s worth of power plants are shut down in one US state because it’s colder than we expected.

It is also important to note that these are not the only recent system failures from extreme weather. Generally, the term energy crisis is used when there is not enough supply to meet demand, which was the case for the two events above. But we have also seen many typical outages resulting from failures in the transmission and distribution system, where we had enough power but we just couldn’t get it to people. For example, in July of 2020, Hurricane Isaias left millions without power, for some lasting over a week, in the middle of a pandemic. 

Observers of this situation can have nuanced debates about it’s implications: does this show the harsh side of deregulation, should gas plants be considered unreliable, or was Rick Perry’s dream of giving coal plants credit for their on-site fuel storage not as silly as many thought? There’s plenty of important and interesting post-mortem analysis to be done here, but there is one key consideration that ought to dominate the conversation following this disaster. 

The bottom line is that complex systems often aren’t great at responding to events that weren’t contemplated in their design. When stressed beyond its capabilities, failure in one component can cause cascading failure throughout the system. And as these systems become both more complex and exposed to more extreme conditions, this risk intensifies. We have a century’s worth of infrastructure that is simply unprepared for a changing climate. It is going to fail more often and more spectacularly. 

With this in mind, resilience will be a critical feature of energy systems moving forward. Our grid must be able to withstand more extreme conditions, while delivering more services (the electrification of heat and transportation), and it must be able to bounce back when things go wrong.

And this issue is not just one for the power system to deal with. Yes, rolling blackouts have swept across Texas and power prices are pegged at $9000/MWh, but natural gas is also in short supply and prices are through the roof, oil wells are frozen and inoperable, a nuclear plant has shut down because its turbine was built without protection from the elements, and the largest oil refineries in North America are closed -- all simultaneously and all due to one extreme weather event.

All of our infrastructure— energy systems, water treatment plants, buildings, highways, ports, bridges, telecommunications, etc. — is not ready.

But why haven’t we already addressed system resilience? After all, we’ve known about climate change for decades now. Beyond the general confusion and polarization of the topic, why aren’t the alarm bells ringing? I believe there are two causes of this: we have trouble imagining the need for resilience in the near term and we’d like to pretend we don’t have to deal with its cost.

There are two broad categories of climate action: mitigation and adaptation. Mitigation involves eliminating our emissions so we don’t cause more climate change, whereas adaptation describes improving our systems to respond to the level of climate change that ends up occurring. For good reason, mitigation gets a lot of attention. We need to reach net zero emissions to deal with climate change.

But adaptation, for which improving resilience is the largest element, has taken a back seat. I believe this is because climate change feels far off. When we contemplate the impacts of a warming world and how it will stress our systems, we imagine worst case scenarios 150 years from now where the earth is unimaginably different from today. However, we tend not to consider all the ways society will be impacted by even slight climate change. Everything we’ve built our society on top of has been predicated on a stable climate, a narrow band of weather conditions, and even small changes can lead to dramatic outcomes. There are millions of Texans who now know this with first hand experience.

Additionally, even if one recognizes the urgent need for resilience, it can be difficult to discuss because addressing it carries a cost. Consider how the narrative surrounding mitigation has changed in recent years. Now that solar and wind are the cheapest form of new electricity, suddenly there is immense support to build more of it. It is far easier to pitch emissions reduction when it also carries a cost reduction. But with resilience, this often isn’t the case. It is intuitive that a stronger power system built to operate through more extreme conditions may also cost more, just like a four-wheel-drive car costs more. But what we conveniently ignore is that unaddressed resilience needs, across all of our infrastructure, will create a far higher societal cost.

Luckily, there is reason to have hope. In subsequent posts I will describe how, alongside other grid hardening activities, distributed energy resources are poised to address this massive challenge and deliver immense societal benefits, and it can actually save us money if we make the right investments. A few recent developments in the distributed energy ecosystem make this incredibly clear and offer a transformative solution.

In the meantime, let’s all follow the situation in Texas closely and do what we can to help people get through this difficult time.