What To Do if Energy Gets Cheap?

Mackenzie Morehead

09 Dec 2025 • 18 min read

Solar has long been marching down an exponential, improving at a learning rate of 25% per doubling for 70 years. Their cost has now fallen by 1,000,000x cumulatively. A scarce few physical technologies like transistors can boast such durable scaling “laws”.

And yet, the global LCOE for solar seemingly started to level off in the last five years and PPAs in the U.S. and Europe for solar have risen by 50-100% since 2019. The rise of electricity prices has ignited the public consciousness to the point that for the first time in decades it is deciding congressional elections.

Which diametrically opposed story is the right one? Will naive extrapolation of solar learning curves continue to outperform overly-involved projections, and we get another 50% reduction in module cost within 5 years which could possibly yield a US LCOE of ~$25/MWh or a minimalist off-grid array at ~$15/MWh? Or will electricity prices continue to march upwards? What are the most interesting companies to build in the respective scenarios, where, and how?

To start to answer those questions, we begin exploring why exactly have costs come down by so much historically?

These resources do a wonderful job of detailing the technical changes that drove cost declines and the history surrounding them. But for a snap-shot understanding see the excerpt below showing the ranked order drivers from 1980 to 2012. Since 2012, each of these trends have continued but the physical scaling of plant & project sizes have been most dramatic.

Why did LCOEs & PPAs stall in the last five years?

While we couldn’t find any academic papers precisely allocating costs, the NREL researchers we spoke with that produce the main industry reports and forward projections reaffirmed the intuition that the biggest drivers have been interest rates, tariffs, and supply chain disruptions (not necessarily in that order). Wood Mackenzie estimated that a +200 bp increase in the risk-free rate raises renewables’ LCOE by ~20%. Meanwhile, this study’s model implies U.S. wholesale panel prices were ~10–35% higher because of tariffs after 2012.

And starting next year, solar is no longer subsidized so that will hike the price another ~30% in the U.S.

What are the arguments against the cost curves continuing?

The module costs are now at commodity prices and only a relatively small fraction of the total system cost. The remaining costs like BOS and soft costs, which have been the hardest to bring down and only contributed 13% of cost reduction relative to hardware. The remainder of the cost is relatively spread out across the stack, especially on utility projects such that pushing on any single vector won’t have a night-and-day impact. For instance, automating installation and thereby cutting fieldwork in half would result in ~10% lower capital costs.

Meanwhile, the industry is in the part of the capital cycle characterized by oversupply with current capacity double demand, factory utilization of <50% and as low as 20%, obsolete factories still operating, and inventories built up to 400,000 MT. In fact, in the aggregate the solar industry doesn’t currently break even. Moving forward, industry consolidation and / or regulation is likely. 33 of the top Chinese producers have already agreed to cut output.

Finally, the manufacturing process is already quite efficient with prices at commodity level so it’s not immediately obvious that exponential cost declines can continue indefinitely. Per a back of the envelope calculation, the idiot index of solar manufacturing in China is roughly 1.

Why we remain confident over a 5-10 year timeframe

We don’t have confidence that prices at least in developed countries will be materially lower in the next few years due to our analysis that prices are artificially deflated due to capital cycle dynamics, that the macro headwinds that have pushed up costs since 2020 are unlikely to meaningfully reverse for at least several more years, and that moving further down the cost curve will be harder than it has been in the past.

With that said, over the medium- to longer-term, we see the continued room for efficiencies in production processes washing out these immediate factors. The latest ITRPV report that records the views of many global experts makes clear that there are still countless innovations to production processes yet to be rolled out that are expected to be scaled over the next 10 years as sampled below.

Moreover, a relatively automated, fully-integrated supply chain in the US is starting to come online, which should help lower our price differential down from 2x the world average.

Meanwhile, companies we’re thrilled about like Planted Solar and AES have real shots at industrializing installations, meaningfully cutting down soft costs.

All told, we don’t expect another 10x decrease in costs but wouldn’t be surprised by 40% lower LCOE for grid-connected utility projects in the U.S. and 60% lower for maximally minimalist off-grid arrays within 10 years.

Even if solar’s cost curves don’t continue to march downwards, hundreds of startups are trying dozens of fundamentally different approaches to energy generation. Over a dozen claiming <$30/MWh LCOE at maturity. Even when seriously discounting the claims of each individual company, there’s still a solid chance at least one will come true.

(P.S. if you’re building an energy technology with a fundamentally differentiated approach, we’d love to talk.)

One cannot overstate the implications of such possibilities. Humanity has never in its history had its most fundamental input be a technology that continuously gets ever-cheaper.

Where and How to Build

If these energy technologies do get cheap, how should one build a company that harnesses that trend?

Connecting to the grid likely won’t make sense. While you save the capex, the grid’s energy prices are only loosely connected to the lowest cost generation and will always significantly lag purpose-built new generation. Incremental grid costs will long be dominated by transmission and distribution costs, not generation. Moreover, wholesale prices are set by the marginal generator (i.e. natural gas) not by solar, and solar-battery hybrid plants won’t be materially cheaper in competing for that flat block for a while. Indeed solar is already seeing its value to the grid decline at <30% market share.

Instead, a startup should likely build its own maximally simple array, at least in the longer-run. Whether or not to redesign its industrial process around intermittent energy and forgo the battery capex largely depends on how much it can shrink the capex factor.

The even more interesting question is what to build.

What are the Most Compelling Things to Build with Energy Too Cheap to Meter

As the fundamental input to all downstream economic activities, cheap energy of course benefits society and specific businesses in innumerate ways. We narrow down from everything to businesses most levered to the cost of electricity. Commodities generally have energy as the largest fraction of operating costs.

We at Compound have an especially high bar for such investments, looking for:

Genuine technological breakthroughs as venture markets are willing to extrapolate one datapoint to many future ones so they will pay significantly up for scientific outcomes, not execution
Sizable and durable differentiation on the key vector of competition (usually cost in commodities and not sustainability alone)
A business model built around that tech that’s compelling or novel to the industry

Some high-level frameworks that guide our thinking in these markets include increasing the product’s embodied energy, using DePIN business models when applicable, being as close to the end application layer as possible, starting with premium markets, physical technologies that scale well, and the ball-bell curve of energy technologies.

Below are some of the specific areas we’re excited about that have a chance to disrupt large industries purely from economic perspective even without a fall in electricity but that would be most beneficially levered to that outcome.

SOECs

Among electrolyzer technologies, solid oxides are the most thermodynamically favored. They utilize high operating temperatures to make the chemical transformations require less energy, as described well elsewhere. Their efficiencies are now nearing CCGTs.

Yet, they have received little startup attention because historically their capital costs were prohibitive and the GTM has been weak. Both of those are changing and a new startup could radically push the frontiers.

While competing approaches are highly mature technologies, SOECs’ performance trajectory remains rapid and malleable. Over the past 10-15 years, their efficiency has improved by over 2.5x and long-term durability by ~100x.

There’s clearly lots of white space left both on the cell level and the stack level. We’ve met with teams that claim their new cells are even 2x more capital efficient than the current state of the art. We see no reason why this can’t be pushed further at both the cell- and system-level.

The cells must have maximal surface area which requires extremely porous 3D architectures, high conductivity, and low impedance interfaces. This multidimensional problem may lend well to the cutting edge materials discovery techniques we detailed here.

Meanwhile, the chronic challenges of working with temperatures as high as 800°C results in lots of steel, heat exchangers, and overall high BOP-related capex. Maybe this is just a naive outsider’s perspective, but the SotA as pictured below still appears wildly over-complicated.

Could engineering principles like those used for the Raptor engine design of using few, highly intricate parts be employed? Alternatively, could the complexity be resolved with materials innovation either by developing materials or manufacturing methods that can more readily handle 800°C or redesign the cell to operate with similar efficiency at say 400°C?

A Raptor 1 to 3 type transformation would redefine the business model to one more favorable to startups. Indeed, steel usage largely explains why chemical engineering is a canonical economies of scale business. Capital costs increase sub-linearly with capacity, by roughly (capacity ratio)^0.6. A plant with 2x the output is only 50% more expensive to build and operate.

What makes SOECs interesting now is not just that decades of research have gotten the tech to commercial applicability but that there’s finally material demand pull. Their end use switched from overpriced green hydrogen to the massive, rapidly growing, price insensitive data center market. As the bellwether of this PMF, Bloom Energy has signed $5B partnerships and hit a $25B public market valuation.

While narratives have focused on gas turbines and nuclear, SOECs are also quite well matched to the problem. They can generate electricity at nearly peak efficiency starting at 1KW. Once at 20MW sizes, then CCGTs become an option and boast 64% efficiency vs Bloom’s current 54%. But turbines’ supply chain has years-long backlogs, whereas Bloom deployed cells to Oracle to power an entire data center in 90 days. Finally, because their solid-state nature means no moving parts or mechanical lag, they can respond to demand at least twice as fast as rotating generators. When combined with supercapacitators, they can ramp 100% in milliseconds which matches datacenter’s frequent 20-150% load swings in milliseconds.

A new startup trying to ride the data center wave would likely use a JV business model of licensing their step-function better cells to incumbents like Bloom, Topsoe, etc. While they use their existing manufacturing scale, you focus not only on the novel cell but also stack design.

Alternatively, the cells could be run in reverse to produce chemicals or fuels from electricity. Two other possible GTMs that don’t necessitate physical economies of scale are producing expensive specialty chemicals like Solugen and Circularity Fuels or using the cells in a simplified system for the conversion of flared methane gas into heavier, high value hydrocarbons like benzene. 140 billion cubic meters are flared a year and any solution would have to be distributed. Solid state membranes have been shown to directly convert 20% of the waste stream into high value, easily shippable liquids.

Either way, the longer-term TAM for SOECs include the fuels and chemicals for use cases for which batteries or other simpler chemical processes can’t compete.

At scale, electricity is ~70% of operating costs.

Cultivated Meat

We at Compound are comfortable investing in out of favor sectors and in fact generally won’t invest in highly legible, consensus areas. Few sectors are currently as beaten down as alternative meat and cultivated meat in particular. We view that not as a signal to run away but to dig in deeper – especially when intuitively our future’s most advanced technology won’t be feeding cows 7 calories to produce 1 calorie of usable harvested meat. The margin of that gross inefficiency leaves plenty of room for opportunity.

Earlier this year, we wrote at length about the dozens of Nobel-prize worthy technologies that next-generation biomanufacturing startups can combine.

Since then we’ve met with the most advanced cultivated meat companies and investors to understand precisely where the economics, technologies, and financials of the first generation lies. We were surprised to come away more bullish on a couple of the companies, at least relative to our expectation that none would have plausible paths towards price parity on anything but the most premium products. Though they still need to actually achieve their projected parity costs at their soon-to-be live manufacturing plants, the most fundamental scientific problems like <$1/L growth media and viable cell lines have largely been solved.

With that said, we certainly do not view the field as set in stone. Startups with no tech debt nor a cap table full of beleaguered investors could quickly reach the technological frontier and then carve out their niche. Unlike the first generation, new startups can leverage the knowledge in the published literature, the open sourced cell lines from non-profits like Tufts and GFI, and can initially purchase essentials like <$1/L growth media from existing startups that specialize in selling such infrastructure while developing core inputs internally over time as needed.

With those tools, the right team could get up to speed on a small seed round and then start working towards solving the remaining core challenges in the field. As proof, Clever Carnivore has become a top five competitor off of $9M in funding.

Remaining challenges include:

Bioprocess optimization at large scale: microcarriers and adherent cell processes appear unscalable, leaving most companies to pursue aggregate suspension systems. However, the limited aggregate size forces either very high-productivity batch cycles or semi-continuous or even continuous bioprocessing. Both require combinatorial optimization of or ideally fundamental innovation in aggregate dissociation, size-based bleed-and-feed, or a move to true single-cell suspension (with its own media and transport-kinetic considerations). Moreover, the larger the scale, the less friendly the environment for cells so creativity may be needed to industrialize.
Customer adoption & marketing: existing brands complain about uptake on the currently offered premium products as being slower than desired, particularly for a venture-funded business. We won’t know customer adoption willingness until products are offered at or very near cost parity. But, we are confident that if a team is able to get to a significantly lower price point, then customers will quickly enough get over the phobia of the lab-grown. Price rules all. However, a team with special aesthetic taste that crafts an inspiring brand could possibly allure customers before getting well below cost parity. Their product could even be developed from the start to fit with a more natural vibe by engineering the cell lines to utilize only media inputs that they naturally use for instance.
Cut meats: novel manufacturing techniques like metamaterial-based injection molding or textile-based methods could unlock cut meats like steaks which command 3x premium over ground beef
Making otherwise unmakeable products: beyond cost, the other way to give customers a reason to put some foreign technology in their bodies is to offer them something they need or innately desire but can’t otherwise get. Researchers we’ve talked to engineered red meat free of alpha-gal allergen or cells enriched for vitamin A. We’re excited about the possibility of making beef that doesn’t raise cholesterol, otherwise boosting its healthiness (e.g. ratio of healthy and unhealthy fatty acids), and changing the aroma or taste (e.g. imbuing the rich smokiness of aged Angus steaks in cheaper cuts or even going beyond anything nature’s made).

We’re also excited about startups using the technologies developed from the first generation to solve higher value products like synthetic blood and approaches that use chemistry to make simple fats.

At scale, electricity is ~33% of operating costs.

Metals, Magnets, and Other Minerals

While this category is slightly tricky for us because at this point it’s getting reasonably crowded and highly legible, we remain excited by some ideas like plasma-based manufacturing methods that can process many different materials and those with one-step processes that can radically undercut incumbent processes on price.

With respect to value capture, we’re skeptical that the business of inventing novel metals will create or capture enough value because many superalloys are already off patent, competitively priced, and the companies that benefit most from high performance materials will always hold market power over new producers or will make them internally.

All of the margin is in the vertical integration of end product making of proprietary materials. As proxies for this statement, consider that:

Haynes sold powders of its superalloys and was sold for $1B, while ATI does fully integrated AM and part making of its superalloys and currently trades at 13x Haynes’ acquisition price. This suggests that proprietary materials aren’t enough; vertical integration is required.
No producer of synthetic diamonds is valuable despite cut diamonds for consumers being the canonical luxury item because the core manufacturing patents have long expired. This suggests premium products with opportunities for branding differentiation aren’t enough; they must be patented.
Very few 3D printing companies have been good outcomes, and only a couple superalloys command a meaningful price premium and that’s generally in relatively niche markets. This suggests that niche markets aren’t enough even with proprietary production methods or proprietary materials.

Together, these three suggest an exciting new startup would need a proprietary production method for the large-scale manufacturing of a high-value material and push towards vertical integration as quickly as internal company dynamics allow.

With respect to integrating towards part making, we’ve seen companies commercializing novel alloying or metalization techniques that for the time being “integrate” by simply sending their material to a conventional CNC machining shop to be made into the final part and even with that outsourcing still capture far more value than if they had tried to sell the material directly.

An example of a company that’d excite us would be one that rises above the long history of failed attempts to beat the Kroll process and dramatically reduces Titanium’s idiot index from ~40x could unlock its widespread use given its superior properties to steel and aluminum for broad markets ranging from precise parts to structural supports.

These ideas become particularly interesting if generalized robotics become commoditized or cheaply contractable in a 5-10 year timeframe. Then, a company with a proprietary method of cheaply making useful materials could possibly integrate all the way to making the end products (e.g. automated shipbuilding with structurally lower costs; 30-40% lighter cars using titanium for crunch roll in front, steel for sidebar roll cages, and magnesium for the rest of the body)

Decentralized Water Desalination

Smac just presented our thesis on decentralized water desalination at our AGM. With desalination nearing cost parity, technology is no longer the core gating factor to resolving constant water shortages. Instead, politics block large-scale coastal plants from being built and even if such a plant did get approved it would still face the terrible business reality of begging the water utility to hook up to its canals. It wouldn’t be the first time we see a decentralized network spawn in the wake of slow, bureaucratic centralized solutions. Especially given the misaligned incentives that exist today for developing at-scale facilities in the US (aggressive planned build-outs in the Middle East are already underway).

With that in mind, one could build small-scale sites attached to minimalist solar arrays to deliver water to the local community. DePIN could even orchestrate the build out.

The data center boom makes this somewhere between viable and necessary. Activist groups are preventing massive campuses from being approved due to local water concerns, and it’s a frequently viral topic on social media. A parallel conversation happened 10-15 years ago with regards to traditional data center power consumption. Google addressed this issue by pioneering the use of virtual power power agreement and tariffs where they hook up to the grid but then pay for the creation of renewable generation elsewhere in that area.

A new startup could invent a similar process for water where data centers connect to the municipal water supply and then pay for an equivalent amount of capacity distributed throughout the community. That sure beats the current greenwashing where they pay for things like conservation of local marshes without adding any new capacity. There’s probably an important narrative framing here as well around removing the concentration of forever chemicals in the water supply, which is becoming an increasingly prevalent topic.

To capture this high-need market, the startup could use minimalist solar arrays and off-the-shelf reverse osmosis membranes (also a very mature technology). Its core competency would be cheaply deploying a simplified system and making it as easy as possible for developers to contract for water generation. Underneath a nice UI, it’d have automated permitting and site review.

Data centers’ immediate needs could get the startup to scale and then over time idiosyncratic local demand could drive DePIN allocation.

Maybe once the distributed network is on its way to saturating small-scale demand, the operations are ironed out and the protocol’s treasury has swelled, the network could shift to the cheapest possible water generation per unit, often entailing larger scales or more technically ambitious systems. One frontier to push on is building RO systems deep underwater. 70% of RO’s energy usage is pushing water through the membranes at pressures of 800psi. That pressure comes naturally at 1,000-2,000 feet under so energy costs can be cut by up to half.

Maybe the company even gets so good at distributing generated water that in the longer-term it leads the way on terraforming uninhabitable areas of the southwest, handling the greenification and then keeping a portion of value from the new city developed on top of the land.

Electrons Everywhere

Each home component that energy flows through is gradually being replaced by electric appliances that are both cost competitive and flat out superior experiences. From heat pumps to thermostats to inductive stoves to rooftop solar to backup power, these technologies save consumers energy costs and turn each house into a fully programmable VPP.

But this still requires customer choice and therefore meaningful SG&A. In a particularly intriguing business model innovation, Octopus Energy is partnering with home makers to build new homes with all these technologies connected through Octopus in exchange for homeowners getting 5-10 years of free electricity.

Even still, 30% of buildings can’t support the weight of silicon-based solar and rooftop solar panels are still a visual blight. We look forward to solar being flexibly woven into the shingles themselves or car rooftops. While some have hyped this for years, a few emerging technologies that are inherently flexible but efficient could enable this offering to be cost-competitive with existing solar. Rooftops manufactured with integrated solar would remove residential solar’s two biggest cost drivers: sales and rooftop installation.

We’re excited about such products and business models that embed electronics more passively into our lives and / or offer consumers delightful product experiences while making our infrastructure programmable.