How do we make solar even cheaper?

Building on the success of the last decade

Over the last 15 years, the cost of energy from solar photovoltaic (PV) generation has fallen precipitously. ‘Solar maximalists’¹ exult in a new golden age of abundant energy that will transform our civilization. The more pragmatic point to burgeoning interconnection queues and the reworking of residential solar rules in Hawaii and California as signs of the rapidly increasing viability of solar PV.

But learning curves² aren’t magic, and ‘solar’ isn’t some monolithic thing whose costs have declined monotonically. In this post I want to dig into what specific costs have been reduced, and where we might hope to see further declines.

I’ll focus here on the US³, drawing on data⁴ from the National Renewable Energy Laboratory (NREL) unless otherwise noted. We’ll look at utility-scale and residential PV in turn, with each demonstrating their own pattern of cost reductions and further opportunities.

---

---

Utility-scale PV

First, let’s establish the classic cost decline chart for utility-scale PV in the US. As noted, the costs have come down enormously (-82% per watt DC between 2010 and 2023, or -12% per year).

Source: NREL U.S. Solar Photovoltaic System and Energy Storage Cost Benchmarks

In this case though, because the cost declines have been so precipitous, it can be hard to tell what’s still left.

To get a better view, we can look at how the split of costs has changed over time. As we can see below, everything has gotten cheaper but some costs have fallen even faster than others (per the compound annual growth rates in red).

Source: NREL U.S. Solar Photovoltaic System and Energy Storage Cost Benchmarks

In particular, module costs have fallen fastest (even in the face of fluctuating tariffs), while (field) labor and 'balance of system’ (BoS) costs (like rails / mounting and conductors / wiring) have proven harder to cut⁵ or even gotten more expensive in recent years. This has left the Labor and BoS as a larger share of the remaining costs.

This underpins the thesis for a number of interesting startups focused here, including:

• Terabase: Construction pre-fabrication and automation (along with related offerings) that purports to significantly increase labor productivity, including a panel wielding robot arm.

• Erthos: ‘Earth-mounted’ solar that trades individual panel yield for BoS and land efficiency. The thinking professionals approach to “pave the world with PV”.

Alternatively, companies like Origami Solar—which manufactures steel racking in the US—have taken the relatively high prices of BoS components and attempted to differentiate on supply chain stability and domestic content (which can help achieve IRA tax incentives) rather than cost per se.

More generally, it will be interesting to see if the industry carves out a role for distinct ‘integrators’⁶ that pre-assemble modules + wiring + some BoS offsite to speed labor in the field and potentially reduce BoS costs, or move in the direction of Terabase where major projects include on-site integration factories.

Surprising (to me at least) is the relatively low non-labor soft costs. As we’ll see in a second, this is a big part of the story for residential solar, but it seems utility-scale interconnection and permitting delays are painful but not a big cash cost relative to the project as a whole.

Overall, this is a story of cost declines across the board, and further progress will likely involve incremental improvements across cost types.

---

---

Residential PV

We can perform the same analysis for residential solar, where costs have declined by -69% per watt DC between 2010 and 2023, or -9% per year.

Source: NREL U.S. Solare Photovoltaic System and Energy Storage Cost Benchmarks

Starting from a higher cost per watt, the declines have been impressive but not quite as steep as for utility-scale PV. The costs declines have also been more uneven, with the biggest improvements in Module and Labor costs.

Source: NREL U.S. Solare Photovoltaic System and Energy Storage Cost Benchmarks

The result is capital costs that are more than twice as expensive per watt as utility-scale solar.

Source: NREL U.S. Solare Photovoltaic System and Energy Storage Cost Benchmarks

Module and Labor costs are similar per watt between Utility-scale and Residential PV.

Inverter costs are meaningfully higher for Residential installations, likely a result of the competitive dynamics (and physical realities) of creating a plug and play system centered on someone’s garage (rather than on a concrete pad as part of a spec’d and procured ~$300M installation in the desert). This may grow more commoditized over time as industry interoperability standards become more widespread.

The story is similar for BoS costs. Theoretically, Residential BoS could be cheaper (there’s already a structure to attach the modules to), but the customized nature of these installations coupled with the relatively low bargaining power of individual installers leads to Residential PV being a price taker for components. The result recently has been a jump in BoS costs from ~$0.36 per watt in 2021 to ~$0.57 in 2023, likely due in part to supply chain disruptions.

As supply chains normalize and components continue to standardize, I would expect to see further improvement in each of these areas.

But what’s in that big brown Other bucket⁷?

For residential solar, the single biggest cost is customer acquisition.

Source: NREL U.S. Solare Photovoltaic System and Energy Storage Cost Benchmarks

In the United States today, residential solar is sold, not bought.

These are complex purchases worth tens of thousands of dollars, often involving a multi-decade contract. The result is sophisticated (and/or expensive) customer outreach (including both direct mail and in-person door knocking) and pipeline development (where leads are qualified, systems are designed, and customers are walked through their options before hopefully signing on the dotted line).

These customer acquisition costs have proven relatively resistant to improvement, and have ended up as the single biggest cost per watt.

Folks are tackling this at multiple stages in the process, including:

• Sales support tooling (e.g., OpenSolar): The sales process is complex. Sales and design tools that flow directly into creating a customized proposal (including potential savings) help reduce sales effort required per customer. (And the tools really are easy to use, as I found out writing my article on NEM 3.0).

• Solar marketplaces (e.g., Haven Energy): The marketplace is fragmented, with many distinct steps and capabilities between initial customer contact and system completion. Connecting customers and installers while smoothing the path from inquiry to system operation (often leveraging similar tooling and support as above) can reduce cost per customer by increasing yield.

• Subscription (e.g., Sunrun): High upfront costs are a big deterrent to potential customers, and ‘subscriptions’ or solar leases have proven an effective way to get customers past this potential hurdle. Per Sunrun’s latest quarterly filings, more than 80% of their customers opt for a subscription versus purchasing a system outright.

The final tailwind will come from rising utility rates. As electricity rates go up, the relative value of a solar (+ storage) installation goes up as well, likely easing the path to customer acquisition⁸.

At the same time, as module costs continue to decline we may start to see different paradigms for solar at home. For example, Germany is seeing growing adoption of ‘balcony solar’ wherein a small number of modules get plugged directly into the home through a normal outlet. While the absolute power output may be limited, this approach eliminates whole categories of costs relative to traditional rooftop solar.

As with utility-scale installations, the truth will likely be a combination of all of the above and more, but there remains significant opportunity to further reduce costs and increase deployment of these resources at all scales.

---

1. I jest somewhat, but I would actually count myself more in the solar maximalist group than not. Really, really cheap electricity—even if only available for part of a day—is a gamechanger, particularly given parallel advances in energy storage.

2. The decrease in costs as produced volumes increase, generally expressed as a % decrease in cost per unit for every doubling of ‘experience’. In the context of solar PV, the experience level is often equated to global deployments in gigawatts of capacity.

3. Australia (and other regional comparisons) are out of scope for this initial breakdown. Ostensibly, residential solar in Oz is a fraction of the cost as in the US, but it’s not clear what the specific differences are based on some quick research. Still trying to figure this one out.

4. Mainly this and this:
Ramasamy, Vignesh, Jarett Zuboy, David Feldman, Robert Margolis, Jal Desai, Andy Walker, Michael Woodhouse, Eric O'Shaughnessy, and Paul Basore. 2023. "Q1 2023 U.S. Solar Photovoltaic System and Energy Storage Cost Benchmarks With Minimum Sustainable Price Analysis Data File." NREL Data Catalog. Golden, CO: National Renewable Energy Laboratory. Last updated: December 12, 2023. DOI: 10.7799/2002868.

While I might quibble with some of the specifics of their methodology, the results are broadly corroborated by other observers like Lazard and Bloomberg NEF and have the benefit of public availability and granular cost attribution.

5. There has been a bit of recent fluctuation in costs per NREL, including a rise in BoS cost from ~$0.22 per watt in 2021 to $0.30 in 2023 and a jump in labor costs from ~$0.13 per watt in 2022 to ~$0.24 per watt in 2023. TBD if these patterns hold once the 2024 data is released.

6. Similar to what we see in battery energy storage with companies like Powin and Fluence playing an ‘integrator’ role in projects. They don’t manufacture the cells and modules themselves, but they design the system architecture / thermal management / battery-management system and get it assembled into a constructable package.

7. After ‘Acquire customer’ the next biggest cost is Profit within Soft costs. The underlying data does not shed much light on where specifically this estimate comes from, and I’m not sure this allows for a true apples to apples comparison between residential and utility-scale installations. More prosaically, with NEM 3.0 and bankruptcies like Sunpower it’s not obvious how much profit is happening right now. This may be more productively understood as the cost to integrate all these different activities and cost centers into a completed and installed system (over an above the ‘Management’ cost already included).

8. Though big installers may see this as an opportunity to accelerate deployments or take share for a given customer acquisition spend, rather than boost profitability by reducing this cost. TBD exactly how this plays out from a competitive perspective.