How the Latest Tariff Threatens to Kill 70+ GW of Renewable Energy Power Plants — by David Riester
Background
We are going through a bit of a rough patch here in the US solar sector, if you hadn’t noticed. The pandemic hiatus of 2020 evolved straight into a severe supply-chain mess, and then came the inflation. Two years of troubleshooting and margin-evaporation then bled straight into perhaps the biggest exogenous gut-punch yet:
On April 1st, 2022, the Enforcement and Compliance division of the International Trade Administration in the Department of Commerce initiated country-wide circumvention inquiries to determine whether imports of crystalline silicon photovoltaic cells, whether or not assembled into modules, which are completed in Cambodia, Malaysia, Thailand, or Vietnam using parts and components from China, are circumventing the antidumping duty (AD) and countervailing duty (CVD) orders on solar cells and modules from China. The circumvention inquiries cover solar cells and modules that have been completed in Cambodia, Malaysia, Thailand, or Vietnam, using parts and components from China, that are then subsequently exported from Cambodia, Malaysia, Thailand, or Vietnam to the United States. This inquiry was prompted by a petition submitted by a previously irrelevant company called “Auxin Solar”, so the inquiry and its sprawling effects are broadly referred to as “Auxin.”
Importantly, Commerce has determined that, if circumvention is confirmed, duties/tariffs would be applied retroactively to April 1, 2022. In such circumstances – where the threat of retroactively applied tariffs overlaps with a review period at Commerce – “provisional” duties are collected. This is what is colloquially referred to as “duty deposits”. Simplified, this means any importer of modules subject to the investigation must give Customs and Border Patrol cash equal to the applicable tariff rate – in essence presuming that Commerce’s final determination will find the module providers in question “guilty” of circumvention. At the outset, the rates in place from the existing 2012 AD/CVD tariffs (the ones which are alleged to have been circumvented) are applied. Commerce intends to issue preliminary findings in late August / early September, including the various rates applied to various countries/manufacturers, and then the final determination is due on January 26th, 2023 [1]. Once final determination is issued, duty deposits are “settled up” (either direction); either money is refunded, or more is due depending on the final applicable rate relative to the duty deposit. Though final supplier-specific tariff rates often arrive with the final determination, Commerce is not forced to finalize those rates at that time – it could feasibly be months, quarters, even years until the rates are settled at the supplier level. In the meantime, the duty deposit rates are substantial, though very wide ranging (~12 – 250%). Most of the rates fall in the 34 – 47% range, all in. [2]
Auxin is already impacting a very large percentage of the market, which we will estimate below. In concept, this group is the universe of projects where three things are true, 1) they plan(ned) to use modules subject to the inquiry, 2) modules had not yet cleared customs on April 1st, but 3) would need to between now and whenever these tariffs lapse (or are otherwise nullified). With ~82% [3] of US modules coming from these sources, and the threat of these duty deposits and/or the actual tariffs existing for many years, this impacts a ton of would-be solar power plants.
This article will stop short of debating the merits of the Auxin petition or Commerce’s decision to take up the investigation, instead focusing on attempting to quantify the current impact of what has already occurred.
Quantifiable Impacts
To begin, let’s look at the impacts which we can measure quantitatively, with a few assumptions.
The Duty Deposit Itself
The most obvious one. The way these tariffs landed in 2012 is described in detail here. Trina and Suntech received special rates (34.29% and 46.41%, respectively), with most other manufacturers subject to 41.20%. Given that the “other” group falls right in the middle of the Trina <-> Suntech spectrum, that feels like as good an assumed all-in rate as any. For the sake of isolating various impacts, let’s assume a module price in the mid 30’s (cents/W) – which is where prices averaged out before re-stoked fears of protectionist tariffs muddied the waters [4]. That’s meant to be a price still associated with a messed-up supply chain and forced-labor problems – but not the added layer of tariff panic [5].
There’s a first number for our stacked up “Auxin Impact” – about 14.5 cents/W
Bonding Costs
But it gets worse, of course. Remember that the duty deposits are just placeholders. The final tariffs levied could, in theory, be considerably greater than the duty deposit rate. If you’re a module supplier enjoying no shortage of demand for your product and facing a bottomless pit of tariff rate risk, would you take the risk of paying whatever tariff comes to pass? Or would you force your buyer to take that risk? Whatever your answer, most of them happen to be doing the latter.
How are they doing this, you ask? Requiring the buyer post a bond against potential tariffs levied seems to be one common method. Estimating this cost is easy arithmetic, though the key assumption is mostly speculative.
Surety bonds charge an upfront fee against total bond coverage. Where I’ve used them in the solar sector, that fee has been in the 2.25 – 3.50% range. We’ll use 3% for the surety fee; that’s easy enough. The other bits are trickier, however.
Total bond coverage is the amount of coverage required (at the module sellers’ discretion) in excess of cash paid to the module seller. Implicitly, this would depend on i) module prices and ii) the tariff rate threshold to which seller requires “coverage”. The first of those is as volatile as ever (and circular referenced with the question at hand, moreover). The second is subject to the whims of a senior risk manager in Shanghai.
I’ll throw my finger in the air here and suggest we assume a 0.40/W module price, and desired tariff rate coverage up to 150%. Here’s that math:
There are at least a couple potential issues with this methodology. The first is that it’s hard to imagine that the risk being covered by the bond is truly 3% risk. I doubt an actuary doing insurance curve math on the magnitude and probability of the tariff risk would land on an insurance premium that is 3% of the policy amount. That said, the affordability of surety bonds has consistently surprised me to the positive over the years. Truthfully, I haven’t been close enough to any bonds papered for this purpose over the last month to have a data point, but would venture the guess that the clearing price for this bond product is higher than 3%.
The second potentially misleading aspect of the above methodology is the implied presumption that such a bond product is even available – at any price – to certain parties. The “principal” (the module buyer in this case) in a surety arrangement guaranties performance. The surety could always go after the principal if the “obligee” (the module supplier) calls on the bond. This is a critical aspect of the surety’s underwriting. No, the principal’s creditworthiness isn’t airtight – if it were, the surety bond probably wouldn’t be required – but in the cases where I’ve seen surety bonds used, the notion that the principal would make good on the obligation in the end…is not crazy. For example: Cypress Creek circa 2017, or SunEdison in 2011 (the times/places where I was using surety bonds). NOT Brand New Solar Company Named Sol______ which has some good pipeline but essentially no balance sheet. Given that a very significant percentage of the “worthy” projects nearing NTP are owned by precisely those types of firms, this problem is far bigger than ~2 cents/W of margin compression.
Supply/Demand Driven Price Hikes
There are direct impacts to a cost stack - like those just described - and then there are secondary, indirect effects of a tariff looming over everything. In the case of Auxin, the big one is opportunistic price increases imposed by module suppliers not subject to the inquiry.
Here’s a simple, important, objective, and indisputable fact: the global module capacity exempt from the inquiry is significantly less than total global demand. Any suggestion that this reality can be promptly remedied is completely, pathetically delusional. [6]
So, let’s say you’re an economically rational [7] module buyer. You’re talking to Trina about buying modules at a base price of 0.40/W, plus 0.144/W of duty deposit, plus 0.018/W of surety costs, for a total of 0.562/W. What price would you be willing to pay for an exempt module of similar quality? 0.56199999… of course. I appreciate in the real world, it’s not quite that simple, yet a basic microeconomics concept broadly observed in real world markets is that if the following two things are true, then sellers of an equivalent alternative product will set their price to just barely beat the next best offer (on a net basis):
- There is less supply than demand
- Price transparency of information is high enough such that the market makers’ net-to-buyer prices are, more or less, known
With those two things being true in the exempt module market, market prices will converge very near the net-to-buyer price paid for modules subject to the inquiry. Game theorists call this the “BATNA Principle”. Best Alternative To Negotiated Agreement. If Adani knows their customers’ BATNAs are 0.562, and their inboxes contain more solicitations to purchase modules than they could ever fulfill, then why would they offer anything meaningfully below 0.562?
You can call that price gouging if it makes you feel better. I doubt any court in a capitalist economy would agree. And yes, there are strategic reasons why a module provider might price more favorably (scale, customer quality, long-term price certainty, etc.). But it’s hard to ignore the reality that this BATNA-driven pricing behavior carried the day in 2017 – 2018 (during the somewhat comparable Suniva section 201 inquiry), so expecting any different here seems pollyannish. Sure, there are anecdotes demonstrating this isn’t happening across the board quite yet…but all the data points I have to date suggest we’re well on our way to that asymptote.
Of course, this phenomenon is not cumulative with the duty deposit and bonding costs – you’re either paying those things or buying the exempt modules at the jacked-up prices. The point is – unless this inquiry is dropped, we’re screwed either way. Sorry.
Delay Costs
As alluded to above, the logic above implies that, while it may be ugly, bid/asks are still converging to market clearing price points where would-be module buyers can always strike some deal to buy modules if they are willing to pay the market clearing price. That’s not the case on the ground, unfortunately. Many project developers and EPCs aren’t even getting access to the supply that’s out there – at any price point. This is usually because i) they aren’t creditworthy enough to address the perceived tariff risk, and/or ii) their projects/company/relationships aren’t generating enough “pull” with the exempt module suppliers, so they are not offered any of the (limited) supply. The reality on the ground is that many folks can’t even get a module company to “pick up the phone” – literally and figuratively.
So what happens then? Well, a lot of projects die simply because the delays imposed by this reality (see below) push past a cliff date. For the rest, the costs of the delay hit the bottom line, or in some cases serve as the economic straw that breaks the camel’s back…leading to project death by margin suffocation. Here are some of those delay costs, assuming delay until final determination in January 2023:
Land Options
Simple: If you must sit tight while you wait for module availability, you will incur costs while doing so. Maintaining site control is the most basic. Most utility scale projects secure their site via lease or purchase options. Let’s assume that the annual option payment for 100 MW worth of buildable land is $25k. (While this number varies considerably, $25k is a number I frequently see in this project size range. In the absence of any decent data set on this, I’m afraid “a number I see a lot” is as scientific as I can get). In most cases, a delay to early 2023 will trigger an additional annual option payment.
Of course, it could be far worse than that. Given that the most directly affected projects are those nearing construction start – when most land options are meant to expire – we should expect pervasive landlord extortion [8], wherein site control is only extended if the project developer pays very handsomely.
IX/PPA Deposits Cost of Capital
Most utility-scale projects require deposits for interconnection costs and later stage projects particularly vulnerable to the Auxin situation will usually have a PPA at such time module POs are put in place.
We scraped together a sample of 24 projects to estimate average IX deposit amounts. This universe consists of projects which are all of the following: 1. Between 50 – 200 MW, 2. Either built, under construction, or have a very high probability of going under construction in the near term, 3. Segue has good transparency into the deposit amounts. Not a perfect sample set, but I think it’s good enough for what we’re doing. The average per watt total deposit amount in that sample is 0.058/W. For PPA deposits, the most common methodology we see in the market is $75k/MW. However, deposits are not required for every project – in some cases it’s not appropriate, or the offtaker doesn’t have the leverage to demand a deposit. So, I’m going to haircut that 0.075/W down to 0.05/W. Between the two, that’s 0.108/W of deposits.
We need two other assumptions: 1. The cost of late-stage deposit capital, 2. The length of the delay.
Segue has a good perch to estimate cost of capital – I’d put this in a range from 9 – 14% [9]. Let’s be a generous and use 9%.
For time, we’ll use 9 months which is about the period between the Auxin news dropping and the final determination. This is probably a little generous, as things are unlikely to immediately settle the moment the final determination lands – it’ll take stakeholders a beat to digest the new playing field and decide how they are going to approach it.
Overhead
Ahh, the quiet killer of so many renewable energy developers. While firms will seek to put many projects directly impacted by the Auxin inquiry “on ice,” there are limitations to how thoroughly this can be done. Many different professionals will need to continue allocating some of their time to the project. “Life support” activities – adapting major equipment procurement plans, keeping landowners and local stakeholders happy, begging EPCs to leave labor allocated, keeping financing parties committed and in-the-loop, etc. – must go on. Here’s a swag at the burden for a 100 MW project (these comp guesses are a real can of worms, please don’t dwell on these – it’s not the point!):
Zooming out a little to show this aggregate impact in the context of the “original” module cost: 
That’s a ~43% increase to the costs of the primary component. I doubt many industries could continue producing their product at anything resembling the same scale/velocity if their main cost input increased by 43% overnight. Zoom out a little further and make an assumption for balance of system costs…
…and you’ll notice that this is still a ~20 – 25% increase to the total COGS (obviously dependent on BOS assumption). My MBA friends tell me that’s a problem.
And we haven’t even mentioned the many secondary and tertiary problems which are nearly impossible to quantify. I’ll offer one anecdote, just to give a flavor. Just this week I had two entirely unrelated conversations with truly top-shelf veterans of the solar industry who’ve recently arrived at a transition point (one sold their company, the other left a firm they’d been at for 5+ years). These are people who have, for over a decade, been real engines behind solar (and storage) plants showing up on the US electrical grid. Each initially planned to commence their next venture rather promptly, but in the last couple of weeks have decided to stay on the sidelines because of the uncertainty in the market. Similarly, hiring freezes are in place at many of the largest solar firms.
The resiliency and creativity of the US solar industry have long been its greatest attributes. This collective strength has had several sources of support (capital overhang, state-level political support, inelasticity of demand), but the primary factor is the remarkable talent base we’ve been spoiled with since day one. And, moreover, solar has tended to attract very gritty – even masochistic – folks who are willing to grind and persevere. When I hopped over from the venture capital industry, this “toughness” was one of the first things I noticed. But everyone has a limit. You can only get punched in the face so many times before you just want to stay down on the mat. Talent drains obliterate industries, even if their impacts are difficult to measure.
How Many Projects Will Be Affected?
The universe of projects most directly experiencing the impacts estimated above is defined by the following 3 criteria being true:
- The project planned to use modules subject to the inquiry
- The modules for the project had not yet cleared customs on April 1, 2022
- The modules for the project would need to be ordered before ~March 2023 to meet PIS dates
Let’s use ERCOT as a case study and slice up the queue in those three ways, extrapolate those findings to the broader US market… and then tally up the bodies. (It’s not a perfect methodology, I know; but, I bet it’ll be in the ballpark).
A few comments on the assumptions in the table:
Using the executed interconnection agreement as the proxy for project viability is a blunt tool, but as good a tool as I can come up with. This is also the most common methodology used by congestion consultants. Surely it will include many projects that die, but it will also exclude many that eventually place in service. Orennia produced some top-shelf analysis [10] earlier this year which suggests those “misses” in the methodology have roughly offset each other, so I’m confident this barometer will get us close enough.
The timing assumptions are subjective. Furthermore, to keep this tolerably digestible, we’re focused on the average of the universe here. This means that the assumptions are too high for projects on the lower end of the size range, and too low for those on the higher end of the same range. Again, I think we’ll be close enough to trust the findings at a high level.
When the chaos/paralysis clears is, perhaps, the hardest prediction to make. On one hand, we all could wake up tomorrow and find that the Dept. of Commerce has dropped the inquiry. On the other, the final determination could be so severe and sweeping that the bloodbath continues for years. The assumption that the air clears by March 2023 is meant to be conservative leaning, but it could still be badly off.
So just under a quarter of the projects in the entire ERCOT queue would appear to be in the Auxin crosshairs. It’s easy enough to zoom out and extrapolate this to the full domestic market, we just need to make an assumption about the total size of that market, and ERCOT’s portion therein.
According to most recent queue data [11] the total amount of solar and Solar+BESS across domestic ISOs is ~536 GW, and ~239 GW, respectively. For conservatism, we will assume that half of the Solar + BESS in various queues is solar (it’s probably more than half). This gives us a total of ~656 GW worth of module demand in the queues. 
Remember the point was to drill down on the ERCOT queue to establish the portion of solar in the queues likely to be directly affected by Auxin (about 24%), and then apply that to the broader domestic market.
OR
One hundred and fifty-five gigawatts! Woof. My initial reaction was “that must be wrong”. I caught myself starting to screw around with the assumptions to see if they were overly punitive…if I could get an answer to spit out that feels more right. Because make no mistake about it: if that is even within the right order-of-magnitude of the lost renewable penetration over the next ~3 years, it a mass extinction event for the US solar industry. [12]
How Many Projects Will Die?
“Ah, but wait!” you say… “just because a project is affected by Auxin doesn’t mean it’s killed by Auxin”. Indeed, this is true. We must filter this down a bit more.
First let’s estimate how many projects could “stomach” ~$0.172/W of lost value. For this we need to know the probability distribution of solar project margins. This data simply does not exist at an industry level, unfortunately. The best proxy I have is some aggregated analysis applied to a large developer pipeline. It covers 128 projects, and the snapshot is from 2019. The range of project sizes is 10 – 300 MW. The pipeline includes projects in 12 states. There are plenty of holes to poke – is this sample representative of today’s fleet of utility scale projects under development, nationwide? If there is better data on this available, I am all ears, but given that the margin distribution jives with Segue’s experience of the pre-Auxin market, I believe it’s a decent data set for our estimations. 
The percentage of the projects below $0.172/W is ~89%. Apply that to the ~155 GW estimated above:
Still a catastrophically large number. However, there are a few reasons why that number likely overstates the universe of solar power plants that will die because of the Auxin inquiry. The most significant among them are:
- Sunk Costs – I always hesitate to make this point, because developers have a bad habit of bastardizing the concept to justify bad decisions. But by the time a developer must secure modules, they have usually incurred considerable non-refundable costs. The correct way to make a “build vs. kill” decision on a go-forward basis is to assess a project’s expected marginal profit by measuring value against costs to be incurred. The margin probability distribution graph above includes all costs, including those which may be sunk at decision time. How meaningful sunk costs are for a project depends on several variables - I’d estimate a range between ~nothing and $0.08/W. On the high end of that scale, the % of projects in the “Dead Projects?” box above shrinks to ~45%, resulting in a “Total GW Killed By Auxin” of ~70 GW.
- Actual Vs. Expected Margin – Decisions like “should I kill this project based on margin X?” are made based on expected Yet the data set used above is mostly actual margins. Developers are an optimistic bunch who tend to see what they want to see. In my experience, actual margins come in below expected margins more often than they hit/exceed the mark. It’s reasonable to expect multiple GWs of projects get built at a loss. Of course, though that may take the edge off solar penetration deceleration, it will only serve to exacerbate the solar company mass extinction event that seems almost inevitable.
- The “Rubber Band Effect” – There are many stakeholders hovering around a would-be solar power plant who want to see it get built. Faced with the choice of i) the project dying, or ii) sharing some of the economic pain, many of them will choose the latter. This reality tends to bring many projects back from the dead. I’ve written an entire article on this phenomenon here.
It's false precision to try and calculate how big an effect these three (and other) factors will have on the magnitude of the inquiry’s project killing spree. For whatever it’s worth, in my head I estimate it will cut the previously calculated ~139 GW in about half, but that’s just a finger-in-the-air gut call.
For the sake of seeing this longwinded thought experiment to some semblance of a final answer, let’s say I’m right. I hereby predict ~70 GW of casualties attributable to Auxin, assuming no dramatic deviation from current course. For context, ~26 GW was installed in the US in 2021. It’s a horrifying prospect.
Conclusion
These are very dark days for the portion of the solar industry focused on getting renewable power plants placed-in-service profitably and rapidly. We are all still trying to wrap our head around i) what just happened, ii) what will happen next, and iii) what it will mean for our companies, our projects, and our careers; and are doing so against a backdrop of helplessness. How bad this ends up being is a function of dozens of factors – almost all of which are controlled by either the federal government or module manufacturers. One thing we can do is attempt to convey the severity of the situation to stakeholders in those groups, while de-mystifying the underlying phenomena driving that severity. Many politicians, bureaucrats, and industry observers are failing to appreciate both (and the guy whose attempt to remedy that just took 10 pages isn’t about to throw stones). While there are many alternative methodologies one could use to make the estimation made herein, I believe it’s important to at least try and demonstrate the magnitude of where the Auxin inquiry is taking our national energy sector. If the horror story told in the preceding pages comes to pass, let’s at least force everyone to look each other – and the data – in the eye as we continue marching down that path.
Author’s note: I had to take a lot of liberties to get through this analysis/experiment. I acknowledge the methodology is rather porous in some spots. But…because this is uncharted territory in so many respects, there are really only two paths to estimate the impact: either 1) Use a survey, as SEIA did, or 2) make a lot of assumptions, some of which are true swags. So go easy on me. Or, better yet, constructively offer your suggestions and data points by emailing
[1] The Secretary of the Department of Commerce can extend this by up to 65 days
[2] For all the gory detail on the sequence and timing: https://www.ecfr.gov/current/title-19/chapter-III/part-351/subpart-B/section-351.226#p-351.226(l)(2)
[3] New York Times. https://www.nytimes.com/2022/04/29/climate/solar-industry-imports.html
[4] EIA Import data
[5] For some reason module prices always seem to elicit disagreement and strong opinions. Please note this analysis isn’t really about module pricing, so if you are tempted to point out how my module price assumption is “wrong”, instead maybe ask yourself if/how the conclusion of this piece would change.
[6] Remember: over 80% of imported crystalline polysilicon modules (the dominant tech by far) are subject to the inquiry.
[7] In the neoclassical sense. The economic rationality principle is based on the postulate that people behave in rational ways and consider options and decisions within logical structures of thought, as opposed to involving emotional, moral, or psychological elements.
[8] If you’ve ever had to send one of your team members out to a gun-waving, whisky-drenched, shirtless tobacco farmer’s home with a briefcase full of cash on a ~$200M closing day, you wouldn’t take issue with this word choice, I assure you.
[9] Including all the closing costs, upfront fees, commitment fees, admin fees, extension fees, etc. For more on this, please read Segue’s writeup on the cost of development capital
[10] “Queue the Energy Transition: ERCOT”, Deng, Horner, Donaldson of Orennia. Orennia.com. Behind a license-wall
[11] From New Project Media, pulled 5/9/2022
[12] Mass extinction events of another variety are probably pulled towards the present too. Call that alarmist hyperbole if it makes you feel better, but that won’t help your grandchildren.
