T1 Energy

IThe business

T1 Energy makes solar modules, the panels that go on a solar farm. It runs one factory, in Wilmer, Texas, called G1_Dallas, and is building a second, G2_Austin, to make the cells that go inside the modules. The company is listed on the New York Stock Exchange as TE, is based in Austin and had 562 employees at the end of 2025, plus about 618 agency workers at the Dallas plant.² It used to be FREYR Battery, a battery developer that went public through a blank-check company in 2021, lost money for three years without selling anything, and changed its name in February 2025 after buying Trina Solar’s American module plant.²¹⁰

That history matters. This is a solar manufacturer with one audited year of sales, 2025, and six months of the next. Before December 2024 it had no revenue. The auditor for 2025 was KPMG, which took over from PricewaterhouseCoopers in 2025.² I have no ten-year record to show, and the page does not pretend to.

What it sells, and to whom

G1_Dallas has a nameplate capacity of 5 GW a year on seven assembly lines, about 4.5 GW for utility projects and 0.5 GW for commercial and home roofs. A gigawatt is a thousand megawatts. The factory made 2.79 GW in 2025, 683 MW in the first quarter of 2026 and 935 MW in the second.⁸¹¹¹² The company expects to be at the high end of its 3.1 GW to 4.2 GW range for the full year, which means about 2.4 GW in the second half.⁴ I read that as a real ramp but not a certain one.

The customer is the unusual part. In the first half of 2026 one customer was 100% of sales, and the same customer was 100% of receivables.³ It is the Trina Solar group, the Chinese company that sold T1 the plant. The Trina group is also T1’s sales agent, its operations adviser, and a big supplier: T1 bought $428.1M of materials and components from the Trina group in the first half against $450.2M of gross module sales to it.³² Trina’s U.S. arm is a lender, too: it is owed the $65.0M production reservation fee, which carries no interest if paid on time.³ And it is a shareholder. It held 50.2M shares, 18.0%, in April, and the Form 4 filings show it sold 22.5M of them in the open market on 21 and 22 May 2026 at $8.13 to $9.20, about $190M.⁷¹³ I do not know who the end buyers are, because the filings do not name them.

Sales were $755.3M in 2025 and $427.8M in the first half of 2026, which makes the last twelve months $996.9M C (calculated).²³ The line to watch is gross margin. It was 7.4% in 2025, 16.4% in the first quarter of 2026 and 19.6% in the second.²⁴ Section V shows that all of it, and more, is a tax credit.

The plan: cells made in America

Today G1_Dallas assembles modules from cells bought abroad. For 2026 T1 buys them from suppliers that have certified they are not owned or influenced by the foreign entities the new law targets, and it says it is still checking some of them.²³ G2_Austin is the fix. It is a 2.1 GW first phase on a 100-acre site in Milam County, with a second 3.2 GW phase planned, and its cells would go into G1_Dallas modules to make them close to fully American.² Three things about it have moved the wrong way. The cost was $400M to $425M when construction began in December 2025, then $425M, and on 28 July 2026 it became $510M, which includes a 20% contingency for Texas construction costs. First cells were due in the fourth quarter of 2026 and are now due in the first quarter of 2027.²⁶ About $180.6M had been spent by 30 June, shown as construction in progress.³ I estimate about $345M is still to be spent E (estimate).

T1 does not yet have the money. Its stated goal is a “comprehensive financing solution, which includes a significant debt component,” first targeted for early in the second quarter of 2026, then for the second quarter, and still open on 12 August.⁸¹⁴⁴ In the meantime it has sold $184M of 4.00% convertible notes in April, $120M of 4.75% notes in July and $50M more of the 4.75% notes on 30 September, and it calls the last two a “bridge.”³¹⁵⁵

Contracts, and how firm they are

At 30 June 2026 T1 had contracts for 5.4 GW: 1.5 GW for the rest of 2026, 1.3 GW for 2027 and 2.6 GW for 2028 and 2029. The figure was 3.3 GW at 31 March, so it rose by almost two-thirds in a quarter.¹⁶ The contracts are priced by the watt, can be adjusted, and require customers to prepay half of each quarter’s price before it begins.³ That is useful, because it means customers fund the inventory. Three gigawatts of the 2026 volume are cost-plus or fixed-margin contracts.¹² The two named buyers of modules built with G2 cells are Treaty Oak Clean Energy, a three-year contract for at least 900 MW, and Clearway Energy Group, 641 MW announced in August 2026.⁸⁴ Both depend on a factory that has not made a cell.

Look at 2027. The factory could make about 5 GW and only 1.3 GW is under contract. The rest has to be sold at prices T1 does not control. In May the company said indicative demand covered more than 100% of 2027 and 2028 output, and in August it said it was “in discussions with multiple parties.”¹⁴¹² I treat that as a hope, not a backlog.

The other pieces

Four smaller things sit on top. T1 bought the TOPCon solar patents it had been licensing from Evervolt, a Singapore company, for $135M in four payments through 30 October 2026, payable in cash or in shares at a 15% discount to a five-day average price, and so far the company has chosen shares. The company says this ends royalties of $25M to $40M a year.¹⁷¹² It closed the purchase of KORE Power, a battery-storage engineering firm, on 1 July for an enterprise value of about $32M.³ It is “exploring value optimization” for its European leftovers, including a Norwegian data-center site with a 50 MW grid allocation and a place in the queue for 396 MW.⁴ Those are discontinued operations with net liabilities of $54.8M at 30 June, which I count against the shares.³ And it put $5M into Talon PV, which is planning a cell factory of its own in Baytown, Texas.⁹

IIThe moat

Two things protect a business: something rivals cannot copy, and customers who find it costly to leave. For a solar module maker neither is easy, and T1’s best protection is not a product advantage. It is a set of rules.

What is hard to copy

Since July 2025 U.S. law denies the 45X manufacturing credit, and a customer’s project credits, to businesses that are owned, financed or controlled by certain foreign entities, or that depend on their parts. T1 spent a year rebuilding to comply: Trina’s stake stayed under 25%, a $274M cash payment cut Trina’s share of the debt, Trina gave up its right to appoint a covered officer, and the patents moved from Trina to Evervolt and now to T1 itself.² On 12 February 2026 the Treasury issued first guidance, which T1 says matches its reading.² That notice covers the share of material cost that may come from foreign entities, and says the rules on who counts as a foreign-influenced entity are still to be proposed.¹⁸ A factory that is compliant, American-run and listed is hard for a Chinese-owned competitor to copy, and that is worth something. I give it real weight.

The second protection is the supply chain T1 is trying to build. It has multi-year contracts, partly at fixed prices, for American polysilicon from Hemlock Semiconductor and for wafers from Corning, and a frame agreement with Nextpower for American steel frames.¹²⁹ As I read it, only G2 would use those wafers. Until it runs, the American content in a T1 module is smaller than the pitch suggests.

What is not protected

Not the product. TOPCon is the standard high-efficiency cell design, and T1 describes two of its three module types as “similar to other commercially available modules.”² Not the customer relationship: the sole buyer is the company that sold T1 the plant, and the supply, sales and support agreements with it run to 2029 or later.² Not the warranty. When Trina’s trademark licence ended in December 2025, so did Trina’s warranty on modules without the Trina name. T1’s own warranty took effect with the first delivery of T1-branded modules in July 2026, backed by insurance and reserves.²³ It has no field record yet. Not the patents either: First Solar says T1 infringes U.S. Patent No. 9,130,074 by importing certain cells, and has asked the International Trade Commission for a general exclusion order covering TOPCon cells. The hearing is scheduled for 16 February 2027.³ T1 contests the claim. The danger is timing: an exclusion order could reach the imported cells G1_Dallas runs on before G2 makes any.

Judgement

A real set of legal and supply-chain protections, held by a business whose products are not different from its rivals’ and whose one buyer is the company that sold it the plant. I would pay for the first. I do not pay for the second, and I do not yet know how long the first lasts, because the credit it protects ends in 2032.

IIIThe competitive advantage

The alternatives to a T1 module are imported modules, American modules from other makers, and not building the solar project at all. T1 is not selling against other technologies. It is selling against price.

Do the accounts show an advantage today? They show one thing, and it is not a margin. The second quarter of 2026 had a gross margin of 19.6%, the highest yet, and it included a $24.4M tariff refund. Without the refund the margin was 9.9% C (calculated).⁴ I estimate the 45X credit inside gross profit was about $95.4M in the half, so without the credit and the refund gross profit was about −$41.6M, a margin of −9.2% E (estimate).³ Management itself says the quarter “reflects margin profile of T1’s contract coverage,” and the 2026 contracts are cost-plus or fixed-margin.¹² That is a business paid for its effort, not for an edge.

Where T1 might win

If G2 runs, T1 would make cells that few Americans make. The company says the scarcity of American TOPCon cells supports prices for its high-domestic-content modules, and that the Section 232 tariff and an onshoring programme would favour a company that is building cells, wafers and polysilicon contracts at home.¹² That is the logic of the plan. A developer building a project that wants the domestic-content bonus has a small number of places to buy.

Where it is exposed

The first exposure is the credit. The company’s run-rate target of $375M to $450M of EBITDA when G1 and G2 are both running is almost exactly what the credits alone would pay. At 7 cents a watt on 5 GW of modules and 4 cents on 2.1 GW of cells, at the 93 cents on the dollar T1 got in June, they come to $404M C (calculated).¹¹²³ The second is demand after 2030. The federal credits for the solar projects that buy T1’s modules end for projects not started by 4 July 2026 unless they are in service by 31 December 2027.² Safe-harboured projects carry demand for a few years. After that, T1 has to sell against imports with no help from either credit. The third is the buyer: a company that is T1’s supplier, lender and the seller of its plant has a lot of room to set prices between the two firms, and the filings show $450M sold to it and $428M bought from it in six months R (reported).³

IVManagement

Management here means a chief executive who came from investing and from the blank-check company that brought the business public, a chief financial officer who came from research and banking, and a board in which one investor, Encompass Capital, has a fifth of the stock. The company says its team has “an execution track record from large-scale industry and renewable energy projects.”² I cannot test that from the filings, and neither of the two top officers has a factory in the biography the company gives.

How they are paid, and what that pays them to do

The chief executive’s 2025 pay was $5,777,846: an $800,000 salary, an $800,000 bonus for individual performance, $420,000 for meeting company goals, and $3,752,000 of stock awards, most of it a one-time grant of 1,000,000 stock units on 1 January 2025 described as a “transaction bonus.”⁷ The chief financial officer received 1,500,000 units in the same way, and his total was $7,014,122.⁷ The long-term pay is stock options and time-vested stock units. I found no long-term award tied to profit, cash flow or return on capital.⁷ Half of the annual bonus rests on company goals and half on individual goals, the proxy amendment does not list them, and the committee can adjust the result.⁷

The incentive is to close deals and to keep the story going. For 2025 the company told investors it expected EBITDA of $25M to $50M, and in November, with the year almost over, it said it still did.⁹ Adjusted EBITDA for the year was −$65.0M, about $90M to $115M short.⁸ The cash bonuses were still paid. I do not read motive into that. The definition of the measure was changed in the fourth quarter, the 2025 year was a sell-down year, and the committee has discretion. But a plan that pays in a year like that does not make the shortfall costly to the people running it.

How much of their own money is in

Mr. Barcelo’s 1.11M shares are worth about $3.0M at $2.70, a little over half of one year’s pay, and the filing counts stock units granted as pay among them. Mr. Calio’s 2.19M shares are worth about $5.9M.⁷ All directors and officers as a group hold 62.7M shares, 22.5%, but 55.8M of them are Encompass’s, reported through Mr. Kantor.⁷ In the Form 4 filings from 31 October 2025 to 30 October 2026, I found no open-market purchase by any officer or director. The only open-market sales were five by Trina Solar (Schweiz) AG on 21 and 22 May 2026: 22.5M shares at $8.13 to $9.20, about $190M.¹³ Trina is not management, but it is the company’s main customer, supplier and lender, and the sales were 45% of its stake. The shares closed at $12.04 on 2 June, $9.48 on 30 June and $2.70 on 30 October.¹³¹

What they did with capital

The capital record is the weakest part of this page. Shareholders paid in $1.36B, shown as additional paid-in capital at the end of 2025, and the accumulated deficit was $1.09B.² The battery plan that the company was created for ended in discontinued operations, with $320.1M of valuation allowances against its assets and a Georgia site that sold for $50M in February 2025.² From 2021 to 2025 the weighted share count went from 75M to 174M, and by 10 August 2026 it was 294.5M.¹⁹³ In no year has it had positive owner cash without a one-time release of inventory.² Since December 2024 it has done the Trina deal, a restructuring to keep the credit, a $135M patent purchase paid in discounted shares and a battery-storage acquisition, each described by the company as a step toward an integrated supply chain.²¹⁷³

Three other items belong here. In August 2025 the audit committee found it had to restate the first quarter of 2025, because it had put the amortization of a customer-contract asset in the wrong place, and it identified a material weakness in controls that is still open in the June 2026 10-Q, covering the information systems and the revenue and inventory processes at G1.²⁰³ In November 2025 the company and “a Company executive and Board member” received grand jury subpoenas from the Department of Justice about sales of company stock in the second half of 2023 on account of that person, followed by a voluntary document request from the SEC; the company says it is cooperating and the filing does not name the person.³ And the 2025 audit report’s critical audit matter concerned revenue on bill-and-hold sales, where modules sit in warehouses for customers.² I do not read any of this as proof of a problem. I do read it as a reason to widen the range of values I show.

VFinancials

The honest history is short. FREYR Battery had no revenue from 2021 to 2023, and the columns for those years are a battery developer, not a solar maker. The solar business appears from 23 December 2024. Its first full year is 2025, audited by KPMG; the first half of 2026 is unaudited.²³ I show what exists. I do not extend it to ten years.

Add the six columns of owner cash and you get −$1,087.8M C (calculated). That is cash the shareholders put in and the business used, and it is why the share count has more than doubled since 2023. In 2025 owner cash was positive, but that year includes a $158.5M release of inventory and a $135.1M rise in payables.² Without the inventory release, owner cash was about −$153M C (calculated).

How much of it is the credit

The 45X credit is the key. T1 gets a federal credit for each module it makes in the U.S. and sells to an unrelated buyer: 7 cents a watt for a module, 4 cents for a cell, $12 a square metre for a wafer, available through 2032 and stepping down from 2030.² The company can sell it for cash. It sold $160M of 2025 credits in December for $0.91 on the dollar, and the last $39.1M in June 2026 for $0.93.⁸⁴ The credit is not shown as a line in the accounts. The company records it as a reduction in cost of sales.²

That lets me back it out. In 2025 the cash received and receivable for credits was about $182.0M E (estimate), which is 6.5 cents on each watt of the 2.79 GW produced after the discount, and 7.1 cents before it, and it sits inside a gross profit of $55.6M.²⁸ So gross profit without it was about −$126.4M. In the first half of 2026 gross profit was $78.2M, which includes a $24.4M refund of tariffs the Supreme Court found invalid, and a credit receivable that grew to $95.4M.³ Without both, gross profit was about −$41.6M, a margin of −9.2% E (estimate).

This is not a criticism of the credit, which is law. It says what the business is today: modules sold at about 31 cents a watt cost more than that to make with imported cells, and the credit pays about 6.5 cents of the gap back. I cannot tell from the filings who keeps the credit under the cost-plus contracts, only that gross profit is positive with it and negative without.¹²⁶⁴ At the full-integration plan, the credit is what makes the EBITDA guide work: $404M of credits against a guide of $375M to $450M C (calculated).¹¹

The quarters

Two points. The first-quarter result was the company’s best at the time, with $9.1M of adjusted EBITDA, and the second was $10.7M only because of the refund; the company itself first said the second quarter would be −$14.5M to −$11.5M before the refund, and then moved it into the figure when closing the books.⁶⁴ And the unit economics improved from the first to the second quarter only if the refund is counted: the company says gross margin rose by about 300 basis points on throughput and fixed-margin deliveries.¹² That describes the reported margin fairly. Without the refund, the margin fell from 16.4% to 9.9% C (calculated).

Cash is not earnings

In the first half of 2026 operating cash flow was −$103.0M and capital spending was $161.8M, mostly G2_Austin. Add stock pay of $8.5M and owner cash was −$273.3M C (calculated).³ Inventory grew by $112.7M and the government-grants receivable, where the 45X credit sits until it is sold, by $59.0M. Deferred revenue, which is customers’ prepayments, grew by $93.7M.³ The company funded the half with $175.7M of convertible notes, $6.8M from option exercises, and $114.3M of its cash.³ It is a growing business that uses more cash as it grows, with the subsidy arriving as a receivable first.

The balance sheet, and what ranks ahead of the shares

At 30 June 2026 T1 had $79.1M of unrestricted cash and $77.3M of restricted cash, held for debt service, customs and rental agreements.³² Against that it owed $173.4M on the bank loan, $65.0M of production reservation fee, $161.0M of 5.25% convertible notes due 2030, and $184.0M of 4.00% convertible notes due 2031: $583.4M of principal. In July and September it added $170.0M of 4.75% notes due 2031, for $753.4M in all C (calculated).³¹⁵⁵ Behind those sit $66.0M of Encompass preferred stock, redeemable on 23 December 2027, and operating leases with a present value of $230.7M, which sit in my cash costs, and $54.8M of net liabilities from Europe.³ Against the $753.4M of debt and $68.6M of preferred, I count $79.1M of cash, about $165M of net proceeds from the July and September notes, $95.4M of credit receivable and $25.3M of tariff-refund receivable, less $54.8M of European liabilities: $310M. Net debt and preferred stock is about $512M C (calculated).³ The market value of the shares at $2.70, counting the shares already owed, is $863M C (calculated), and the enterprise value is about $1.4B C (calculated).¹³

The wall in 2030 and 2031 is what matters. If T1 stays below the conversion prices, $4.46, $6.80 and $6.93, the converts are repaid in cash: $161M in December 2030 and $354M in 2031. By then the credit is falling. I come back to this in Valuation. T1 also has to find a way to pay for $345M of G2 construction before then, and a covenant on the bank loan requires a debt service ratio that it said it met through the date of the 10-Q.³

The shares

The share count is higher than the cover of the 10-Q suggests, and I build it item by item. On 10 August 2026 there were 294,527,066 shares.³ After that, 5.0M were issued to Evervolt on 30 September, and another 12.2M that the company owes Evervolt were deferred because they would take it past 4.99% of the stock: 2,246,377 for the 30 September payment and 9,962,645 for the 15 October payment, priced at $3.01 a share.²¹ The last $18M was due on 30 October, and I assume about 7.8M shares for it, using $2.70 as the five-day average price E (estimate). That is 319.6M shares C (calculated). It leaves out 38.8M shares for the preferred stock, which at a conversion price of $1.70 to $1.90 holders would take if the stock were above that; and about 3.4M shares that Trina has the right to buy at $1.70 if they do.³²¹ Another 7.4M stock units and 5.6M options were outstanding at the end of 2025, and I do not count them, because their cost is in my stock pay line.²

VIValuation

First, how this goes badly wrong

Before any number, I ask what would most likely hurt a buyer here. There are three ways, and they are different things.

The credit goes, or the plan misses it. In my base the credit is $374M of the $319M of EBITDA in 2029, which is all of it and more. The rules that decide who can claim it are new, and the 10-K said in March that no regulations had been proposed or finalized.² If Treasury or the courts read the foreign-entity tests differently from T1, or Trina or a supplier turns out to be a prohibited entity, the credit is lost for the years affected and the customers’ project credits are at risk too. Even if T1 qualifies, the credit steps down by a quarter each year from 2030 and ends after 2032.² My bear case has the credit but G2 slips, cells are scarce and the module margin stays negative.

The cells never arrive, or arrive late and cost more. G1_Dallas runs on imported cells until G2 produces. First Solar has asked the trade commission to bar imports of TOPCon cells, with a hearing on 16 February 2027.³ G2 has already moved from the fourth quarter of 2026 to the first quarter of 2027 and from $425M to $510M.⁶ A further slip leaves T1 paying more for the same plant with a worse module.

The funding takes the equity. T1 has about $79M of unrestricted cash and needs about $345M more to finish G2, while $0.8B of debt and preferred stock ranks ahead of the shares.³ In my base T1 borrows $213M of new money in 2027 at 10%, repays it from the good years, and then in 2030 and 2031 pays $515M to the convert holders as the credit falls away, ending 2031 with $284M of new debt. The company has already sold stock to Evervolt at a 15% discount, and its stock closed at $2.12 on 26 October.¹⁷¹ A raise at a low price would dilute everyone who holds today, and my per-share values do not fully show it.

What an owner can take out

Reported results are a poor guide, so I count only cash. For every year I start with EBITDA, which is the credits T1 collects, plus the margin on modules without the credit, plus the extra the cells earn, less the cash cost of running the company. I take away stock pay in full, capital spending and the build in working capital, and then interest, loan repayments and the preferred stock. What is left is what could be paid out to owners. In my base T1 collects $1,567M of credits from 2027 to 2032 and the rest of the business loses $488M over the same years; EBITDA adds up to $1,079M. After interest, repayments and capital spending, owners receive $69M in total, $0.22 a share, and most of it comes in 2033, out of the $400M I assume is left once the new debt is repaid.

Reported results and this number are far apart. Management’s adjusted EBITDA for 2026 is positive, at $19.8M for the half, and mine for the second half is $35M E (estimate). Subtract stock pay of $9M, capital spending of $190M and a $20M build in working capital, and the number that matters, owner cash, is about −$184M for the half, before interest and loan repayments E (estimate).⁴

What the price implies

At $2.70, counting the 319.6M shares described in Section V, the shares are worth about $863M. Add $753M of debt and $69M of preferred stock and take away $310M of cash and receivables, and the enterprise value is about $1.37B C (calculated). That is 3.7 times the low end of management’s run-rate EBITDA of $375M, and 4.3 times my base EBITDA for 2029, the best year in my base.¹¹ The three analysts who publish estimates expect a loss of $0.32 a share for 2026, $0.03 for 2027 and a profit of $0.09 for 2028, with a range for 2028 of −$0.26 to +$0.44.²²

Read that plainly. To get to $2.70 I need a business that earns about $183M a year more than my base in every year from 2028, including the years when the credit has gone. That is a company with a profit of about $183M without any subsidy at all, from a plant that today loses money before the credit. Or I need something left in 2033 that is worth $2.8B. Either is what the bull case assumes, and the bull case is $2.22.

What I can’t judge

Eight inputs sit outside my reach. First, what the plant earns without the credit: the filings do not print the credit, and I have backed it out from the balance sheet. Second, who keeps the credit under the cost-plus contracts. Third, what price T1 gets for the roughly 3 GW of 2027 volume that is not under contract. Fourth, whether the foreign-entity rules, which are still new, are read the way T1 reads them. Fifth, when and on what terms the G2 money arrives. Sixth, whether sales to the Trina group count as sales to an “unrelated person,” which the credit requires: T1 treats them so and sold its 2025 credits, but accounting and tax tests of “related” differ and the filings do not explain. Seventh, whether the 4-cent cell credit applies to cells made at G2 and put into modules at G1; the filings do not say, and about $70M a year of my base credits depends on it.² Eighth, whether the Trina group also earns the 5% sales bonus on adjusted EBITDA in the sales agency agreement, on top of the 5% operations fee I deduct: if it does, my base falls to $0.03 and my bull to $2.13.² With this much debt ahead of the shares, a small change in one input moves the equity from nothing to a few dollars. I show a wide range instead of a precise number.

Nothing in the grid reaches $2.70. The top right cell, $180M more EBITDA a year to 2032 and $1.0B left at the end, is $2.20. The grid says how far the price is from what a business which loses money before the credit and gets a lower credit each year can support.

The discount rate

I use 15%. The basis is the 10-year Treasury yield of 4.94% on 30 October plus a beta times a 4.20% equity premium.²³²⁴ I measured the stock’s weekly beta against the S&P 500 at 2.06 over five years, 2.78 over three and 3.28 over two.¹²⁵ A common adjustment pulls beta toward 1. It gives 12.1% on the five-year figure, 14.1% on the three-year and 15.6% on the two-year. I take 15%, in the upper half of that range, because the stock has dropped by more than three-quarters since 2 June and the cash flows depend on a law that can change. It is high against the 10% I use on the large companies on this site. The sensitivity table shows what it does: very little to the base and a lot to the bull case.

Three cases

One point on how I handle money T1 does not have. In every case, when cash would fall below $50M, T1 borrows new money at 10% up to $450M. If that is not enough, it sells new shares at $2.00 in the base, $0.90 in the bear and $3.50 in the bull, and each later payout is split over the larger share count. That is how dilution enters. The $66M of preferred stock converts into 38.8M shares, with 3.4M more going to Trina, if a share is worth more than about $1.90; otherwise T1 pays it back in cash in December 2027. The three convertible notes turn into shares at $4.46, $6.80 and $6.93 only if the shares are worth more than that when they mature.³¹⁵ In none of my cases are they.

I weight the cases 40%, 40% and 20%, for a value of about $0.48. Those weights are my guesses. They are not used in any value above. My base reaches $319M of EBITDA in 2029, which is 85% of the low end of management’s run-rate and two years later than its 2027. My bull reaches $429M in 2030, below the $650M to $700M the company gives for 5 GW of cells, because I phase the cells in and keep the 2032 credit step-down.¹¹ In the bull, the plants earn $320M a year with no credit. That is the part of the bull case I am least able to support.

Other uses of the same dollar

Treasuries pay 4.94% for ten years.²³ On my base, a buyer at $2.70 gets back $0.22 of cash to 2036 and no value after, so the return is −30.9% a year. Even my bull returns 12.1%, a little under the 15% I require, because the price is above it. The nearest comparable on this site from the same 13F is Bloom Energy, which sells power equipment to the same buyers: its base return is −6.3% a year on my figures, and its price is well above its base value too. GE Vernova returns 5.8% a year on its base. Neither is cheap on my numbers. This one is more expensive per unit of value than either, and the money is committed to a business with less cash behind it.

The margin of safety

My margin-of-safety price is $0.07, 75% of my base value. Because my bear value is zero, no price is safe from a total loss, and I do not claim one. The point of the number is the distance. It says the share price is about 39 times a level at which a buyer would pay for the base case with a quarter to spare. That distance is not a forecast of the share price. It says how much of the price is a bet on the bull case.

Two different questions

How good is the business, and how good is the price? On the business: it is a subsidised assembly plant with one customer, a sound plan to add cells and a balance sheet that has to borrow its way through the next three years. On the price: it is above all three of my cases. A business whose value depends on a law that ends in 2032 and on a funding round that has not closed can be a poor purchase at almost any price that leaves nothing for the ways it fails. At $2.70 I read this as a poor price for a business of uncertain quality. That is my reading of these filings. It is not a rating.

Where this could be in 5 and 10 years

For the base to hold at five years, G2 has to run from mid-2027, the credit has to arrive in full to 2032, and T1 has to get through 2031, when $354M of converts fall due, without selling shares at a low price. At ten years the base has nothing left to value: in my base the plants are worth $400M in 2033, which goes first to the new debt and then to owners, and the model has no business after that apart from $28M a year of rent on idle plants. A year-10 value of zero is a statement about my assumption, not a forecast of what is left. The numbers that drive this are the credit, the margin on cells, the cost of the money and the discount rate. The most likely way the base is wrong is that the cells arrive later or earn less, because the company has already moved the date once and the cost once.

The bear differs from the base mainly in whether the cells and the financing arrive on time and what a module earns without the credit. In the bear the equity is gone by 2031. The bull is a different business in 2033, one that earns $320M a year without a subsidy; its year-10 value is $4.06 and it has paid out $2.78 per share before then. These are estimates of value, not forecasts of the share price, and the price can sit above or below value for years.

VIIThe case for the price

What most people miss

Three facts cut in the company’s favour. Customers fund the inventory: contracts require them to prepay half of each quarter’s price before it starts, and deferred revenue was $198.6M at 30 June.³ The credit is cash: T1 sold the last of its 2025 credits in June at $0.93 on the dollar, better than the $0.91 in December.⁴⁸ And some debt is cheap: the $65.0M production reservation fee carries no interest if paid on time, and the converts pay 4.00% to 5.25% against 7.2% on the bank loan. The price of that money is dilution if the stock recovers.³

The bull case

In plain words: the country needs power, solar is the fastest source, American content now commands a premium, and T1 is one of few companies building both the modules and the cells. Management expects annualized adjusted EBITDA of $375M to $450M once G1 runs at 5 GW and 2.1 GW of G2 cells are ramped, and $650M to $700M with 5 GW of cells. CEO Dan Barcelo wants T1 to be “a leading U.S. energy producer and cash-flow powerhouse.”⁸¹¹ On Nasdaq the mean price target is $7.75, range $5 to $10, with three positive ratings, one neutral and none negative.²⁶ Encompass Capital reported 34,968,169 shares plus preferred stock convertible into 38.8M more in December 2025.⁷ Situational Awareness LP, the fund of Leopold Aschenbrenner, held 4,900,000 T1 shares, worth $46,452,000, at 30 June 2026, about 1.7% of the shares then outstanding, per its 13F-HR filed on 14 August 2026. A 13F does not show shorts or non-US holdings, and this says nothing about what the fund holds now or will do.²⁷³

Against reality

Today’s numbers support the near part, not the far part. G1 made 2.79 GW in 2025 and is running at about 3.7 GW a year, contracts cover 5.4 GW to 2029, and the credit has been sold for cash twice. If cells arrive by mid-2027, G2 reaches about 2 GW by 2029, modules keep selling at a small positive margin and the credit lasts to 2032, my base follows: $319M of EBITDA in 2029, 85% of the low end of management’s run-rate. That supports $0.09 a share, because the credit steps down and $0.8B of debt and preferred stock is paid first. The company missed its own 2025 EBITDA guide by $90M or more, and G2 has slipped a quarter and cost 20% more, so I use management’s plan as the bull, not the base.⁹⁸⁶

Multiply the steps out. My guesses: cells by June 2027, 65%; G2 financed without a large low-priced share sale, 55%; the credit lasting to 2032 as T1 reads the rules, 75%; no First Solar bar on imported cells, 75%; 4.5 GW a year sold at a profit after 2029, 50%; a profit after the credit ends, 30%. Multiplied, they give 3% E (estimate). Allowing for overlap, perhaps 5%, below the 20% I give the bull.

What a 10x would take. Ten times $2.70 is $27 a share: $9.8B for 362M shares with the preferred stock converted, about $10.3B of enterprise value. At 8 times EBITDA that needs about $1.3B a year, twice management’s best case of $650M to $700M, which relies on a credit that ends after 2032.¹¹ In five years it means 58% a year. The shares’ best close since 2021 was $16.08 in October 2022, and they rose 5.6 times from $2.16 in December 2024 to $12.04 on 2 June 2026, then gave back 78%.¹ It is close to impossible, and I would not build a case on it.

Where I come down

The prepaid contracts and the cash sale of credits moved me; management’s run-rate guide did not, because it is about what the credit alone would pay. Evidence that G2 runs on time, that the financing closes without a large sale of stock and that the plant makes money before the credit would raise my values most.

VIIIRisks, and what would change my view

The three ways this goes badly wrong are at the start of Valuation. These are the rest.

• One customer. The Trina group was 100% of sales in the first half of 2026 and 78% in 2025, and it is also T1’s supplier, lender, sales agent and former plant owner.³² If it buys less, or the prices between the two companies move, T1 has no other revenue to fall back on.
• Related-party terms. $428.1M bought from the Trina group and $450.2M sold to it in six months; $125.7M of payables are to related parties; the commission and support agreements run to 2029.³² I cannot see the arm’s-length test.
• Foreign-entity rules. A change in reading by Treasury, the IRS or a court could cost the credit. T1 says it complies. The 10-K said in March that no regulations had been proposed or finalized.² The next subsection sets out the tests and what failing them would do to the value.
• Patents. First Solar’s case could bar imported TOPCon cells, and it names T1.³ A settlement would cost money, and an exclusion order would cost output.
• Offtake dispute. RWE Investco sued in December 2025 over a long-term offtake agreement; T1 has cross-claimed and sued RWE’s parent on a guarantee of up to $100M. The 2025 write-off of $53.2M of customer-contract value relates to a dispute over a long-term offtake agreement, and the filings do not say whether it is the same one.³²
• Customs duties. Bills from U.S. Customs for 2024 imports by an entity T1 bought, which T1 contests and says would be covered by indemnification under the Trina deal.³
• Controls and investigations. A material weakness is still open, and the company and an executive have been asked for documents by the Justice Department and the SEC.³
• Overhang. The company has registered for resale 32,258,059 shares behind the July 4.75% notes and 13,440,860 behind the September notes, plus 13.6M and 7.2M Evervolt shares.²⁸³ Whether anyone sells I cannot say, but the supply is there.
• Demand after 2030. The federal credits for the solar projects that buy these modules end for projects not started by 4 July 2026 or in service by the end of 2027.² Projects already begun carry demand for a few years.
• Prices. The 10-K says announced U.S. module capacity could exceed installations by the end of 2026.² T1 itself sold into a “weaker than expected” market in December 2025.²
• The stock itself. It fell 78% from the 2 June close of $12.04 to $2.70, and 72% in the four months from the 30 June close.¹ Even if I am right on value, the path can be violent both ways.

Geopolitical risk, and what it means for the value

Four outside decisions carry this page: Treasury on who may claim the credit, the President and Commerce on import duties, Beijing on what Chinese suppliers may export, and Trina on what it buys.

The July 2025 law denies the 45X credit to a prohibited foreign entity, and in 2026 at least 50% of a module’s direct material cost must come from outside such entities, rising to 85% after 2029.²⁹ Trina Solar is a Chinese company, so it counts as one. T1 clears the ownership and debt tests with room: by my count Trina holds about 27.7M of 319.6M shares, 8.7%, against a 25% limit, and its $65.0M fee is 8.6% of $753M of debt, against 15%.¹⁸³¹³ Two tests are open. Notice 2026-15 says an IP licence with a specified foreign entity made or changed on or after 4 July 2025 gives it effective control. T1’s licence from Trina’s parent was changed on 29 December 2025 and moved to Evervolt, and T1 says it complies, but the full definition waits for regulations not yet proposed.¹⁸² And $428.1M of first-half purchases came from the Trina group. The notice looks through a reseller to the maker, and the filings do not say who made those goods.¹⁸³

Tariffs help and hurt. Section 301 duties of 50% on Chinese cells and 25% on modules, September’s final antidumping rates of 65% to 123% on cells from India, Indonesia and Laos, and a Section 232 action from 4 December 2026 with import floors of $0.22 a watt for cells and $0.38 for modules plus a 15% duty all lift the price of the imports T1 sells against.²³⁰³¹ T1 receives about $0.31 a watt from Trina.³⁶ But the filings do not say where G1’s imported cells are made, so the same duties may raise T1’s own cost until G2 runs, and a floor helps T1 only if Trina agrees to pay more: 3 GW of 2026 volume is cost-plus or fixed-margin.¹²

On 16 October Beijing proposed restricting export of polysilicon, thin-wafer, cell edge-passivation and perovskite technology, and of the technology to make the equipment.³² TOPCon is not named, and T1 does not say who builds G2’s tools or what Chinese support they need. I treat it as a watch item and model it as G2 a year late.

The price already assumes the credit is kept and the tariffs help: $2.70 is above my bull. By price alone, my base reaches it only if module prices rise about 5.7 cents a watt, about 18% on today’s $0.31, in every year from 2027 to 2032.

T1 Energy short report: Culper Research

Culper Research is a short seller. It says it is short T1 and expects the shares to fall, so it profits if the price drops.³⁴ Its report is dated 21 January 2026; an investor’s notes quote its title as beginning “T1 Energy (TE): The ‘American Made’ Charade Masking Chinese Supply and Control”.³⁵ I could not open it: Culper’s PDF (opens in a new tab) returns an access error to automated requests and I found no readable archive.³⁶ What follows is secondhand, from TheFly and Investing.com.³⁴³⁷

Culper says T1 is a “failed EV battery SPAC”³⁴³⁷ that sells itself as American-made while depending on Trina and not qualifying for the 45X credit;³⁴ that Trina had accrued $59.7M of commissions and fees in nine months, 83% of T1’s gross profit;³⁷ that RWE ended an offtake agreement on 19 December 2025 and sued,³⁴ and that T1 disclosed neither;³⁷ and that customs probes of past imports could create liabilities of $205M to $570M.³⁷ I leave out what it says about individuals. T1 has not answered Culper by name in any release, filing or call I found. Its 17 February 2026 release said Treasury’s first guidance “supports T1’s analysis of its eligibility for Section 45X tax credits”.³⁸

What is established, and what is not. T1’s 10-Q for the third quarter of 2025 shows $59.7M of nine-month costs with the Trina group; over gross profit of $71.7M that is 83% C (calculated).⁴⁰ T1’s filings date the RWE action 19 December 2025, served 7 January 2026, in San Francisco Superior Court, but not when, or whether, RWE ended the contract.²³ RWE’s federal complaint against a Trina affiliate (14 August 2026) says it ended the contract and sued on 19 December 2025, in case CGC26632187; that is an allegation, not a finding.⁴² The 10-Q of 14 November 2025 reported a potential dispute over an acquired customer contract and a $53.2M write-off, without naming the customer; its credit amendment counts the RWE offtake contract “solely until the termination thereof”, and an exhibit to the 30 December 8-K lists it among “Current Offtake Agreements”.⁴⁰³⁹ Neither mentions a suit; the first filing I found that describes it is the 10-K of 31 March 2026.² I found no filed securities class action; the 10-Q of 12 August 2026 lists no shareholder suit.³

Where it touches my model. I reran my model one change at a time. These are sensitivities, not forecasts; bear, base and bull stay at $0, $0.09 and $2.22. With the credit lost to 2032, base is $0.00 and bull $0.63 E (estimate), row one of “Five outside events” above, odds under 1 in 10. My base is thin, about $28M of equity value C (calculated): a one-off cost of about $44M in 2027 takes it to zero E (estimate). If the $31.7M of customs bills fell on T1 with no indemnity, base is $0.02 and bull $2.15 E (estimate); at Culper’s secondhand $205M and $570M, bull is $1.84 and $1.32 E (estimate). Trina-group costs at the 2025 level of $74.5M in 2027 to 2029, against my 2026 pace of about $37M a year C (calculated), give base $0.00 and bull $2.00 E (estimate). A $100M win on the RWE guaranty would give base $0.32 and bull $2.43 E (estimate); I do not count it. RWE’s complaint calls its fee and prepayments each “tens of millions of dollars”; it is suing T1’s subsidiary and a Trina affiliate for their return, and a Trina guarantor for up to $40M. The split is not in T1’s filings.⁴² The disclosure claims, the controls findings and the officer’s exit change no input.²

So what

For someone valuing T1, the dispute turns on three open questions: whether Treasury lets T1 keep the credit, who pays the 2024 customs bills, and whether the RWE claims cost T1 much. Culper is right if the credit is denied or the duties or RWE claims land on T1 at a size that matters; T1 is right if the credit stands, the Trina deal covers the duties and the RWE claims cost little. T1’s 10-K says it believes its legal matters will not be material, and also that outcomes cannot be determined.² Trina’s place as the only customer and the RWE suits are knowable now; the three answers are not. Culper profits if the price falls; T1’s management and insiders are paid in its stock. Watch T1’s results on Thursday 12 November 2026⁴⁴ and the 10-Q, then the federal RWE case, where briefing is set to run to 17 December.⁴¹