What this inquiry is about. When a supplier's sales fall and management says its customers have only delayed their orders, everyone who owns the stock is making a bet: that the money is late, not gone. This inquiry tests that bet at two publicly traded suppliers that blamed customer delays in the past year. Microvast makes batteries for commercial vehicles and builds battery storage systems. Photronics makes photomasks, the quartz stencils used to print chips. Where they stand: Microvast's shares closed October 30 at $0.57, down from $2.80 at the end of 2025, and in August Nasdaq gave it 180 days, to about late February 2027, to lift them back above its $1 minimum; its latest brokerage rating, in August, is a buy with a $4 target. Photronics' shares, at $27.58, are about half their May high, though the three brokerage ratings issued since late May are all buys, with targets of $41 to $44. Today neither company's books can tell late from gone, because the answer sits in their customers' files.
What we add is a way to know in advance where the answer will appear in each company's own reports, and when. At Microvast, an accounting rule does the work: a company can never reverse a cut to the value of its inventory. So if the storage units it wrote down by $29.0 million, after building them for one customer, ever sell at the old price, the gain will stand out as a burst of profit. At Photronics, the tool comes from climate science. Each possible cause of the drop leaves a different fingerprint across its product lines. And to call the drop a delay, its sales of masks for the most advanced chips must average $77.08 million a quarter for three quarters, more than it has ever sold in one. We have not seen either approach applied to these companies.
The takeaway for investors. We have written down, before the results, what each company's next reports must show: Microvast's by about March 2027, Photronics' by about June. On the evidence so far, the delay looks likely at Photronics and doubtful at Microvast. And the method travels: whenever a supplier blames its customers' timing, you can name the line in its own books that will tell you whether the money is coming.
Where this began. The Vista Research Desk reads newly filed complaints, the documents that open a lawsuit, as I do, because finding the right expert for a litigator is my day job. In October 2026 we read two securities complaints. Graham v. Microvast Holdings, Inc., in the Southern District of Texas, is a shareholder class action under the federal securities laws alleging that the company's public statements misled investors by overstating its ability to reach its margin targets, in light of inventory management issues and delays in its customers' vehicle rollouts, and to finish expanding its factory in China by the end of 2025. Cooper v. Photronics, Inc., in the District of Connecticut, is a shareholder class action alleging that the company's statements about demand for its most advanced masks misled investors; the complaint's theory is that the pipeline of new chip designs had stalled. Those are allegations, not findings. We take no side on either case, predict nothing about either, and nothing in this inquiry rests on either complaint. What the desk sent me from each was the same question: when Microvast's vehicle customers in Europe put off their programs, and when Photronics' chip designers put off releasing new designs, was the revenue that failed to arrive merely late, or was it gone? This inquiry uses only what the two companies have reported about themselves.
Both are suppliers, and their customers are the ones who know whether the money is coming. Microvast makes lithium-ion battery cells, assembles them into modules and packs, and sells them to the companies that build commercial vehicles: buses, delivery trucks, mining and port equipment. It also builds energy storage systems, large battery installations that store power for a grid or a factory, for the owners and developers of such projects. Photronics makes photomasks for the companies that design chips and the factories that print them, and for the makers of flat-panel displays, whose screens are also printed through masks. It does not name its customers: its two largest bought 29% of its sales in fiscal 2025, and about three quarters of its sales in its latest quarter were earned in Taiwan, China and South Korea. Its annual report says what that means for its sales: "The demand for photomasks is primarily correlated with new product design activity and to a lesser extent scaling up of manufacturing of end products." A chip factory running flat out, printing a design it already has, buys few new masks; a designer releasing a new design buys a whole set. In both cases the supplier is paid on the customer's calendar: Microvast when a vehicle maker's program is ready to take batteries, Photronics when a designer's new design is ready to be printed.
Here is what happened, as far as the filings say. Microvast reported that its sales in the last quarter of 2025 fell 15.0% from a year earlier "as a result of regulatory shifts in South Korea and delays in customer platform ramp up in EMEA." A platform is the industry's word for a vehicle or machine design that a maker builds in volume once it is ready, and a program, the word we use, is a maker's plan to bring one into production; EMEA is the company's term for Europe, the Middle East and Africa. For the first quarter of 2026 it cited, among other causes, "OEM-caused platform delays in the transportation and mining sectors in Europe and APAC, respectively," that is, vehicle programs in Europe and mining programs in Asia, an OEM being the vehicle or equipment maker that buys the batteries. The filings do not say which vehicle makers delayed, why, or whether any cancelled a program or scaled one back; they say the platforms were delayed and that the delays were the customers', and in March 2026 that the European delays were beginning to resolve. Separately, over 2025 the company cut the book value of storage units it had built for a customer by $32.5 million, $29.0 million of it in the fourth quarter, writing that the inventory "was specifically manufactured for a customer and did not generate any associated revenue during 2025." It has not named that customer, described the order, said whether the order still stands, or said what the units will become.
Photronics reported that in the quarter ended May 3, 2026, its revenue from chip masks fell 10.8% from the quarter before, "primarily due to delayed design releases," and most of that fall came from masks for the most advanced chips, down 20.5%. Its chief executive named three causes in one sentence: chip factories running so full that new product launches were being pushed back, shortages of memory chips and the costs they put on the makers of the devices that use them, and geopolitical uncertainty. One quarter later the company reported "a recovery of some of the temporarily delayed semiconductor design releases" and its investor deck called the recovery modest. The filings do not say how many designs were delayed, whose they were, or how many have since been released. No cut to the book value of its goods followed at Photronics, because a mask is made only after the design arrives; what it built ahead was capacity, not goods, and its reports since have shown no cut to the value of that capacity either. That is the difference between the two suppliers. Microvast built goods for one buyer, earned nothing on them in 2025, and cut their value when it no longer expected to recover their cost; Photronics built only the means of making masks, so a delay costs it orders but leaves nothing on its shelves to mark down.
Now the two words, concretely: late and gone, which the tests below call deferred and lost. For Microvast's vehicle business, late means the European vehicle makers take the batteries after all, in a later quarter, so that Europe's revenue for 2026 ends up at or above 2025's; gone means the programs never take them, and the revenue never arrives. For the storage units, late means the customer, or someone paying about the agreed price, takes them; gone means they are sold to someone else for far less, or scrapped. For Photronics, late means the delayed designs are released in a later quarter and their masks ordered then, so that the dollars missing from the spring show up in the quarters that follow; gone means the designs are dropped, folded into later ones, or printed through another mask maker's plates, and the dollars never show up at all. In the quarter the delay happens, a sales figure cannot tell these apart. It records what did not arrive. Whether it is coming, it cannot say.
Delay was the word Microvast used for its vehicle customers in Europe and Photronics for its chip designers, and in both cases it may well be true. Here is where the answer will show, and when. For Microvast the place is one line of profit in its annual report for 2026, due about March 2027: gross margin, which is gross profit as a share of sales, and which should show what became of the storage units; Europe's revenue in the same report should show whether the vehicle makers came back. For Photronics it is a sum: five quarters of sales of masks for the most advanced chips, complete when it reports its fiscal second quarter of 2027, about June. Late and gone move the two places differently: at Microvast, late would show as a burst of profit on that line, and gone as silence or a second cut; at Photronics, late fills the sum, and gone leaves it short.
For Microvast the instrument that reads that line is what economists call make-or-buy. Microvast sells standard battery products to hundreds of customers, but these storage units were built to one customer's order, and goods made for one buyer are worth less to anyone else. They earned nothing in 2025, and when the company no longer expected to recover their cost it cut their value on its books, by $32.5 million for the year, a write-down in the accountant's word. Because that cut can never be reversed, a later sale near the original price would show as an unusually large profit, and a sale near the written-down value would leave little trace. The stake is high: customers in Europe brought nearly half of Microvast's revenue in 2025, and in August the company said that substantial doubt about its ability to continue as a going concern, the accountant's phrase for a company that may not be able to pay its bills for the next twelve months, had not been alleviated.
For Photronics the place is a sum across quarters, and the instrument that reads the cause is fingerprinting, Klaus Hasselmann's method from climate science: write down in advance the pattern each candidate cause would leave, give the most weight to the quiet places, the lines that usually move least, where a change means something, and see which pattern the data match. Of the three causes management named, full factories that slow the release of new designs should fall hardest on masks for the most advanced chips, the high-end line, touch the mainstream line for older chips less, and leave display masks alone, and the quarter ended May 3, 2026 comes close: high-end down 20.5%, mainstream down 3.4%, high-end display up 12.5%. The fit leans toward a delay, though we made it after the figures were out; the prediction written in advance is Test E, one of five dated tests below. A delay can still end in a cancellation, so the main test is a sum: five quarters of high-end chip mask revenue from the quarter of the fall, against the $356.41 million that five quarters at the $71.3 million record would have brought. Two quarters in, the last three must average $77.08 million; even the lower bar, from the growth trend before the fall, needs $71.91 million a quarter, above the record. The stake is the business Photronics is building toward: the high end's share of its chip mask revenue rose from 30% to 39% in three fiscal years, and its chief executive says its investments in the United States and Korea are meant to strengthen its position at the high end of the market over the coming years.
We wrote five tests down and froze them on October 30, 2026, before either company's next results, so that no bar can move once the figures arrive. "Cannot tell" is a legitimate outcome for every test here, and we will report it as one.
Microvast Holdings is a publicly traded company whose shares trade on Nasdaq under MVST. It was founded in 2006 and is headquartered in Stafford, Texas. Its shares have traded publicly since July 2021, when it combined with Tuscan Holdings Corp., a company formed in 2018 for the purpose of merging with another business. Its filings shorten energy storage systems to ESS, and so do we from here. What it says sets it apart is that its own technology runs through the whole battery, from the four core materials of a cell (the cathode, anode, electrolyte and separator, as I remember from peeking at my friend Steve Chin's exam paper in eighth-grade physics at Central Junior High) through the cells, modules and packs to the electronics that manage them. Its batteries, it says, can charge from 10% to 80% in 10 to 30 minutes under suitable conditions, a speed it calls critical for fleets in ports, warehouses and mass transit that cannot sit idle. Is that unique? By its own account, no: its annual report concedes that other manufacturers can charge "as quickly as" its batteries, and rests its edge on how long its cells last, some, it says, up to 20 times longer than standard lithium-ion cells. It cites patents and licenses, among them a cathode design licensed from Argonne National Laboratory in 2017. Passenger cars come last in its plan, after buses and taxis, and only where cities build fast chargers its batteries can use. By its own account, demand for batteries for electric commercial vehicles was once concentrated in Asia and the Pacific, and the company has been working toward Europe and North America.
Winning a vehicle maker takes time. The company expects two to four years from a customer's first introduction to commercial manufacturing, and the long-term supply or framework agreement that typically results sets no minimum purchase; the customer commits through purchase orders for specific quantities, and by the company's own account even those can be adjusted, cancelled, deferred or suspended. All of its cell capacity at the end of 2025, about 3.5 gigawatt-hours a year, was at one factory complex in Huzhou, China; a plant near Berlin has assembled modules since 2021. A gigawatt-hour is the energy a billion-watt load would use in an hour. Revenue in 2025 was $427.5 million, up 12.6%, on 1,879.5 megawatt-hours of batteries sold, 16.5% more than the year before. The company had 314 customers, but the five largest brought 58.4% of revenue.
The geography matters, because the first of our tests is about Europe. Microvast's annual report breaks revenue down by where the customer sits, and the table's label is "Europe," though the company's own discussion calls the region EMEA. Europe brought $211.9 million in 2025, 49% of the total and 12.9% more than in 2024, Italy and France most of it. The order backlog at year end, the orders taken but not yet delivered, was $196.1 million and, in the company's words, "primarily composed of long-term transit and logistics partners in Europe."
By its reported figures, the year had gone well until the fourth quarter. In that quarter sales fell to $96.4 million from $113.4 million a year earlier, a drop of 15.0%, and gross margin fell from 36.6% to 1.0%; of the 35.6 points it fell, 30.1 were the $29.0 million of the storage write-down booked in that quarter. The company attributed the fall in sales to the regulatory shifts in South Korea and to the platform delays in Europe. Korea is a regulatory story and a different mechanism, and we leave it out of the test; Europe is the customer-delay story, and the one we follow.
That write-down is the heart of the Microvast half of this inquiry. In its 2025 annual report, the Form 10-K every publicly traded American company files, Microvast disclosed a write-down of $32.5 million for the year on energy storage inventory. Inventory is carried at the lower of its cost and what it would fetch less the cost of selling it, and the shortfall is charged against profit. Microvast described this charge in a sentence that deserves slow reading: "This ESS inventory was specifically manufactured for a customer and did not generate any associated revenue during 2025." It set the charge apart on its own line of the income statement, calling it material and unusual. Without it the year's gross margin would have been 36.2% instead of the reported 28.6%; $29.0 million of the charge landed in the fourth quarter; the other $3.5 million fell earlier in the year, by our reconciliation of the company's figures and not by any sentence of the company's own.
The same 10-K reports the one dispute with a storage customer the company has disclosed, and it belongs beside the charge. Clenera Battery Holdco had a supply agreement with Microvast for what the filing calls "custom-made battery containers." In November 2024 it began an arbitration, a private case decided by a hired arbitrator instead of a court, claiming that Microvast must refund about $36 million for its "failure to deliver the containers by the contractual deadline." On December 3, 2025, the arbitrator issued a final award to Clenera of about $42.9 million; the two companies settled confidentially on February 12, 2026; and by August Microvast reported that it had made the payments. At the end of 2025 the balance sheet carried $35.9 million of "Advance from customer" among the sums it owed but had not yet paid, its accrued liabilities, with a note pointing to that dispute. The filings do not say whether the written-down storage inventory is those containers. We do not assume it. If it is, the delay that mattered was, by the terms of the claim and the award, the supplier's own; the customer has been repaid; and a later sale could only be to someone else.
The rest of the balance sheet tells of a company that built and then waited. Total inventory fell 37.6% over 2025, to $89.4 million, as work in process and raw materials ran down and the write-down took its share; finished goods moved the other way, up 47.1% to $27.6 million, more product made and waiting. Receivables rose 29.1% while revenue rose 12.6%; by our computation receivables at the end of 2025 equaled 133 days of the year's sales, against 116 a year before: customers were taking longer to pay. Advances from customers, the deposits a buyer pays before delivery and a supplier's natural protection when it builds to order, fell from $43.7 million to $5.6 million. Of the $43.7 million held at the start of the year, only $5.6 million, by coincidence about the same figure as the year-end balance, was earned as revenue; most of the rest, $35.9 million, became the advance owed back to Clenera. Here the deposit did not protect the supplier. It went back to the buyer.
Then came 2026, and the first chance to check the word. First-quarter revenue fell 48.0% to $60.6 million, on 273.9 megawatt-hours sold against 535.7 a year earlier; the company cited regulatory and geopolitical shifts in India and Korea and the OEM-caused platform delays in Europe and Asia. Europe fell to $43.2 million from $60.1 million. The second quarter was better: $87.3 million, down 4.5% from the year before, with Europe up to $52.4 million from $38.9 million and Italy alone at $39.8 million. At the half, revenue was $147.9 million, 28.8% below the first half of 2025; Europe's $95.6 million against $98.9 million was 3.3% below, though it was now 65% of the whole.
The test follows from those numbers. For Europe's 2026 revenue to match 2025's, the second half of 2026 must bring $116.3 million, against $113.0 million in the second half of 2025. The second quarter's pace, carried across a full year, comes to $209.6 million, 1.1% under the 2025 figure. The company said in March that "The momentum in EMEA is encouraging as we continue into 2026, particularly as previous vehicle platform delays in the region begin to resolve," and by August spoke of Phase 3.2, a second production line at Huzhou of 2 gigawatt-hours a year built mainly for its next generation of cells, as its main milestone, and gave no regional forecast. It has not given a revenue figure for 2026. We give one: the 2025 figure for Europe is the bar, and the second half decides whether it is cleared.
Two more facts turn an accounting question into a question about the company, the first its own account of its finances. The 10-K reported substantial doubt about its ability to continue as a going concern, and management's judgment that its plans would probably alleviate it; the auditor singled out that evaluation as a critical audit matter, the label for the parts of an audit that required the most judgment. By the second-quarter report in August the company's own position had changed: "substantial doubt about our ability to continue as a going concern has not been alleviated," because parts of its plans depended on factors outside its control. It counted $104.2 million of its $118.6 million of borrowings as due within twelve months, against $143.1 million of cash and restricted cash. Then, on August 31, it reported a Nasdaq notice: its shares' closing bid price had been below the exchange's minimum of $1 a share for thirty consecutive trading days, and it has 180 days to cure that, possibly more, failing which Nasdaq would be expected to notify it that its shares are subject to delisting, removal from the exchange, which the company may appeal. Inventory built for one customer looks different at a company that needs the cash.
The second fact is quieter. In its first-quarter report Microvast wrote that it was "deferring ESS development" to put its capital into commercial vehicle work that pays sooner. The written-down inventory is ESS inventory. A company deferring storage development to save cash has put its money elsewhere. What it does with the storage inventory it already holds is what we will watch the books for, though a sale near its written-down value may barely show. The other company's clue is of a different kind: no charge, no stranded units, only a pattern across the lines of a revenue table.
Photronics is a publicly traded company whose shares trade on Nasdaq under PLAB. A finished chip design needs a set of photomasks, dozens for an advanced chip. The moment a designer declares a design finished and sends it off to be turned into masks is called the tape-out, by the usual account from the days when the design left the building on magnetic tape. Photronics is a merchant mask maker: it makes masks to order for chip designers, for the foundries, the contract factories such as those in Taiwan that make chips for others, and for the makers of flat-panel displays. It had about 636 customers in its fiscal 2025, which ended October 31, 2025, and revenue of $849.3 million, 72% of it from chip masks and 28% from display masks; its investor deck calls its Hsinchu plant the largest commercial mask maker in Taiwan. Its backlog for chip masks is a week or two: "Our customers generally order photomasks on an as-needed basis," the 10-K says, so each quarter's revenue depends mostly on the orders that arrive in it, and when customers postpone new designs, Photronics waits.
Photronics splits its chip masks into two lines, and the definition is fixed in the filings. High-end means "photomasks that are 28nm or smaller," where 28 nanometers names a generation of chipmaking, a node, by roughly the size of its smallest features; mainstream means masks for "IC nodes greater than 28nm," an IC being an integrated circuit, a chip; those are the older and larger generations that still account for most of the designs made in volume. High-end masks are harder to make and carry higher prices, so the high-end line is where the company's growth and ambition lie. Over three fiscal years high-end chip mask revenue rose from $194.9 million to $238.9 million while mainstream fell from $456.3 million to $376.2 million. A rising high-end share can mean more advanced designs released, or simply fewer old ones, and the two must be kept apart.
The quarter before the fall was a record. In the fiscal first quarter of 2026, the three months ended February 1, high-end chip mask revenue reached $71.3 million, up 8.3% from the quarter before and 18.6% from a year earlier; the company credited the rise to accelerated demand in Asia before Chinese New Year. Total revenue was $225.1 million, and the company gave its forecast for the next quarter, what companies call guidance, as $212 million to $220 million.
The next quarter missed the low end. Revenue for the three months ended May 3 was $209.9 million, down 6.7%, and high-end chip mask revenue fell to $56.7 million, a drop of 20.5% from the record. Mainstream chip masks fell 3.4% to $90.9 million. Display masks went the other way: high-end display rose 12.5% to $52.8 million, on what the company called strong demand in the Asian IT display market, while the small mainstream display line fell 25.1% to $9.6 million. The 10-Q, the quarterly report, gave the cause in one phrase: "primarily due to delayed design releases." The CEO's statement gave the three causes in one sentence: "In the near term, certain design releases have been delayed due to elevated fab utilization rates, which are extending new product launch timelines, memory supply constraints and related cost pressures for OEMs, and geopolitical uncertainty." A fab is a chip factory, its utilization is how full it is, and the OEMs are the device makers who would have launched new products.
The third fiscal quarter brought some of it back. Revenue for the three months ended August 2 was $216.0 million, $1 million above the top of the $207 million to $215 million the company had forecast for it, and high-end chip mask revenue rose 20.9% to $68.5 million. That was 3.9% short of the record, but it was 44.3% of all chip mask revenue, the highest share in the fifteen quarters we assembled, and the company called it "a record percentage for our high-end I.C. business at 44% of total I.C. revenue." Mainstream chip masks fell again, 5.2% in the quarter and 8.5% from a year earlier, to $86.2 million. The CEO spoke of "a recovery of some of the temporarily delayed semiconductor design releases that occurred back in our fiscal Q2." The investor deck called it a modest recovery and expected the third quarter's conditions to continue in the fourth. Guidance for the fiscal fourth quarter, which ends on October 31, 2026, is $207 million to $227 million.
A record share, though, can come from the high end growing or from everything else shrinking, and this one took both. If mainstream revenue had held at its second-quarter level instead of falling, the third quarter's high-end share would have been 43.0%, just under February's 43.1%. The shrinking mainstream line made the record, and the high end's recovery made the rest of the rise. A record share is not the same as a record; the dollars were still below February's, and the company's own words were "some" and "modest." That is why the test we freeze adds up dollars over five quarters. Which of the three named causes left this mark, and whether the designs it held back are coming at all, are two different questions. In the quarter they are asked, the books answer neither.
Set the two companies side by side and the shape is the same. Each built ahead of its customers' programs, Microvast storage equipment to one customer's design, Photronics new mask-making tools and new capacity in Texas and Korea; each was told by its customers, in effect, not yet; and each has reported a stretch, a year for Microvast and a quarter for Photronics, in which the not-yet looks like a never and the ledger cannot tell the two apart. Microvast books revenue when the customer takes control of the batteries; Photronics books nearly all of its revenue as it makes masks already ordered; either way, an order delayed and an order cancelled leave the same hole in the quarter. The decision, meanwhile, is the customer's. The Clash put its question in a title in 1982, "Should I Stay or Should I Go," and spent the whole song asking it without once answering; a supplier's books hear about that much.
I know something about building the supply before the demand, because I did it. In the early 1990s my co-founders and I spent three years recruiting professors into a database of experts while we held day jobs, and on the night we quit those jobs we went to the Pump Room in Chicago to celebrate. Near the end of the wine one of us mused about who our first clients would be, and the dinner turned from celebratory to tense. As I set it down in The Scholars' Treasure, "Who forgets to think about the most fundamental part of a business plan: customers?" (ch. 6, p. 65). Two chapters earlier I had written of the database we were so proud of: "The only problem was that we had no clients, nor did we even imagine who our clients would be" (ch. 4, p. 50). Three years of work had produced a very good answer to a question no one had yet asked us. The customers came, eventually and in full, so our revenue was deferred and not lost. But nothing on our books in 1996 could have told you which it would be, and that is the problem here.
Ecclesiastes says that "For everything there is a season, and a time for every purpose under heaven" (Ecclesiastes 3:1, WEB), and the next verse names "a time to plant, and a time to pluck up that which is planted." The verse holds both cases: a supplier that builds ahead has planted, and the question of this inquiry is whether what follows is the harvest or the uprooting. The French economist Frederic Bastiat, writing in 1850 about what is seen and what is not seen, made the point that an economist's whole skill lies in counting the things that did not happen because of the thing that did: the window the boy broke is seen, and so is the glazier's trade; the shoes the shopkeeper would have bought with the same money are not. A design released late is seen, in the quarter it finally ships. A design never released is Bastiat's unseen: no mask, no order, no line in any table, only a total that stays lower than it should have been. That is why a test for lost designs has to be a sum, over enough quarters that the unseen shows up as a shortfall. And it is why, at a supplier that has already written its units down, the test can be sharper still: there the unseen has been given a number, and a rule.
The first model we bring to the Microvast case, from Vista's Decision Library, the toolkit of models on our site, is the one economists call make-or-buy. It belongs to Oliver Williamson, who shared the 2009 Nobel prize in economics for his analysis of the boundaries of the firm, and to Benjamin Klein, Robert Crawford and Armen Alchian, whose 1978 paper gave it its sharpest form. The idea begins with a plain observation: some investments are worth much more in one use than in any other. The gap between what the asset earns in its intended use and the most it could earn in its next-best use is what Klein, Crawford and Alchian called the appropriable quasi-rent: appropriable because, once the investment is sunk, the customer can bargain over that gap, knowing the supplier's alternative is the scrap price. They called what follows postcontractual opportunistic behavior, the problem Victor Goldberg had named the hold-up, and argued that the threat of it decides whether a company makes a part or buys it, and what contract it writes when it buys.
Their famous example was General Motors and Fisher Body. In 1919 GM bought 60 percent of Fisher Body and signed a ten-year contract to buy substantially all its closed car bodies from it; in 1926 it bought the remaining 40 percent. In the 1978 telling, the contract's cost-plus pricing and Fisher's refusal to build its body plants next to GM's assembly plants let Fisher hold GM up once GM depended on it, and GM solved the problem by buying the rest of Fisher. It is a wonderful story and it may not be true. Ronald Coase, who as a young student had toured American plants, General Motors among them, in 1931 and 1932, went back to the record in 2000 and concluded that the hold-up account was wrong: the two companies had worked closely, Fisher had built plants near GM's, and the acquisition had other causes. Klein replied in the same issue of the same journal, pointing to a dispute over the location of a plant in 1925. We take no side. The lesson survives the dispute: an asset built for one buyer gives that buyer power over the seller, and the seller's protection is a contract with teeth, a deposit, or ownership.
The rule for what happens after a write-down is what makes the asset legible. Microvast states it in its 10-K, and it binds every company reporting under American accounting rules, so we quote it word for word: "At the point of the loss recognition, a new, lower cost basis for those inventories is established, and subsequent changes in facts and circumstances do not result in the restoration or increase in that newly established cost basis." American accounting allows a company to confess that its goods are worth less than they cost; it does not allow the confession to be withdrawn, however cheerful the later news.
Take an illustration of our own, in round numbers. A unit costs $100 to make for one customer, who has agreed to pay $130; to anyone else it is worth $30. If the customer walks away, the unit is written down to $30 and stays there. If the customer then comes back and pays the full price after all, the sale shows a margin more than three times what it would have shown on time. The write-down that stays written down turns a late sale into a conspicuous one.
Microvast's written-down storage units are what the economists call a relationship-specific investment, an asset worth most to the one buyer it was built for, whichever customer that was; and the write-down is the quasi-rent's shadow on the books, a shorter one: in the illustration the quasi-rent runs from $130 down to $30, the write-down only from $100 down to $30.
The test has limits. Late will show as margin, a quarter in which Microvast's gross margin jumps above its range, or as a note that the previously written-down inventory was sold to its customer at the agreed price. Lost may show as a sale to another buyer at about the written-down value with almost no margin on it, as a second charge when the units are scrapped or written down further, or as silence. Accountants record a write-down as a reserve set against the goods and release it, that is, take it off the books, only when the goods leave; so a release, if one is ever reported, will tell us only which of those things happened. The first two quarterly reports of 2026 are silent: no sale of the written-down units, no release, and finished goods down $9.7 million over the half year, to $17.9 million, with no cause given. Gross margin in those quarters was 31.6% and 29.5%, below the 34.7% to 37.6% of the first three quarters of 2025, and nowhere near a burst. The units, as far as the filings show, are still waiting.
The second Library model is the one Inquiry No. 18, The Paper Railway, used to read the cancellations of data center power agreements, and it comes from climate science, where Klaus Hasselmann shared the 2021 Nobel Prize in physics for it and called it optimal fingerprinting. Here it runs across one company's product lines: Photronics reports four lines and five named regions every quarter, and its management named the candidate causes itself.
Three causes were named for the fall in the second fiscal quarter, and we add a fourth that management did not name and a skeptic must. Each would leave a different mark, and we did not write the marks down before the second quarter's figures were published, so the reading below is a fit made after the fact. Crowded fabs that slow the release of new designs should fall hardest on high-end chip masks, because advanced designs need scarce slots at the leading foundries, should touch mainstream chip masks less, and should leave display masks untouched, because display makers are different customers in a different industry; by region it should show in Taiwan, where the foundries are. Memory shortages raising costs for device makers should delay device launches, touching high-end and mainstream chip masks together and perhaps high-end display masks too, since the same devices need screens. Geopolitical uncertainty should spread across lines and regions alike. Lost demand across the board should show as a broad fall that does not come back, with the high-end share falling as well as the dollars; designs lost at the high end to a rival, or to a captive mask shop, one a chipmaker runs for its own designs, would look like the crowded-fabs pattern, and only the cumulative test can tell the two apart.
Hasselmann's second instruction, to weight toward the quiet lines, shapes the reading here without deciding it. Each product line has its own ordinary quarter-to-quarter swing, measured over the quarters before the fall. The mainstream display line is small and lumpy, between $6.7 million and $12.8 million a quarter since fiscal 2024, so its 25.1% fall carries almost no information; a line that noisy falls by a fifth, or rises by half, in ordinary quarters. High-end display is large and rose by about one and a quarter times its usual swing. High-end chip masks fell by a fifth, nearly twice their usual swing; mainstream chip masks fell by a thirtieth, about half of theirs. Photronics reports regions by where its own plants earned the revenue, so Taiwan means its Taiwan plants, which serve the foundries there; Taiwan fell by an eighth, the United States by a twelfth, and China, Korea and Europe hardly at all. Taken together, the lines lean toward the crowded-fabs cause: the largest move, and the only one well outside its own noise, was the high end's fall, where that cause says it should be. They do not settle it: in the third quarter Taiwan recovered 5.3% while China fell 12.5%, mainstream chip masks fell again, and the company blamed softer display sales on launches held back by tight memory supplies, so the memory cause is showing up as well. The fifth of our frozen tests is therefore a prediction about the quiet lines in the quarters to come: if the cause is what the fit suggests, high-end display and most regions should hold; a broad fall across lines and regions would mean the cause had changed to demand, Test E would fail, and Test D would be void, set aside unscored.
One honest caution. The fingerprint tells a delay from a general loss of demand, and no more. Two lighter instruments say the rest: how a delay that never ends should read in a sum, and how far to trust a company's own calendar.
Two more Library models bear on the case. The first is time-to-event, the survival analysis of medicine and engineering, which follows a group of patients or parts and asks what share is still waiting at each moment. For a set of delayed designs, the curve of the share still unreleased should fall toward zero if every design eventually ships; if it flattens above zero, the height at which it flattens is the share that never comes back. Photronics cannot report that curve, since its customers' pipelines are not its to disclose, but the five-quarter total reads it where it ends, assuming a delayed design pays for its masks once, when released: a curve that flattens at 10% leaves the total short by a tenth of the delayed designs' revenue.
The second is the planning fallacy, Daniel Kahneman and Amos Tversky's name for the human habit of forecasting a project from the inside, by its plan, instead of from the outside, by how long projects like it have taken. Microvast's own record supplies the outside view. In 2021 it bought a building in Clarksville, Tennessee, to hold up to 2 gigawatt-hours of capacity it "initially expected to be in operation by the third quarter of 2023"; in 2024 it suspended construction for lack of money, and now expects a smaller pack assembly line there to deliver in 2026. Phase 3.2 is the subject of our third test. The 10-K, filed in March 2026, said "We expect equipment installation to conclude in 2026, followed by commissioning and pilot production," while the same day's release anticipated "serial production in 2026 after the ramp up period." Commissioning is the testing of the installed machines, pilot production means small trial runs, and serial production is steady output for sale. By May commissioning and pilot production were under way; by August they were complete and the company anticipated continued ramp-up in the second half. Each report has recorded progress, and each has described a line not yet in serial production. Against the company's own history the outside view says the date slips, which is why the test asks for the 10-K to say serial production in so many words, and for the batteries sold in 2026 to exceed the 1,879.5 megawatt-hours sold in 2025. That second bar is high. The first half of 2026 sold 717.2 megawatt-hours against 947.2 a year earlier, so the second half must sell 1,162.3, 24.7% more than the 932.3 of the second half of 2025.
One caution, because the figure is easy to misread: the 1,879.5 megawatt-hours is what Microvast sold in 2025, and the company reports no production volume and no utilization rate for any line, old or new.
The same question reaches the licensors of the reusable circuit blocks designers buy, whose license fees follow new designs the way a mask maker's revenue does; the makers of tooling and dies, built to one customer's part; the contract manufacturers who put up a line for one product before it is launched. Each builds ahead, each reports a hole that a deferral and a loss fill identically, and each has somewhere in its filings a number, a running total or a margin, that would later tell the two apart. We name no company, having checked none at source, and we confess a hope that a Vista client in one of these trades commissions us to do it properly, primary research and all; it would be, we say with a nearly straight face, the best study of its kind in the world. The method is the thing: find the relationship-specific asset, find the rule that governs its value on the books, and write down before the data what a late sale and a lost one would look like. The rule is American. Under the international standard, IAS 2, a company must reverse a write-down, up to the original cost, when the goods regain value, so at a supplier reporting that way the late case shows first as a reversal of the charge, and the place to look moves with it. Both of our companies report under American rules, so for them the place holds.
So the quarter in which Microvast sold $96.4 million of batteries and kept about $1 million of it, and the quarter in which Photronics called 44% a record while the dollars behind it sat below the record, pose the same question from two sides of the ledger: a supplier that planted, and a harvest that had not yet come in. Both companies blamed delays at their customers. By the dates we have set, their own books should show whether delay was the right word, or only the first; if they stay silent, we will say so.
None of what follows is a recommendation to buy, sell or hold anything. Each hypothesis carries the evidence for and against it; the result that would falsify each is in the frozen tests below.
Microvast: Europe's revenue was deferred, so that Europe's 2026 revenue in the 10-K's geography table is at least the $211.9 million of 2025. For it: a European second quarter 34.7% above the same quarter of 2025, and a European backlog. Against it: a first half 3.3% behind, and the need for a second half larger than 2025's. Test A.
Microvast: the written-down storage units will sell at the agreed price, so that the 2026 10-K shows a sale to the customer they were built for, or at about that price, or a burst of gross margin, which the write-down rule produces only if the units sell well above their written-down cost. For it: little. Against it: two silent quarterly reports, the decision to defer development of the product line they belong to, and the one storage contract the filings describe, which ended with the customer repaid. Test B.
Microvast: Phase 3.2 reaches serial production in 2026, the company's own hypothesis. For it: the progress each quarterly report has recorded, from installation to commissioning. Against it: the record at Clarksville, and each report so far describing a line short of serial production. Void if a restructuring or going-concern event halts the line, which after August is a live possibility. Test C.
Photronics: the delayed designs were deferred, not lost, so that five quarters of high-end chip mask revenue, from the quarter of the fall through the fiscal second quarter of 2027, total at least $356.41 million. For it: the fingerprint, which leans toward a delay, and a third quarter within 3.9% of the record. Against it: three quarters that must average $77.08 million, above any the company has reported, and the company's own words for the recovery, "some" and "modest." Test D, void if foundry utilization falls by the measure stated there.
Photronics: the quiet lines hold, high-end display and most regions, as a delay confined to advanced chip designs would predict; a broad fall across lines and regions would mean the cause had become demand, and would void the cumulative test. Test E.
What follows was frozen on October 30, 2026, before Microvast's third-quarter results, expected about mid-November, and before Photronics' fiscal fourth-quarter results, expected about early December. Tests B, D and E were re-frozen the same day after a fact-check, still before either report; Tests A and C were unchanged. Every figure comes from the program we wrote for this inquiry, which checks each threshold and refuses to run if one is changed. "Cannot tell" is a legitimate result for every test and will be reported as one.
Microvast is scored on its 10-K for 2026, expected about March 2027, and its fourth-quarter release.
Test A, Europe. The geography table's label is "Europe"; the discussion calls it EMEA; 2025's figure is $211.944 million. Confirm, deferred: 2026 Europe revenue at or above $211.944 million. Falsify, lost: below $201.347 million, more than 5% under. Cannot tell: from $201.347 million up to $211.944 million. Void if the company changes the geography table so that 2026 and 2025 cannot be compared like with like. Where it stands: first-half 2026 Europe revenue $95.625 million against $98.935 million, 3.3% behind; the second half needs $116.319 million to confirm, against $113.009 million in the second half of 2025. Korea and India stay out, a different mechanism.
Test B, the written-down storage inventory, $32.507 million in 2025, Note 4. Confirm, deferred: the filing discloses that the previously written-down ESS units were sold to the customer they were built for, or at about the price first agreed, or reports a quarter whose gross margin was lifted by such a sale. Falsify, stranded: the units are scrapped or written down further, or sold to another buyer at about their written-down value, or new customer-specific ESS impairments follow. Cannot tell: the filing is silent. A reserve cannot be released before the goods are sold or scrapped, so a release tells us only which of those happened. If the units are the containers of the Clenera contract, the customer they were built for has been repaid, so only a sale to someone else at about the agreed price could confirm; we say so now. The first- and second-quarter reports of 2026 are silent so far.
Test C, Huzhou Phase 3.2. Confirm: the 10-K states serial production and 2026 megawatt-hours sold exceed 1,879.5, the 2025 sales volume. Falsify: the line is still described as commissioning or pilot production. Cannot tell: serial production stated, volume not exceeded. Void: a restructuring or going-concern event that halts the line.
Photronics is scored through its fiscal second quarter of 2027, with results expected about late May or early June 2027.
Test D, cumulative high-end chip mask revenue, with the 10-Q definition of 28nm or smaller held fixed, over the five quarters from the fiscal second quarter of 2026 through the fiscal second quarter of 2027, taken from the product tables. Confirm, deferred: at or above $356.41 million, five quarters at the fiscal first-quarter 2026 record of $71.282 million. Cannot tell: from $340.90 million up to $356.41 million; the lower figure is a constant growth rate fitted to the nine quarters from the fiscal first quarter of 2024 to the fiscal first quarter of 2026, 1.95% a quarter, carried across the five test quarters. Falsify, lost: below $340.90 million, provided the foundries are still full. Still full means this. United Microelectronics and Semiconductor Manufacturing International, two large publicly traded foundries, each publish a capacity utilization rate every quarter, and the most recent rates they have published before Photronics' fiscal second-quarter 2027 results must be, on average, no more than five percentage points below the rates they published for the calendar second quarter of 2026, the quarter of the fall: 85% at United Microelectronics and 93.7% at Semiconductor Manufacturing International, an average of 89.35%, so the line sits at 84.35%. Taiwan Semiconductor describes its utilization only in words and is not used. More than five points below that line, and the cause has changed and the test is void. Where it stands: two quarters in, $125.161 million; the remaining three quarters need $77.08 million a quarter to confirm and $71.91 million a quarter to avoid falsifying, both above the record.
If the designs delayed in the fiscal second quarter were only late, the five quarters should add up to at least what five quarters at the record would have brought, and the chart shows how far the first two have come.
Why the two paths differ deserves a sentence. The flat path holds the February record for five quarters; the trend path carries the pre-fall growth of about 2% a quarter forward. Because the record quarter sat above its own trend, pulled forward by demand ahead of the holiday in a quarter the company says is usually its slow one, the flat line is the higher bar, and the band between them is where a real but incomplete recovery would land. We call that band "cannot tell" and mean it.
Test E, the quiet lines, a prediction about the quarters still to come. If the cause is crowded fabs and slow design releases rather than lost designs, the lines the delay did not touch should hold. Holds, confirmed: in each of the fiscal fourth quarter of 2026 and the first and second quarters of 2027, high-end display revenue stays at or above $43.613 million, its lowest quarter of fiscal 2025, and no more than two of the five named regions, Taiwan, China, South Korea, the United States and Europe, fall more than 10% from the quarter before. Falsified, a broad fall: a quarter in which three of the four product lines and four of the five named regions are each down more than 10% from the quarter before; a fall that broad points to demand itself, and voids Test D. Cannot tell: neither.
The data that we're missing is...
Likely, we can get some of it from primary research, which is what Vista is all about: asking the people who saw the programs and the pipelines from the inside. Four things would settle the most.
The first is the storage customer. Microvast does not say which customer the written-down storage inventory was built for, its quantity, or its value on the books today; the one storage contract the filings describe, Clenera's, ended in an award, a settlement and a refund of the customer's advance. Whether the written-down units are those containers, and whether units built to one site's specification can be sold to anyone else at more than scrap, is what decides Test B, and the filings do not say.
The second is the European platform schedules. Microvast's customers are vehicle makers whose platform dates are their own. The filings say the delays were the customers', and the backlog is European, but no filing gives the dates the platforms were due, the dates they are now due, or which of them has since launched. Those dates are the difference between a second half of $116 million, which confirms, and one below $106 million, which falsifies.
The third is the design pipeline. Photronics says its customers delayed releases; it does not say how many designs, at which nodes, for which end markets, or how many have since taped out. A tape-out planner at a chip designer, or an account manager at a foundry, knows whether a delayed design was rescheduled or quietly shelved, and that is the one number the cumulative test is waiting a year to learn.
The fourth is fab utilization itself, which the falsify clause of Test D depends on. The published rates of the two foundries named in Test D are the source we will use; what they cannot say is whether a given customer's design was delayed by the factory's calendar or by the customer's.
The people who negotiated these supply agreements, scheduled these tape-outs and ran these programs know what the filings cannot say, and finding and asking them is what Vista does. Every interview is a paid written conversation with Vista's research desk, at the expert's hourly rate, never a call. We would most like to hear from these. Like any good reporter, we will also try to learn whose order the written-down storage units were, and why it ended; the filings do not say whether they are Clenera's.
As a matter of general past practice, when a vehicle or storage program slipped, what happened to the units the supplier had already built to its specification, who bore the cost, and how often were they eventually taken at the original price? We would also ask what deposits were usual, and what take-or-pay terms, under which a buyer pays for an agreed quantity whether or not it takes delivery, and what share of slipped programs launched within a year, within two, or never.
The answer that would change our view"Units built to a customer's specification were almost never taken late at the original price; the supplier discounted them to a second buyer or scrapped them, and the program that slipped usually changed its specification when it came back."
As a matter of general past practice, how were builds scheduled against customer platform dates, and what protection did a supplier usually take when it built ahead? We would also ask how long platform delays in the industry typically ran and whether delayed volumes were usually recovered in the following year.
The answer that would change our view"Builds ran ahead of firm orders as a matter of course across the industry, and delayed volumes were usually replaced by new programs rather than recovered."
As a matter of general past practice, when customers reported delayed tape-outs, what share were released within one quarter, within three, and never? We would also ask how a forecast of releases was built from customer pipelines, how often a delayed high-end design came back at a different node, and whether share moved to a rival mask maker during a delay.
The answer that would change our view"Of the tape-outs customers called delayed, a third never came to us at all; they were cancelled, folded into a later design, or went to a captive mask shop."
In general, when a fab was full, how were new design starts prioritized against production of existing designs, and how long did a new design wait? We would also ask whether a design that waited two quarters was usually released as planned, redesigned for a different node, or dropped.
The answer that would change our view"When the fab was full the new designs waited, and a meaningful share of those that waited more than two quarters were never started."
As a matter of general past practice, what made a company postpone a tape-out, how long a postponement typically ran, and how the cost of masks figured in the decision to release or cancel a design.
The answer that would change our view"We postponed tape-outs when memory prices made the product unprofitable, and most of the postponed designs were cancelled when the next generation came due."
Every conversation would draw on past, general experience only: no current employees of the companies named, nothing confidential, nothing about either company's internal decisions, and every Advisor, as Vista calls the experts it interviews, screened for conflicts before a word is said. For a client engagement, Vista would combine this inquiry with those interviews: two or three Advisors, structured written conversations, and one brief that says where they agree, where they split, and what would change the answer.
The bet. A reader who takes Microvast's account of 2025 at face value is betting that the European platform delays were delays, that Europe's 2026 revenue matches or beats 2025's $211.9 million after a first half 3.3% behind, that the storage inventory written down by $32.5 million finds a buyer at about the price once agreed, and that the second Huzhou line reaches serial production while the company still has the cash to run it, at a company whose own August report says substantial doubt about its going concern has not been alleviated. A reader who takes Photronics' account at face value is betting that the designs delayed in the spring of 2026 will be released rather than cancelled, so that five quarters of high-end chip mask revenue total at least $356.41 million, which requires the next three quarters to average more than the company has ever done in one.
The payoff. If Microvast's bet is right, the proof arrives as a European revenue figure at or above $211.9 million and, if the storage units sell at the agreed price, as a quarter of gross margin far above the 29% to 32% of 2026 so far; if it is wrong, the proof arrives as a shortfall in Europe, a resale at the written-down value or a second charge on the storage units, and a line still described as commissioning. If Photronics' bet is right, the five-quarter total clears $356.41 million and the high-end share's rise will have come from more releases; if it is wrong, the total falls short while the foundries stay full, and the record share will have been the mainstream line shrinking.
Our assessment. For Photronics the evidence leans toward deferral and does not yet prove it: the fit across the product lines favors a delay caused by crowded fabs, the third quarter recovered most of the fall, and the company's own words are "some" and "modest." For Microvast the evidence so far runs against the bet: Europe is behind for the year, the storage units have been silent for two quarters, the company is deferring development of the product line they belong to, the one storage customer it describes has been repaid, and its own going-concern language has moved the wrong way. In the Pump Room our revenue turned out to be deferred and not lost, and nothing on our books could have told you so in advance; the difference here is that we have written down, before the data, what would.
What settles it, and when. Microvast's third-quarter results in mid-November, unscored but a signpost for Europe; its 2026 10-K and fourth-quarter release, about March 2027, scored against Tests A to C; Photronics' fiscal fourth-quarter results in early December and each 10-Q through the fiscal second quarter of 2027, scored against Tests D and E; and, before any of that, the conversations above with the people who bought the batteries and scheduled the tape-outs.
This inquiry began with two securities complaints the Vista Research Desk read in October 2026, Graham v. Microvast Holdings, Inc., filed in the United States District Court for the Southern District of Texas on July 21, 2026, and Cooper v. Photronics, Inc., filed in the United States District Court for the District of Connecticut on July 6, 2026. The complaints are the provenance of the question: "Where this began" says in one sentence what each alleges, and this inquiry takes no position regarding either case, says nothing about what anyone knew or when, rests nothing on either complaint, and uses only the companies' own reports to the Securities and Exchange Commission. The Vista Library models it uses are make-or-buy and fingerprinting a cause, with time-to-event and the planning fallacy in support.
In the quarter the delay happens, a sales figure cannot tell. It records what did not arrive, not whether it is coming. This inquiry names, before the results, the place in each company's own later reports where the answer will show: for Microvast, its gross margin and its revenue from Europe in its annual report for 2026, due about March 2027; for Photronics, the sum of five quarters of sales of masks for the most advanced chips, complete about June 2027.
Microvast cut the value of storage units it had built for one customer by $32.5 million in 2025, $29.0 million of it in the fourth quarter. American accounting never lets a company reverse such a cut, so the lower value becomes the units' cost for good. If they later sell near the original price, the sale will show as an unusually large profit; if they sell near the written-down value, it will leave little trace.
Masks for the most advanced chips would have to bring at least $356.41 million over the five quarters from the fiscal second quarter of 2026 through the fiscal second quarter of 2027, which means averaging $77.08 million a quarter for the last three, more than Photronics has sold in any quarter. A broad fall across its other product lines and regions would point to weaker demand instead.
Every figure in this inquiry is computed in code from the companies' filings, with each input traced to its source in the inquiry's working files. Microvast's figures come from its 10-K for 2025, its 10-Qs for the first and second quarters of 2026 and its 10-Q for the third quarter of 2025, and its earnings releases for the fourth quarter of 2025 and the first and second quarters of 2026, all read at the SEC's site; the run rate is four times the second quarter; days of sales are net receivables over annual revenue times 365, our computation and not the company's; the $3.5 million of the storage charge outside the fourth quarter is reconciled by arithmetic to the third-quarter provision and the year's standard write-downs, which no filing states in those words; the Clenera matter is stated from the 10-K's commitments and contingencies note, its accrued liabilities note and the two 2026 10-Qs; the Nasdaq notice, the $1 minimum and the cure period are stated from the Form 8-K of August 31, 2026. The round-number illustration of the write-down rule is ours. Photronics' figures come from its 10-K for fiscal 2025 and fiscal 2024, its 10-Qs from fiscal 2024 through the third quarter of fiscal 2026, and its earnings releases and investor decks for the fiscal third quarter of 2025 through the fiscal third quarter of 2026; the fifteen-quarter series of product-line revenue is built from the 10-Q tables, with fourth quarters derived as the fiscal year less nine months and checked against the 10-Ks' own quarterly tables; the trend is a constant growth rate fitted to the nine quarters before the fall; the flat path holds the fiscal first-quarter 2026 record for five quarters; each product line's ordinary swing is the spread of its quarter-to-quarter changes over the quarters before the fall, fiscal 2023 through the first quarter of fiscal 2026; regions are as the company reports them, by the location in which the revenue was earned. No earnings call transcript was used, and nothing said only on a call appears here. Share prices are daily closes from Massive market data through October 30, 2026; analyst ratings and targets are as StockAnalysis, TipRanks and MarketBeat reported them on November 1, 2026. The tests were frozen on October 30, 2026 in the model, which asserts every figure and every band; Tests B, D and E were re-frozen the same day after the fact-check, before either company's next report.
This inquiry is research, not investment advice. It states what the filings and models show under labeled assumptions; nothing here is a recommendation to buy, sell or hold any security, and the decision belongs to the reader. Russ W. Rosenzweig, Vista's founder, owns no shares of Microvast, Photronics or General Motors and holds no options on any of them. He owns shares of Taiwan Semiconductor (TSMC), named once in Test D only as a foundry that publishes no utilization rate. Holdings through mutual funds and exchange-traded funds are not counted. How this inquiry was made: written by Russ W. Rosenzweig with Vista's AI research desk. It began with two securities complaints the desk read in October 2026, named in "The idea," which prompted the question and are described in one sentence each as allegations; the facts were gathered from Microvast's and Photronics' filings and releases at the Securities and Exchange Commission, every figure was recomputed in code, the five tests were frozen before the next results, and the draft was fact-checked and read for plainness before I read it. I read it and listened to it before it was published. Why I write with AI. Why I publish these inquiries.