The NoScript Test

The NoScript Test

A strange thing happened while I was reporting this. I wanted to read a market-research firm's note on how to choose AI stocks, and the page wouldn't let me. An anti-bot screen took over: enable JavaScript, it said, or there is no article. Just an empty page making a demand.

That is nearly the whole story of the 2026 software sell-off in one screenshot.

In the first two months of 2026, software stocks fell as if someone had thrown a switch. The MSCI index tracking the world's big software companies lost more than a fifth of its value in that window. The stated cause was AI: a burst of new AI coding models had convinced a lot of investors that software itself was about to become a commodity, that the thing underneath every interface was about to be automated away. Money managers who had spent three years paying for a narrative suddenly demanded something they could audit. Non-GAAP earnings — the adjusted version that lets a company set aside stock-based compensation and report a smaller loss — stopped being acceptable as a basis for a price. They wanted GAAP profits, the plain number.

There is a way to think about this that is not about fear at all. A web page is made of two things. HTML is the content — the text, the structure, the actual information. JavaScript is the behavior — the drop-downs, the smooth scrolling, the charts that make a plain document look like a product. Turn off JavaScript and the page still renders, but a lot of it goes dead, and you discover what the page is really made of. Some pages are rich documents that used the script only for decoration. Others were mostly script all along, and what survives is a little text and a lot of white space.

The market turned off JavaScript on AI software in 2026. What you now see in every chart, every earnings release, every multiple is the raw HTML. The honest question is which companies were documents, and which were script.

The cleanest test case is a company whose script was visible even in its ticker. C3.ai trades under the symbol AI on the New York Stock Exchange, which tells you what its product was, sort of. It went public in December 2020, at the crest of the first AI frenzy, and at its peak traded at more than ninety times sales. The story — a pure play on the adoption of artificial intelligence by big companies — did the heavy lifting for years. Here is the tell: during all of that, the underlying business grew at an annualized rate of about 6 percent over five years, and across the last two years it was already shrinking. Even at the height of the boom, the document was thin. A perfectly ordinary software business was wearing an AI costume, and the costume was priced as if it were the whole company.

The costume started coming apart last autumn. C3.ai's fiscal year ends in April, and fiscal 2026 was a demolition. In the first quarter, the company reported $70.3 million of revenue, missed estimates, and withdrew its fiscal 2026 guidance. The reasons given were a sales reorganization and the health of the founder and longtime CEO, Thomas Siebel, who stepped down from running the company.

By winter the decline was unmistakable. Revenue fell 46 percent to $53.3 million in the quarter ended in January — against the $72 million to $80 million the company itself had forecast three months earlier. In the spring quarter, the reported revenue of $51.6 million was down 52.5 percent from a year earlier. That is not a rough patch. That is a business that lost half its revenue in a year.

Here is the detail that should make you distrust every AI headline you read for the next year. When C3.ai reported that quarter in June, analysts had marked their own estimates down so far that the company beat them — revenue came in 2 percent above consensus — and the stock jumped 11 percent. The market celebrated a number that was only a win because everyone had already rewritten their expectations to match the new, smaller reality. Beating a floor that investors built themselves because they no longer believed the story is not strength. It is an echo.

For the whole fiscal year, revenue came in at $250.3 million, down about a third from the year before. The company reported a GAAP operating loss of $498.5 million — nearly double its annual revenue, once you recognize that the plain number includes about $264 million of stock-based compensation. That is exactly why the company prefers the non-GAAP version, and exactly why the market has stopped accepting it.

The response was a crash diet. Management laid off 26 percent of its staff, flattened the sales organization, and promised $135 million of annual cost savings. In May, Siebel came back as chief executive, citing his health as largely resolved, and described the recent sales results with the kind of bluntness that gets quoted everywhere: entirely unacceptable, to the point of surreal. Then the company guided to about $225 million of revenue for fiscal 2027. Read that again. The leadership's own plan is for the business to get smaller for another year.

This is the moment where the story turns seductive, and you should watch it closely. C3.ai now looks cheap by every measure that has ever made a stock look cheap. In April, with the stock down about 40 percent for the year, the price-to-sales ratio — share count times price, divided by annual revenue, a rough way to compare software companies — sat at 3.7, near the lowest it has ever recorded and far below the ninety-plus times sales of the 2020 peak. When a famous company trades at its lowest-ever multiple, the reflex is that the market has overdone it.

The reflex is wrong, because a price-to-sales ratio is only as meaningful as the number underneath it. C3.ai's revenue is not static; it is falling fast. Today the company is worth about $1.35 billion on roughly $250 million of trailing revenue — about 5.4 times sales, by my arithmetic. The stock went down and the multiple went up, because the denominator fell faster than the price. If revenue halves again while the price doesn't move, the multiple doubles without the headline price getting uglier. "Cheapest ever" and "shrinking underneath" are describing the same moment, and the valuation tells you nothing about which one you're buying.

Strip out the roughly $575 million of cash and investments the company still holds and the operating business sells for about $770 million — around three times trailing revenue that is still shrinking double-digit, with no profit in sight. That is not a stage for a turnaround story. Paying that for a company that cannot say when the decline ends is a bet, not a bargain.

Bulls will tell you the pieces are there: cash and investments of $575 million against a $1.35 billion market cap, $135 million of annual cuts, and a legendary founder back behind the wheel to sell again. I used to put a lot of weight on that last part, so I watched what actually happened after May. The stock has not recovered; it trades near the bottom of its 52-week range, below its 200-day moving average, around $9, in what investors call a downtrend. And the insider filings tell a quieter story. This summer the returning founder has been selling shares through the Form 4 filings that insiders must file when they trade their own company's stock, including an August sale worth about $4.8 million. When the person whose return was supposed to be the whole fix is selling into the fix, the page has not re-rendered.

But here is the detail that keeps this from being just a short story about one broken company. Not every part of C3.ai was script. In the winter quarter, bookings in its federal, defense and aerospace segments rose 134 percent and made up more than half of all new contracts. Governments buy software through formal procurement, on requirements and contracts, not on PowerPoint charisma. That part of the company never needed the AI costume and never needed the founder to charm a CEO over dinner — and it is the only part still growing. The commercial pipeline, the part that was sold on the story, is the part that emptied out. That is the cleanest map of what the movie was showing: when you strip the script, the order that survives is the one that never ran on narrative.

Now step back, because this matters well beyond one ticker. The market turned off JavaScript on all of software in 2026, and it hit strong companies as hard as weak ones. Palantir spent the year posting record quarters and watching its stock fall more than 40 percent from its high — a real document getting repriced as if it were script. That is the danger of reading every low price as a bargain: the same sell-off contains both a genuine discount and an honest photograph of a broken business, and the price alone will not tell you which is which. The labels lag too. AInvest's aggregate signal still labels C3.ai a Hold, and the consensus the market trades on still models per-share losses in the twenty-to-thirty-cent range for the quarters ahead. The ratings and the forecasts are all rendering the document as it looked three years ago, not as it is.

So here is the test I now run on any washed-out stock, and you can run it too. Call it the NoScript test. First, strip the label: pretend the company has a boring name and a meaningless ticker, and ask what it actually sells and whether customers would keep paying if the CEO retired and the analysts deleted their price targets. Second, measure the multiple against a denominator that is actually holding — check whether revenue grew last quarter, rather than whether the multiple is low. Third, treat a "beat" as real only if the expectations were set by reality, and not by a market that already stopped believing and marked every number down to the floor.

For C3.ai specifically, the number worth watching is narrow and it is not the price. It is whether the commercial decline has stopped — whether revenue can hold around the $52 million per quarter that management now guides to, and whether the government strand keeps growing fast enough to matter. If the decline halts, the current price makes a turnaround story credible. If it keeps halving, the "cheapest ever" multiple will quietly become "even cheaper than we thought." The market turned off the script, and it has been honest since. The least you can do is read what it is showing you.