In 1937, seven million cubic feet of contained hydrogen ignited over Manchester Township, N.J., bringing down a German airship the size of a football field in half a minute as nearby movie cameras whirred. "This is the worst thing I've ever witnessed," said one broadcaster.
The vessel's name, Hindenburg, never caught on as a popular brand, as you might imagine. There's no Hindenburg peanut butter or antacid tablets. You surely wouldn't want the name for an investment business. Unless you're a short seller. This unusual breed bets against stocks, and then publishes reports convincing others that those stocks are unrelenting disasters. A short seller would object to naming his firm for the Titanic on the grounds that there were 706 survivors.
And so it is that there's a short seller called Hindenburg Research that has gone after an artificial-intelligence highflier called Super Micro Computer. It isn't super for Super, because apart from Hindenburg's name, there's its record. Four years ago in this column, I reported on my conversation with Trevor Milton, founder of clean energy big-rig start-up Nikola, which was pushing back against a searing Hindenburg report. Investors who bought shares would have "one of the funnest rides they've ever had," Milton told me. I recommended that readers "sit this fun ride out." Shares are down 99%. Milton was sentenced late last year to four years in prison for securities and wire fraud.
In May of last year, Hindenburg issued a report arguing that the unit price of publicly traded Icahn Enterprises was "inflated by 75%." It's down 74%.
Hindenburg's latest turd blossom, dated this past Tuesday, alleges that it has found "glaring accounting red flags" at Super Micro, including "evidence of undisclosed related-party transactions." Super Micro hasn't commented, but said in a Wednesday filing that it will delay its annual report to assess "the design and operating effectiveness of its internal controls over financial reporting." Shares have collapsed from a high of over $1,200 in March to a recent $440.
Let's put aside the merits of Hindenburg's specific claims and check in with a guy who turned bearish on Super Micro long before the short attack, to look for clues on avoiding other stock blowups. Mehdi Hosseini at Susquehanna Financial Group studied electrical engineering and designed chips for National Semiconductor, now part of Texas Instruments, before becoming a stock analyst. He has been following Super Micro since 2013, and says his Sell call late last year was early, but also that the company hasn't been trading on fundamentals. "This is a meme stock," he says.
Super Micro says it "delivers the broadest selection of AI systems and solutions," which sounds like an ideal fit for the moment. But then, I can claim truthfully, more or less, that in my 20s, I made my own computer. What I really did was assemble one using plug-and-play parts recommended on some website for cash-strapped nerds. The job took 45 minutes and a screwdriver.
I'm not saying that I know as much as Super Micro. I'm saying that Nvidia knows a heck of a lot more, and it's not really a subjective matter. It's a matter of using research-and-development spending as a proxy for smarts. Preliminary results for Super Micro's latest fiscal year show R&D of $348 million, or 2% of revenue, versus Nvidia's more than $6 billion, or 10% of revenue. These two companies don't bring nearly the same thing to the AI arms race.
Super Micro makes computer systems using its own brand of boxes with key innards from other manufacturers. A partnership with Nvidia made it an AI stock darling. But most of the innovation occurs "upstream," with Nvidia and its chip-making partner Taiwan Semiconductor, says Hosseini. "Super Micro doesn't really do the innovation," he says. "They are a contract manufacturer with willingness to commit working capital." In other words, Super Micro pays upfront for memory, storage, cases, screws, and whatever else is needed for a partner like Nvidia to sell more high-value chips. It's a relatively weak hand that limits pricing power. It's also expensive during periods of extreme growth.
In Super Micro's latest fiscal year, its revenue doubled to $14.9 billion, judging by preliminary, not official, really-don't-rely-too-heavily-on-these-yet results. Earnings were seemingly robust, at $1.34 billion. But free cash flow was negative by $2.6 billion. "Would It Take Another $2B+ of FCF Burn to Double Rev Again?!" asked Hosseini in the title of an Aug. 7 investor note. Super Micro raised funds this year by issuing convertible bonds and shares, and might have to raise more, he wrote.
Earnings and free cash flow are meant to measure roughly the same thing over time. It's just that earnings pretend that the timing of revenue matches up neatly with the timing of associated costs, to make it easy for stock investors to compare the two. With free cash flow, there's no pretending. Nearly 30 years ago, a landmark accounting study found that companies with high "accruals" -- think high earnings and low free cash flow -- go on to produce poor stock returns, on average. More recent studies suggest this "accrual anomaly" has weakened, perhaps because it has been traded away by hedge funds exploiting it.
But free cash flow is still a useful check on earnings. Maybe a frequent cash burner is making a daredevil bet on growth that will pay off marvelously, as it did for Netflix. But maybe not.
If the Super Micro selloff tempts, Hosseini says to set aside the accusations and judge matters, for now, by what is known. The company was delisted by Nasdaq in 2018 for delayed financial reporting, and restated results the following year. Some executives from then have been rehired. The founder is both chief executive and chairman, limiting board independence.
Also, "customers may abandon this partner because of these irregularities," Hosseini says. In other words, Hindenburg could be a disaster for revenue no matter what we learn about its claims.