PSIX: The Data-Center Darling That Fell 74% and Still Has a $70 Price Target

Somebody at Power Solutions International needs to explain the math. The stock ripped 385% in 2025, rode the data-center power narrative to a 52-week high of $121.78, and then proceeded to lose three-quarters of its value, bottoming out at $31.76. And yet the analyst consensus target still sits at $70.37 — implying the Street thinks this thing more than doubles from here. Somewhere between the euphoria and the epitaph is the actual truth about this company, and August 6th is when we find out which story is real.

Let's start with what made people fall in love with PSIX in the first place. This is an Illinois-based engine and power-systems maker, founded in 1985, that suddenly found itself standing at the intersection of "boring industrial manufacturer" and "AI infrastructure play." Data centers are projected to consume something like 1,600 TWh of electricity by 2035, and PSIX makes the packaged power systems — engines, cooling, electronics, the whole vertically integrated kit — that could plug into that build-out. Add a 21.2% forecasted annual revenue growth rate through 2029 and a trailing P/E under 10x at various points, and you can see why quant screens and momentum chasers piled in.

Then reality showed up. Q1 2026 numbers, released May 11th, were fine on the surface — $128.6M in net sales, $7.3M net income, $0.32 diluted EPS — but underneath, margins are cracking. Net profit margin slid from 15.7% to 14.3% in a year. Oil-and-gas demand, still a real chunk of the business, has softened. Ramp-up costs at the Wisconsin facility remain "elevated," which is corporate-speak for "we're spending more than we planned and it's showing up on the income statement." None of that screams data-center supercycle. It screams a legacy engine maker trying to retrofit a growth story onto a cyclical business.

And then it gets uglier. Two separate securities class-action lawsuits, filed in March and April 2026, allege the company misrepresented the strength of its data-center demand and glossed over manufacturing inefficiencies. Say what you want about the merits — lawsuits get filed on flimsier grounds all the time — but the optics are brutal when you pair them with the fact that CEO Constantine Xykis sold his entire stake back in November 2025, right as founder Gary Winemaster was also trimming his position. Insiders selling into strength while the marketing deck talks up hyperscale demand is exactly the pattern that makes plaintiffs' attorneys salivate and makes me want to keep my hand on my wallet.

Here's where it gets genuinely interesting, though: the stock is cheap on paper. Non-GAAP P/E around 13x, revenue growth still north of 38% year-over-year, trailing EPS of $4.44, ROE reportedly around 76% historically. This isn't a broken shell — it's a company with real unit economics that the market has decided to punish first and ask questions later. Short interest is reportedly running near 20% of the float, which means any credible good news out of the August 6th print — margin stabilization, a clean read on order-book conversion, anything that quiets the "data-center demand was fiction" narrative — could trigger a genuinely violent squeeze. That's the coiled-spring case.

But binary events cut both ways. If management shows up on August 6th and margins keep deteriorating, or if guidance confirms the lawsuits' central allegation that data-center orders were oversold, this stock has plenty of room to keep falling — support isn't even clearly established below $30. Beta of 2.00 tells you this name doesn't do anything halfway.

My read: this is not a "back up the truck" situation, and it's not a "run for the hills" situation either — it's a "wait for the tape to tell you something" situation. The valuation is genuinely tempting. The insider selling and litigation are genuinely disqualifying until proven otherwise. Accumulate on dips toward the low-$30s if you must be involved, but don't confuse a cheap stock with a safe one. PSIX has already taught this year's momentum crowd one brutal lesson about the gap between a narrative and a balance sheet. Don't make it teach you a second one.