A small U.S. systems integrator posted 523% revenue growth in early 2025 by assembling complete liquid cooled AI racks. The SEC filings it sits inside reveal how narrow its customer base really is.
The recurring trap in small-cap AI infrastructure is the gap between a capacity announcement and the revenue that capacity actually produces. TSS, Inc., a U.S. systems integrator that assembles complete liquid-cooled racks for AI data centers, just turned in the kind of quarter that makes the gap easier to measure. Revenue was $99.0 million in the first quarter of 2025, up 523% year over year, with EPS of $0.12 against $0.00 a year earlier. Whether that line keeps compounding is now a falsifiable question, because the company has announced a new factory to expand capacity, filed a second earnings quarter, and put a fresh SEC annual filing on the table.
TSSI sits in a specific lane of the AI infrastructure buildout. It does not make the GPUs, the coolant chemistry, or the rack-scale heat exchangers (CDUs) that move facility coolant to individual servers. It integrates them. Management's positioning, captured on the company's Q1 2025 earnings call and reinforced in its November 2025 investor presentation, frames the product as a finished rack: chips, networking, power conversion, and rack-scale cooling plumbed into a unit a data-center operator can install. Liquid cooling is not optional for AI training clusters above a certain density, because air handling cannot shed the heat. That is the wedge the 523% growth line is built on.
Two questions make the story falsifiable. How concentrated is the demand behind that number, and how exposed is the supply chain delivering it. The SEC annual 10-K filing is where those answers should appear, in the customer-concentration risk factor and the supplier-dependency disclosures. The first-quarter 2026 release will show whether the new factory floor translated into sequential revenue growth or whether the ramp absorbed the costs and supply-chain friction hardware integrations typically face when chip or CDU lead times slip. The surrounding paperwork, including the company's press-release index and a separate SEC exhibit filing, fills in the customer and end-market detail that any capacity story has to clear.
The lesson worth pulling out is broader than this one ticker. When a small-cap integrator announces AI infrastructure capacity, the falsifiable evidence lives in three places: order-book concentration disclosed in 10-K risk factors, sequential conversion of announced capacity into revenue quarter over quarter, and supplier concentration on the GPU and CDU inputs. TSSI's filings will show that, or they will show the limits of the story. Either result is a more useful data point than the 523% line on its own.