Snowcap Research is short shares of Sterling Infrastructure Inc (NASDAQ:STRL).
A regional excavation contractor dressed up as an AI infrastructure play.
We are short Sterling Infrastructure, a poster child for the AI bubble in public markets. Data center exposure appears exaggerated. Backlog growth is not supported by contract win data pulled from a third-party publication. Worse, margins appear grossly inflated through abnormal accounting revisions, corroborating overstated DC exposure. Sterling’s valuation is so exuberant that it now looks expensive even compared to AI darlings.
Sterling Infrastructure Inc. (“Sterling” or the “Company”) is not a data center infrastructure company. It owns a collection of regional contractors that specialize in site preparation and excavation services – clearing and grading land before foundations are laid.
In 2022, Sterling rebranded one of its segments as “E-Infrastructure” and began positioning itself to investors as a “picks-and-shovels” play on the AI boom. Smart move as its stock has increased nearly twenty-fold since then, outperforming even marquee AI beneficiaries like NVIDIA.
Management has promoted this message heavily. In its most recent earnings call, management used the term “data center” 14 times. Three years ago, it was barely mentioned.
We think this narrative is a mirage. Subsidiary contract data pulled from Engineering News-Record (“ENR”) – a construction industry publication – combined with a detailed review of Sterling’s underlying projects, and forensic accounting analysis, leads us to believe Sterling has materially overstated the scale of its data center business, has inflated margins through aggressive accounting, and is dramatically overvalued.
We see 60–80% downside from current levels.
1. Exaggerated Data Center Exposure. Management is coy in its disclosures, but their implication is unmistakable: data centers make up the bulk of Sterling’s E-Infrastructure business, and the remainder consists of other high-margin “mission critical”1 customers like chip fabrication plants and “next-generation” manufacturing.
a. In-depth Review of Underlying Projects Pulled from Subsidiary Websites Suggests Number of Data Center Projects Exaggerated by 5x. In a recent earnings call, management indicated that Sterling has worked on at least “100 or so data centers”. We undertook a thorough review of hundreds of individual projects showcased on the websites of Sterling’s subsidiaries. In total, we counted just 18 data center projects undertaken since 2020 – a fraction of the figure implied by management2. We believe the showcased projects are likely exhaustive because we cross referenced them with projects referenced on Sterling’s social media pages and in state planning filings (where available). Even using generous assumptions, we struggle to reconcile the size and quantum of these projects with Sterling’s purported data center revenues.
b. Recent Subsidiary Disclosures in ENR Show Almost Half of E-Infrastructure Revenue is from Low-Margin Warehouses. We uncovered recently published data from Sterling’s flagship subsidiaries Plateau Excavation (“Plateau”) and Petillo Companies (“Petillo”) (together responsible for 90% of Sterling’s E-Infrastructure revenues3), which reveals that 40% of their revenues in FY24 are from generic warehouses. This is alarming because warehouses typically earn much lower margins than data centers and other mission critical projects4. Management’s claim that the “vast majority” of E-Infrastructure revenues are mission critical appears to be nonsense.
c. Petillo: #2 E-Infrastructure Subsidiary Completed Just 1 Data Center Project. Most strikingly, the second largest subsidiary in Sterling’s E-infrastructure segment5 – Petillo – appears to have completed just one data center project since it was acquired in 2021. Petillo is one of Sterling’s trophy subsidiaries, that it apparently has almost no data center exposure is a huge red flag in our view.
d. Project Sizes Have Shrunk by 50%, Crushing Margin Expansion Narrative. By our calculation, the average size of data center projects disclosed by Sterling’s subsidiaries has halved in recent years6. This is a head scratcher as Sterling’s primary explanation for its margins expanding is that project sizes are growing. If project sizes are shrinking, does that imply the margin story unravels?
2. Phantom Backlog: Subsidiary Data Undermines the Growth Story. Sterling’s E-Infrastructure revenues have been stagnant. Instead, the growth narrative hinges on its reported backlog—which has supposedly grown at a 40% CAGR over three years. Management explicitly attributes this to new contract wins outpacing revenues – a sign of a healthy demand pipeline7.
a. New Contract Wins Are Declining, Not Accelerating. New contract wins self-reported by Sterling’s subsidiaries to ENR contradict this narrative. They show that contract wins at Plateau and Petillo declined 10% in FY24 – the most recent year for which data is available, shattering Sterling’s claim that data center demand is accelerating.
b. We Cannot Reconcile 75% of Sterling’s Claimed Backlog Growth with New Contract Wins. Using the ENR disclosures, we calculate that cumulative new contract wins have eclipsed revenues by just $159m since year end 2021; barely 25% of Sterling’s $600m of E-Infrastructure backlog growth over the same period. With data indicating contract wins are below reported figures, we believe that stated backlog growth is not the indicator of healthy project wins purported and may even be overstated.
3. Inflated Margins; 45% of EBIT from Abnormal Accounting Revisions. Even if one were to accept Sterling’s data center narrative at face value, the Company’s reported margins seem at best unsustainable, and at worst, inflated.
a. Impossible 31% Segment Gross Margins. If we assume Sterling earns approximately 10–15% gross margins on its non-data-center work – consistent with industry norms – the implied margins on its purported data center projects would need to exceed 45% to reconcile with reported segment profitability. Industry peers we spoke with told us that even in the current environment, 20%-30% gross margins are typical for data center projects. One competitor openly expressed skepticism about Sterling’s margins, describing them as “probably not sustainable”.
b. Subjective Accounting Revisions on Future Profitability Estimates Drive ~45% of Reported EBIT. We unmask Sterling’s abnormal margins. Buried deep in its financials, footnotes reveal that ~7% of Sterling’s revenues are from “changes to contract estimates”. These are essentially an accounting assumption about the expected profitability of future work and are flagged by Sterling’s auditor as a key area of uncertainty. Yet their impact on Sterling’s margins is staggering. By our calculation, these revisions drove 45-48% of Sterling’s EBIT in FY24 and 9M25. Not only are the revisions way out of sync with peers, they also represent a substantial and uncharacteristic uptick from previous periods. Best case, these revisions reflect abnormal contract profitability that is not sustainable. Worst case, they reflect aggressive recognition of revenues that could reverse.
c. Revolving Door in the CFO Suite. Particularly troubling, given the extraordinary degree of discretion embedded in Sterling’s accounting, is the instability in the CFO role. Since May 2024, the Company has seen two CFO departures in quick succession — the most recent lasting less than a year. The Company was forced to bring back its retired predecessor on an interim basis before appointing a permanent replacement in July 2025 — Sterling’s third CFO in under 18 months.
4. Illusory Cash Flow. Investors have likely taken comfort in Sterling’s cash flow, viewing it as validation of the margin story. This is a mistake. Since 2022, Sterling’s cash flow has benefited from a cumulative one-time benefit of $550 million from advance billings – which seem abnormally high vs peers. Strip these out and Sterling’s cash conversion collapses to just 40% in 9M25.
a. Adjusted Cash Conversion Shatters Sterling’s Comparison to Specialty Peers. This is also critical because sell-side analysts use Sterling’s higher cash conversion as crucial justification for valuing it in line with specialty peers – which trade at much higher multiples.
5. Eye-Watering Valuation. In the past year, Sterling has undergone a dramatic multiple re-rating. The stock now trades at ~23x NTM EV/EBITDA, well above traditional contractor peers such as Granite Construction (NYSE:GVA) and approaching valuations typically reserved for specialty service providers like Quanta Services (NYSE:PWR), which enjoy durable competitive advantages, proprietary technical expertise, and meaningful barriers to entry across its end markets.
Sterling possesses none of these attributes. Even if investors believe Sterling’s margins are sustainable, a re-rating in line with civil contractor peers implies 60% downside to its stock price. We see even further downside should Sterling’s margins normalize in line with industry peers.
a. Even Against Other AI Darlings, Sterling’s Stock Looks Expensive. If we apply an industry average multiple to Sterling’s non-E-Infrastructure earnings, Sterling’s E-Infrastructure business is trading at nearly ~29x NTM EV/EBITDA. This is a substantial premium to AI darling NVIDIA (NASDAQ:NVDA) and on par with Vertiv (NYSE:VRT) – a pure play data center infrastructure business with vastly superior EBITDA growth prospects.
Ultimately, we believe Sterling is a regional contractor dressed up in an AI veneer, trading at a valuation that assumes it is something it is not. The stock is acutely vulnerable as the narrative fades and fundamentals reassert themselves.
Read the full report here by Snowcap Research
