Business

Know the Business

RadNet is the largest freestanding outpatient diagnostic imaging operator in the U.S. — an asset-heavy, regional-density roll-up where operating margins sit near 4% and capex runs a chunky 10% of revenue. Almost all the current equity value above the cash-on-cash yield of the imaging business is a market bet on the Digital Health (DeepHealth) segment — an AI/SaaS unit that grew 41% in FY2025 but is still losing roughly $32M a year. The playing field is consolidating into RadNet's favor, but the market is underwriting a software re-rate that has not yet shown up in segment margins.

1. How This Business Actually Works

The economics are a fixed-cost scanner utilization game layered on a payor-negotiation game. Each MRI, CT, or PET machine costs roughly $0.5M–$2M upfront and runs for ten years — almost all center-level profit comes from running those machines more hours per day on a patient mix skewed toward advanced modalities. The bottleneck is not demand; it is finding enough technologists and radiologists (a structurally tight U.S. labor pool), scheduling patients efficiently across a regional network, and negotiating payor rates that leave something for the operator after 53% of revenue goes to salaries and professional reading fees.

Two structural realities sit inside those two charts:

First, commercial insurance (55% of revenue) is the only payor class where negotiated rate can actually move up. Medicare (23%) is fee-schedule price-taking; Medicaid (3%) is worse; capitation (6%) is pre-paid flat-fee per member. Commercial rate pickups — together with a product-mix shift toward PET/CT (prostate-cancer and Alzheimer's tracers) — are the only levers management pulls to beat baseline scan-volume growth.

Second, labor is 53% of revenue and growing faster than revenue. The California healthcare minimum-wage increase landed in October 2024; FY2025 salaries plus professional fees rose 10.1% against total-revenue growth of 11.5%. Any slowdown in volume means wage deleveraging, which is exactly what compressed FY2025 EBITDA margins to 15.4% from 17.7%.

The unit economic flywheel

The analogy that best fits RadNet is a regional grocery chain, not a tech platform. Same-store volume and density are everything; the operator that controls the cluster collects a modest but recurring margin the smaller competitor cannot match.

Digital Health: a separate engine, still loss-funded

DeepHealth is the lit AI/SaaS business inside the plumbing. FY2025 revenue grew 41% to $92.7M on a customer base that jumped from 486 to 2,075 after the iCAD acquisition closed in July. Annual Recurring Revenue ended the year at $75M. But 45% of segment revenue is still internal (sold to RadNet's own centers), the operating loss nearly doubled, and management expects "net loss in the near term" as three acquisitions (iCAD, See-Mode, CIMAR UK) are integrated. This is a build-phase SaaS business funded by the imaging operating cash flow — not yet a profit engine.

2. The Playing Field

Pure-play public peers for outpatient imaging are unusually thin. Akumin went private in 2023. Alliance HealthCare Services was acquired by Akumin. Alliance/RadNet-style roll-ups are otherwise private or embedded inside hospital systems (HCA, Tenet). What is left in the public peer set is asset-heavy healthcare services operators with similar payor exposure and multi-site roll-up economics — dialysis (DVA), skilled nursing (ENSG), post-acute rehab (SEM), hospitals (THC), and outpatient PT (USPH). Apples-to-apples this is not, but it is the best available lens.

Three relative facts jump off that table:

RDNT has the lowest ROIC and the highest capex intensity in the peer set, yet trades at the second-highest EV/EBITDA. At 1.7% ROIC and 10.4% capex/revenue, the imaging-center business alone looks like a capital sinkhole priced like a compounder. That gap only closes if a reader believes either (a) the 15.4% EBITDA margin rebounds toward the 19%+ it printed in FY2019 and FY2021, or (b) DeepHealth embeds a software-grade margin and re-rates the stock.

Hospitals (THC) and dialysis (DVA) — the lowest-multiple names — actually deliver better operating metrics on almost every dimension. THC runs 21% EBITDA margins, converts 12% of revenue to FCF, and trades at 7.2× EV/EBITDA. The market is plainly not paying RadNet for the imaging business as a hospital-equivalent cash engine; it is paying for a growth story that the numbers underneath the imaging segment do not yet validate.

Against ENSG — the other high-multiple operator here — the comparison is informative. Ensign trades at a similar 20.7× because it has consistently delivered 8%+ ROIC and 7% FCF margin from a better-run SNF roll-up. RadNet wants that re-rating; it does not yet have the return profile to earn it on imaging alone.

The chart makes the anomaly explicit: every other peer with an above-15× multiple also posts above-8% ROIC. RDNT sits alone in the top-left — a high multiple on a low-return core business.

3. Is This Business Cyclical?

Not in the traditional consumer sense. Imaging volume is driven by demographics (aging population, cancer screening, Alzheimer's diagnostics), not the business cycle. FY2020 revenue fell 8% during peak COVID, then snapped back 23% in FY2021 — that one pothole is the only recession-shaped event in the last 15 years of the revenue series.

Where the cycles do hit is more subtle — and matters more for profitability than for revenue:

1. Reimbursement cycles. CMS Physician Fee Schedule updates reset Medicare rates annually. RadNet notes 23.4% of imaging revenue came from Medicare in FY2025 and another 2.5% from Medicaid; a negative CMS update hits margin with no offset. The 2022–2024 multi-year reimbursement pressure, compounded by wage inflation, is one reason EBITDA margins have drifted from the 21.2% FY2021 peak down to 15.4% in FY2025.

2. Wage cycles. California's healthcare minimum wage increase (October 2024) is the single biggest reason FY2025 margins compressed. Because California is RadNet's largest state, what looks like a national wage tailwind hits RadNet disproportionately, and cannot be repriced to commercial payors in the same year.

3. Credit / capex cycles. Long-term debt sits at $1.06B against $313.8M of EBITDA (about 3.4× gross, ~1.0× net of the $767M cash balance after the 2024 debt restructuring and equity raise). Interest expense is $70M annually. A higher-rate refinancing cycle — or a slower equity market that closes the stock-issuance window that funded 2024's deleveraging — would compress equity holders' upside fast.

4. Seasonal Q1 softness. A mechanical — not macro — cycle: high-deductible health plans reset in January, northeastern winters cancel appointments, and Q1 is the reliable low point of every year. This is not a thesis-moving item but it matters for reading quarterly prints.

4. The Metrics That Actually Matter

Five numbers explain more about RadNet than any P&L or balance-sheet ratio. Watch these first; everything else is downstream.

RadNet's own scorecard on these metrics

Capital-intensity reality check

Over seven years RadNet generated roughly $376M of cumulative FCF against $1.0B of cumulative capex — a reminder that the operating cash flow this business produces mostly goes back into the ground. That is the price of the density flywheel, and it is also the reason the stock is better judged on long-run compounding than on any single year's "cash earnings."

5. What I'd Tell a Young Analyst

Imaging is a plumbing business; stop modeling it like SaaS. Fixed-cost scanners, unionized-adjacent labor, regulated rates, and 10-year equipment cycles — none of that is going to become a 30% operating-margin story by force of will. The core case for owning RDNT is that the density flywheel steadily widens the gap between RadNet and sub-scale independent centers, that California / Mid-Atlantic / NY payor contracts modestly expand, and that cost inflation eventually gets recovered at renewal. That is a 10–14% EBITDA-CAGR outcome — solid, not thrilling.

Your entire valuation edge hangs on DeepHealth. Strip Digital Health out and you are paying a 20-plus multiple for a 15% EBITDA-margin, 1.7% ROIC asset-heavy roll-up. That math does not work. The lit thesis is that DeepHealth OS, FDA-cleared AI for breast/lung/prostate screening, and the iCAD/See-Mode/CIMAR stack eventually become a 70-75% gross margin subscription business that independently deserves a software multiple. Track external ARR growth, customer count, and gross margin inside the segment every quarter — these are the numbers that either confirm or kill the re-rating case.

Wage inflation and the California minimum wage are the silent thesis-killer. If you believe FY25 margins are a permanent reset rather than a one-time October-2024 rebase, the whole compounding story is two turns of multiple too expensive. Watch Q1 and Q2 FY26 for evidence that pricing finally recovers the wage step-up.

Three things that would genuinely change the thesis. (1) CMS proposing a structural cut to outpatient imaging reimbursement — immediate -200 bps margin risk with limited offset. (2) DeepHealth hitting a gross-margin inflection and demonstrating external growth equal to or greater than internal — that is when the multiple is earned. (3) A $150M+ acquisition funded with debt at 8%+ rates — if equity markets cool and management falls back on leverage, the risk profile of the whole stack shifts.

What the market is probably underestimating. The FDA-cleared AI moat on mammography workflows (via iCAD and Kheiron) is more commercially defensible than generic imaging AI. 1,500 iCAD customers is a real installed base, and competing FDA clearances take 2–3 years and clinical trials to build.

What the market is probably overestimating. How quickly DeepHealth can swing to segment-level profitability. Three acquisitions in 2025, integration costs running through FY26, and management's own "near-term losses" guidance mean the SaaS re-rate is at least a 2027 event, not a 2026 event. The patient investor has a chance to let earnings catch up to the multiple; the impatient one should wait for a cheaper entry.