Story

The Full Story

Five years ago, RadNet described itself in one line: the largest operator of freestanding outpatient imaging centers in the United States. Today it describes itself as two companies — a 418-center imaging network and a global AI‑powered health informatics platform ("DeepHealth") with 2,075 customers, 22 FDA clearances and $75M of Annual Recurring Revenue. The core imaging engine has delivered in every meaningful quarter. The Digital Health story is louder, more ambitious, and — for now — less proven.

Management raised guidance in seven of the last eight quarters. In the same period, Digital Health operating losses more than doubled and net income attributable to common stockholders turned negative. Both things are true at once.

1. The Narrative Arc

The growth is real — revenue compounded ~12% annually since 2021. But the segment breakout shows what's actually new. Imaging centers carried 99.9% of revenue in 2021, 97.5% in 2024, and 95.5% in 2025 (gross; ~50% of Digital Health is intersegment). The narrative shift is sharper than the dollars. FY2021's 10-K opened with "freestanding, fixed-site outpatient diagnostic imaging." FY2025's opened with the same line — but the Business section now devotes more space to DeepHealth OS, 22 FDA clearances, and "cloud-first, AI-powered" than to multi-modality density.

**The inflection is 2024, not 2025.** Q4 2024 was when management launched DeepHealth OS + SmartMammo + TechLive as products, renamed the AI segment to "Digital Health", and completed a $230M stock raise + debt refi. The 2025 acquisitions (iCAD, See-Mode, CIMAR, and the March 2026 Gleamer deal at €230M) are the check-writing phase of a strategy locked in twelve months earlier.

2. What Management Emphasized — and Then Stopped Emphasizing

What got promoted: AI moved from "a side division" in 2021 to the name of a reportable segment in 2023 to a standalone rebrand in 2024 to the lead item on every 2025 press release. "DeepHealth OS" did not exist as a product name until Q4 2024; by Q4 2025 it appears in almost every paragraph of shareholder communication. ARR — a SaaS metric RadNet had never reported — was introduced in FY2025.

What got retired: The capitation story (pre-determined per-member fees) was once a differentiator — "highly skilled at assessing and moderating the risks associated with the capitation agreements." Capitated revenue has shrunk from 11.3% of total in 2021 to 6.2% in 2025, and management rarely mentions it now. COVID recovery language, front-and-center in 2021–2022, is completely absent from 2025 communications. The "Imaging On Call" teleradiology subsidiary was quietly divested in 2020 for $1,000. The Baltimore Ravens / Orioles sports sponsorships are being "renewed through 2026" with less fanfare than their previous framing as a brand pillar.

3. Risk Evolution

What became more important. California's 2024 mandated minimum wage increase for healthcare workers is now a dedicated paragraph in the 10-K and, per FY2025 commentary, actively pressured Q1 2025 salaries. PE-backed imaging-chain competition entered the risk lexicon in 2023 — "private equity-backed chains" was added to the competitor list that used to say only "independent imaging operators." Labor competition for radiologists and technologists is the single risk whose tone has hardened most, driving the TechLive remote-scanning investment as a defensive response. Wildfire/natural-disaster risk was generic boilerplate until Q1 2025, when Southern California fires and a nor'easter cost ~$22M of revenue and ~$15M of Adjusted EBITDA in a single quarter — management now describes California location concentration explicitly.

What became less important. COVID disruption disclosure shrank from dominating the risk section in 2021 to generic pandemic-preparedness boilerplate by 2024. Interest-rate risk faded after the April 2024 refinancing (maturity extended to 2031) and November 2024 repricing — by 2025 the leverage ratio is ~1.0x and Berger stops reciting it on every call.

What is newly visible. AI integration and IPR&D write-off risk is not a stated risk factor but a revealed one: the Aidence IPR&D (Veye lung-nodule platform) failed to achieve US FDA approval without a new submission and triggered a $3.9M impairment in September 2023. Management disclosed it matter-of-factly in the mda — an early signal that the European AI acquisitions were not the seamless plug-ins they had been positioned as in 2022.

4. How They Handled Bad News

RadNet's pattern on bad news is pre-announce, itemize, move on.

**Q1 2025 wildfires and storms — best-in-class disclosure.** Management pre-announced the ~$22M revenue / ~$15M Adjusted EBITDA hit when releasing Q4 2024 results in late February 2025. Q1 2025 reported exactly those numbers. Net loss was $37.9M (vs $2.8M the prior year). Full-year guidance was still raised — because Q2, Q3 and Q4 had to overdeliver to get there, and did.

Where the record is less clean is the Aidence IPR&D impairment. The $3.9M charge is noted inside the FY2023 MD&A, not called out on the earnings call or press release, and not reconciled against the January 2022 purchase price allocation commentary that had framed Aidence's lung-nodule IPR&D as a near-term FDA filing. Investors who didn't read Note 4 would never know the AI acquisition thesis had slipped. It is the one case in the five-year record where bad news was genuinely buried.

The 2025 shift to net loss ($17.6M attributable loss for the full year, a swing from $3.7M of income in 2024) is being framed as "integration of iCAD, See-Mode and CIMAR progresses" — an investment narrative, not a miss. That is a defensible but convenient framing, and the reader should flag it: Digital Health segment operating losses expanded from $14.9M in 2024 to $32.3M in 2025 as acquisitions stacked up.

5. Guidance Track Record

No Results

Management credibility score (out of 10)

Explanation. On the operating side — Imaging Center revenue, Adjusted EBITDA, procedural volumes, balance-sheet deleveraging — Berger's team has delivered or exceeded every material guidance range they set across 2024 and 2025, including the one quarter they pre-announced would be ugly. That is rare. Where credibility is still being earned is product-level AI milestones: the single datapoint so far (Aidence Veye Lung Nodule IPR&D) failed, and the big promises — DeepHealth OS as an industry platform, ARR scaling from $75M to $140M in twelve months, iCAD and CIMAR integrations unlocking international breast-cancer screening — are all ahead, not behind. A 7 is charitable on the AI track but fair on the whole record.

6. What the Story Is Now

RadNet is no longer a "diagnostic imaging center operator that is dabbling in AI." It is a bet that the same management team that ran a $2B outpatient imaging business cleanly can use its own scale (24 million procedures, 418 centers) as a testbed and reference customer for an AI platform it is rapidly buying into existence — iCAD, See-Mode, CIMAR, Aidence, Quantib, Kheiron, and (post-year-end) Gleamer, all stacked under DeepHealth OS.

What has been de-risked since 2021:

Operational resilience: procedural volumes recovered from COVID, grew ~14% YoY in Q4 2025 with PET/CT up 28%. The core works.
Balance sheet: net debt/Adjusted EBITDA fell from ~2.0x (end 2023) to ~1.0x (end 2025); $833M of cash at mid-2025; maturities pushed to 2031.
Guidance execution: near-flawless on the imaging side over 2024–2025.
AI adoption inside RadNet itself: Enhanced Breast Cancer Detection at ~45% adoption, generating $40-per-patient out-of-pocket revenue; TechLive on 255+ MRIs.

What still looks stretched:

Digital Health segment losses doubled year-over-year even as revenue grew 41%. Near-term losses are guided to continue as iCAD, See-Mode, CIMAR and Gleamer integrate.
The ARR bridge to $140M by end-2026 implies ~85% growth, much of it from just-acquired customers management has not yet retained through a renewal cycle.
"Most comprehensive and broad collection of clinical AI solutions of any company worldwide" is a big claim for an imaging-center operator with 12 months of platform revenue history. Roll-up integration risk is real — and the Aidence Veye precedent shows it is not theoretical.
Shareholder dilution: weighted diluted shares went from ~57M (Q1 2023) to ~76M (Q4 2025), largely financing the AI strategy (iCAD was an all-stock deal; the March 2024 raise was $230M).

**What to believe vs discount.** Believe the imaging-center numbers and the near-term EBITDA guidance — that part of the machine is proven. Discount the ARR trajectory and the "global AI platform leader" framing until (a) organic Digital Health net-new customer adds accelerate independently of acquisitions, and (b) two or three of the acquired product lines demonstrate commercial durability without a RadNet-internal reference customer carrying them.

The story in 2021 was simple: buy more imaging centers, integrate, run them well. The story in 2026 is not simple — and the people telling it have earned the right to be heard on the imaging half, and not yet on the AI half.