RadNet
The biggest chain of outpatient scan centres in America: 442 places to get an MRI, CT or mammogram for less than a hospital charges.
The scores
Growth3/4 →
Same-centre revenue grew 11.1% and revenue 25% in Q2 2026, but much of the total was bought, with acquisitions every year.
Business quality3/4 →
Regional scale and a lower price than hospitals take share, but hospitals, private chains and dozens of AI rivals compete directly.
Financial position3/4 →
Net debt about $616M, 1.8 times EBITDA, term loans due 2031 and an undrawn revolver; acquisitions use more than free cash.
Execution record3/4 →
Beat and raised imaging guidance through 2025 and twice in 2026, but on adjusted figures while GAAP earnings fell to a loss.
Per share capture2/4 →
Shares up 47% since 2021 through offerings and an all-stock deal; revenue per share grew 17% while revenue grew 73%.
Potential3/4 ↑
A bounded upside on observable steps: scans shifting from hospitals, joint ventures signed, and a CMS rule pending that would help.
Risk profile3/4 →
One serious dependency, payer rates set by CMS and insurers, plus an 81-year-old founder and over 100 centres in California.
Valuation2/4 →
About 18.5 times trailing adjusted EBITDA and 55 times free cash flow: pays for growth and rising margins for years.
Evidence quality3/4 →
Results, payer mix and debt are filed and the price gap measured by CMS; the share shift and AI benefits are inferred or management's word.
If it works 0.8× · if it fails you keep about 62%.
The Verdict
We rate RadNet Unattractive. The business is solid: insurers pay its centres far less than a hospital for the same scan, so patients keep moving to it and its centres grow about 10% a year without counting new ones. The price is full: about 18 times its yearly earnings before interest, tax and depreciation, which counts on years of growth and margins finally rising, though they have not risen in years, an improvement paid for up front. As a bet it is priced for success: if its AI software grows as hoped the shares are worth less than today, and if it fails shareholders keep about three-fifths.
The long caseRadiologists are scarce and scans keep growing, so software that reads images and drafts reports could sell widely, starting in RadNet's own centres.
The evidence under this - 8 claims
No body publishes the outpatient share of scans by year, so the shift is inferred; part of the growth is new demand, such as prostate and amyloid PET scans, now over 25% of RadNet's PET volume.
Adjusted EBITDA leaves out $59.3M of trailing stock pay, and partners' minority stakes, $298.0M at book, would add to the value; both make the multiple higher, not lower.
Imaging software, the part of the segment not bought recently, still grew 17.3% in Q2 2026 (C11).
The company
When a doctor orders an MRI, a CT scan or a mammogram, the patient can usually go to a hospital or to a standalone centre. RadNet runs 442 of those centres, clustered in California, New York, New Jersey, Maryland and a few other states, and says it is the largest such chain in the country.
MRI and CT are most of the money: MRI alone brought 37.7% of its imaging payments in 2025. Its contracted radiologists read the images, and the report goes back to the doctor.
The evidence under this - 4 claims
Its same-centre margin barely moved in Q2 2026 (C04) and capitation revenue has fallen two years running (C02), neither what a widening scale advantage would show.
How it makes money
RadNet is paid per scan, mostly by private insurers ($1.13B of 2025's $2.04B revenue) and Medicare ($477M). Some health plans instead pay a fixed monthly fee per member, whatever the number of scans.
A complex scan pays much more than an X-ray, so growth in MRI, CT and PET is what lifts revenue. A small software arm also sells imaging and AI programs to hospitals.
The evidence under this - 3 claims
Where the money comes from, stream by stream
| Stream | Share of sales (FY 2025) | How it earns it |
|---|---|---|
| Commercial insurance | 55.4% | Scans paid by private insurers under negotiated contracts. |
| Medicare and Medicaid | 25.9% | Scans paid at government fee-schedule rates. |
| Capitation | 6.2% | A fixed monthly fee per member from health plans and medical groups, whatever the number of scans. |
| Other payers and fees | 12.5% | Workers' compensation, personal injury, other payers, management fees and other revenue, including Digital Health's outside sales. |
Who runs it
Howard Berger, 81, has run RadNet since 1992 and built it by buying chains and partnering with hospitals. He owns under 1% of the shares, and all directors and officers together about 5%. Doctors' groups owned by senior managers staff centres in six states.
Pay rewards staying more than results: 2025 bonuses were set at the board's discretion and long-term awards are shares that vest with time. Insiders have sold, not bought, this year.
The evidence under this - 3 claims
Why it wins, and who could beat it
Why customers come back
The edge is price and density. Medicare's own figures show a hospital department paid about $106 for a bone-density scan an office does for about $30, so insurers prefer the cheaper centre, and RadNet has the most centres where it operates.
The evidence it works: scans at its existing centres grew 9.6% in Q2 2026, while CT use across the country is growing only slowly. That gap means it is taking work from hospitals.
The evidence under this - 5 claims
No body publishes the outpatient share of scans by year, so the shift is inferred; part of the growth is new demand, such as prostate and amyloid PET scans, now over 25% of RadNet's PET volume.
Its same-centre margin barely moved in Q2 2026 (C04) and capitation revenue has fallen two years running (C02), neither what a widening scale advantage would show.
The threat
Hospitals still do most scans and can open centres of their own; HCA lists imaging centres among its rivals. Private chains such as SimonMed, with more than 170 locations, chase the same patients. In software, the FDA has cleared over 1,200 radiology AI products, so RadNet's AI arm sells into a crowded field.
The evidence under this - 4 claims
The rivals: HCA, GEHC, and 3 more that are private, in-house or listed abroad.
Who else does this, rival by rival
| Ticker | Rival | Where it competes | How far along | Priced against it |
|---|---|---|---|---|
| Private | SimonMedPriced against it: Private | Imaging Centers | On sale | Private |
| HCA | Hospital outpatient departments, such as HCA'sPriced against it: One segment only | Imaging Centers | On sale | One segment only |
| Private | Other private-equity-backed imaging chainsPriced against it: Private | Imaging Centers | On sale | Private |
| GEHC | GE HealthCare and LunitPriced against it: One segment only | Digital Health | On sale | One segment only |
| Private | Aidoc and TherapixelPriced against it: Private | Digital Health | On sale | Private |
How big it could get
How big this could get
Americans had about 93 million CT scans in 2023, and RadNet did about 1.2 million of them, near 1.3%. CT use is growing slowly, so the room is in taking scans from hospitals rather than in the market growing.
Nobody this search could read counts MRI or PET scans by where they are done, so the full market cannot be sized from outside.
The evidence under this - 3 claims
Who has signed
Most revenue is scan by scan, with nothing signed in advance. The firmer parts: 157 centres run as joint ventures with hospital systems, health plans paying fixed fees worth $125.5M in 2025, and about $37M of software contracts signed in the first half of 2026.
The evidence under this - 3 claims
The results
The numbers
Sales grew 19% over the twelve months to Jun 2026, faster than the 12% in the year to Dec 2025.
Operating margin is 3.1%, down from 6.1% in the year to Dec 2023.
Yearly sales growth has ranged from +8.7% to +23% over the last 5 years.
Each dollar put into the business earns 1.7% a year after tax, down from 6.9% in the year to Dec 2021.
It turned 3.2% of sales into free cash, $72.4M over the last twelve months.
It owes $606M more than the cash it holds, about 2.0 years of the cash its operations bring in.
The share count rose 47% in 4.5 years, and pay in shares costs 2.6% of sales.
All the numbers
Filed figures from the company's 10-K and 10-Q, read 2026-10-10. Fiscal years end in December.
| Dec 2021 | Dec 2022 | Dec 2023 | Dec 2024 | Dec 2025 | 12 mo to Jun 2026 | |
|---|---|---|---|---|---|---|
| Size and growth | ||||||
| Sales | $1.32B | $1.43B | $1.62B | $1.83B | $2.04B | $2.27B |
| Sales growth | +23% | +8.7% | +13% | +13% | +12% | +19% |
| Profit | ||||||
| Operating margin | 6.3% | 3.2% | 6.1% | 5.7% | 3.0% | 3.1% |
| Net margin | 1.9% | 0.7% | 0.2% | 0.2% | −0.9% | −0.9% |
| What growth costs, per sales dollar | ||||||
| Pay in shares | 1.9% | 1.7% | 1.7% | 1.6% | 2.7% | 2.6% |
| Capital spending | 10% | 8.4% | 11% | 10% | 10% | 10% |
| Cash | ||||||
| Cash from operations | $149M | $146M | $221M | $233M | $299M | $310M |
| Free cash flow | $11.6M | $27.0M | $44.3M | $45.0M | $85.6M | $72.4M |
| Balance sheet | ||||||
| Cash and short-term investments | $135M | $128M | $343M | $740M | $767M | $726M |
| Debt | $755M | $852M | $830M | $1.02B | $1.09B | $1.33B |
| Net cash | −$620M | −$724M | −$487M | −$276M | −$323M | −$606M |
| Returns and shares | ||||||
| Return on capital | 6.9% | 3.3% | 5.8% | 5.4% | 1.6% | 1.7% |
| Shares (diluted) | 53.4M | 57.3M | 64.7M | 74.8M | 75.2M | 78.7M |
What growth costs
Growth has not made RadNet more profitable. Revenue rose by more than half from 2021 to 2025 while operating profit fell, pay in shares doubled, and capital spending on centres and scanners now takes most of the cash the business brings in.
The evidence under this - 2 claims
The last report (9 Aug)
Quarter ended June 2026, filed figures.
The June 2026 quarter looked strong. Revenue rose 25% to $622.7M and earnings before interest, tax and depreciation, the company's main measure, 22.7% to $99.7M. Management raised its yearly forecast for the second time.
The weak spot was the margin. The share of each revenue dollar kept as those earnings rose by less than a fifth of a point, as salaries rose 6.8% and rolling out its own AI cost money. The software arm grew 56.5% and still lost money.
What it did to the thesisIt confirmed the growth half of the argument: patients keep coming and the forecast rose again. It did nothing for the half the price needs, because margins barely moved, so the question of whether growth ever makes the business more profitable stays open.
The quarter in detail
Sales growth from the previous quarter, filed figures.
Earnings per share $0.29 against $0.20 expected; $0.10 as filed, the gap being what the company adds back.
The evidence under this - 5 claims
Q1 was cut by about $9M of adjusted EBITDA by weather (C33), so the full year needs a second half of $186M to $199M of imaging adjusted EBITDA, the strongest half the company has reported.
What management did with the money
Management spends the cash on new centres, scanners and acquisitions. In the first half of 2026 it paid $315.7M for businesses, including a French AI firm, and borrowed $248.9M more. Earlier deals were paid with new shares, which is why each share owns less of the company than in 2021.
Insiders in 2026Q2: 5 purchases and 5 sales, from their Form 4 filings.
The evidence under this - 3 claims
The price, and where we disagree
In short. Full. The price pays for growth near today's pace lasting years and for margins rising, which they have not done, while little cash reaches shareholders.
The shares have swung hard: from near $80 last November to $50.82 in June and back to about $71 after two raised forecasts.
Third-quarter results on 9 November show whether the second half earns what the forecast needs, and Medicare's final 2027 rules follow. On 26 October RadNet also agreed to buy teleradiology firm ONRAD for $350M in cash and 5M shares; the profit it claims for it is unchecked.
The evidence under this - 6 claims
Q1 was cut by about $9M of adjusted EBITDA by weather (C33), so the full year needs a second half of $186M to $199M of imaging adjusted EBITDA, the strongest half the company has reported.
Imaging adjusted EBITDA margin is about 16%, so a business RadNet describes as earning about a third of its revenue as adjusted EBITDA is a different kind of profit pool, and the figure is unaudited and unchecked by any outside party.
How we read the multiple
A buyer pays for years of growth and rising margins.
1. The shares are worth about $5.6B; with about $616M of net debt the business is valued at about $6.2B. 2. It earned $335.6M before interest, tax and depreciation over the past year, on its own measure, which leaves out pay in shares. 3. So the business costs about 18 times those earnings, and about 25 times the reported figure above. 4. Free cash flow this year is forecast at about $102M, under 2% of the shares' value.
Against its own history. Buyers pay well above the company's usual price for each dollar of earnings, for growth that has not yet made the business more profitable.
The evidence under this - 3 claims
Adjusted EBITDA leaves out $59.3M of trailing stock pay, and partners' minority stakes, $298.0M at book, would add to the value; both make the multiple higher, not lower.
Priced against its rivals
None of the listed rivals above sells the same thing to the same buyers as a large part of what it does. A price comparison needs at least two, so the price is judged against its own history instead.
The bet: what it pays if it works
The bet is on Digital Health, the software and AI that read scans. It loses money and most of its recent growth was bought, so the imaging centres are counted separately, at the price this thesis would call fair.
If the software keeps growing about a fifth a year, the shares end up worth less than today, because the price already pays far more for it than that is worth. If it never earns, shareholders keep the centres, about three-fifths of today's value.
| Digital Health revenue in two to three years | Customers at today's recurring revenue each needed | At the stated multiple |
|---|---|---|
| $140.00M | about 3.8K at 100% kept | 0.8× |
| $200.00M (the big win) | about 5.4K at 100% kept | 0.8× |
| $280.00M | about 7.6K at 100% kept | 0.8× |
To pay 2× (a Fair bet) it would need about $2.47B of Digital Health revenue (about 66.8K customers at today's recurring revenue each); to pay 5× (a Cheap bet), about $8.17B of Digital Health revenue (about 220.9K customers at today's recurring revenue each), at the same multiple.
Each row is Digital Health alone, plus the rest of the company (Imaging Centers) counted at about $4.92B.
The steps it has to take
- Done Breast-screening AI in use across RadNet's centres
- Done iCAD and Gleamer bought and folded into Digital Health
- In progress Digital Health recurring revenue above $140M due 2026-12
- Not started AI drafting half of RadNet's reports due 2027-06
- Not started Digital Health profitable after research spending due 2027-12
- Not started Digital Health margin above 20% due 2028-12
The detail behind the label
The evidence under this - 6 claims
Imaging margins rose only 0.17 points in Q2 2026 despite a 2.4-point shift toward advanced scans; management blamed salary inflation and the cost of rolling out its own AI (C36).
Imaging software, the part of the segment not bought recently, still grew 17.3% in Q2 2026 (C11).
Q1 was cut by about $9M of adjusted EBITDA by weather (C33), so the full year needs a second half of $186M to $199M of imaging adjusted EBITDA, the strongest half the company has reported.
Adjusted EBITDA leaves out $59.3M of trailing stock pay, and partners' minority stakes, $298.0M at book, would add to the value; both make the multiple higher, not lower.
Where we disagree with the market
A growing outpatient imaging leader with AI on top; nine of 11 analysts rate it a buy.
The patients are real, but margins have not moved in years and growth reaches shareholders diluted.
Analysts expect earnings per share to nearly double in 2027. The filings show a loss over the past year after pay in shares and deal costs, margins flat near 16%, and 47% more shares since 2021.
Against us: growth is real and broad, the forecast has been raised twice this year, and a Medicare rule cutting hospital pay would send more patients its way.
What analysts say, and the evidence
The evidence under this - 6 claims
Adjusted EBITDA leaves out $59.3M of trailing stock pay, and partners' minority stakes, $298.0M at book, would add to the value; both make the multiple higher, not lower.
Q1 was cut by about $9M of adjusted EBITDA by weather (C33), so the full year needs a second half of $186M to $199M of imaging adjusted EBITDA, the strongest half the company has reported.
What could go wrong, and how we would know
The risks
- AI could lift margins. If AI drafts half of reports by mid-2027, as management plans, reading costs could fall and margins rise, which would make us wrong.
- Medicare could pay less. Medicare is about a quarter of revenue; management says the proposed 2027 rule costs under $1M, but the final rule is not out.
- Hospitals could compete harder. If hospitals cut prices or keep their higher rates, the flow of patients to cheaper centres could slow.
- A founder of 81. Berger has run the company since 1992 and his doctors' groups staff many centres; no successor is named.
- Salaries keep rising. Technologists and radiologists are scarce, and pay rose 6.8% at existing centres in Q2 2026.
The evidence under this - 6 claims
Imaging margins rose only 0.17 points in Q2 2026 despite a 2.4-point shift toward advanced scans; management blamed salary inflation and the cost of rolling out its own AI (C36).
The rule is a proposal; the final rule is due in November 2026, and RadNet's 10-K says rate cuts reduce revenue unless volumes or mix make up for them.
What we still do not know
- What share of US scans is done outside hospitals, and how fast is it rising?
- How much does RadNet earn per scan against a hospital, insurer by insurer?
- How much of the software arm's revenue would stay if RadNet stopped buying from it?
- Does ONRAD earn the profit RadNet says, and how is the cash price paid for?
The evidence under this - 4 claims
No body publishes the outpatient share of scans by year, so the shift is inferred; part of the growth is new demand, such as prostate and amyloid PET scans, now over 25% of RadNet's PET volume.
Its same-centre margin barely moved in Q2 2026 (C04) and capitation revenue has fallen two years running (C02), neither what a widening scale advantage would show.
Imaging adjusted EBITDA margin is about 16%, so a business RadNet describes as earning about a third of its revenue as adjusted EBITDA is a different kind of profit pool, and the figure is unaudited and unchecked by any outside party.
What would change our view
- Unattractive→NeutralA cheaper price on unchanged earnings
It needs:
- ○The business is valued at 14 times its yearly earnings before interest, tax and depreciation, from about 18.5 today.about ≈$53.66 today, 24% below the last close
- Unattractive→NeutralMargins finally rise with growth intact
It needs all of:
- ○The imaging centres keep 17.5 cents of each revenue dollar as those earnings for a full year.checked at the next results, due by 9 Nov
- ○Scans of MRI, CT and PET at existing centres keep growing at least 8% a year.checked at the next results, due by 9 Nov
We would also be wrong if
- Cash after acquisitions pays for growth without new shares for two years.
- Medicare makes hospitals take the office price for simple scans from 2027.
What is coming
- 9 November 2026 Q3 2026 results: Whether the second half earns the $93M to $99M a quarter of imaging adjusted EBITDA the guidance needs, and whether the margin moves.
Every dated event ahead
- 9 November 2026 Q3 2026 results: Whether the second half earns the $93M to $99M a quarter of imaging adjusted EBITDA the guidance needs, and whether the margin moves.
- 30 November 2026 CMS final 2027 Medicare rules: Whether site-neutral pay for hospital imaging is finalised and whether RadNet's Medicare rates hold.
- 23 February 2027 ONRAD acquisition closes: Whether the deal closes on the stated terms and, with the ONRAD call, what its profit and the funding of the $350M cash look like.
- 15 August 2027 Q2 2027 results: Whether AI drafts half of reports and the imaging margin rises, as management targets.
Since we wrote this
- 26 OctNewsRadNet agreed to buy teleradiology firm ONRAD$350M cash, 5M shares and an earn-out for a business RadNet says earns $60M to $65M a year; unchecked, so the view held.
- 11 OctThesis reviewDigital Health declared an early business, with a bet on itThe cheap-diagnosis Futures theory names RadNet for its AI reading tools, a segment that loses money and whose growth was mostly bought.
- 10 OctThesis writtenUnattractive, medium convictionWritten a month before Q3 results, after the shares recovered from a third lower in June on a second guidance raise.
Sources 23
Sources - 23
Every claim above cites one of these, or states the reasoning that derived it. The quote is frozen at the date it was read: a link can rot, and an agent fetch returns a reading of a page rather than the page. Backs counts the whole claim ledger, so it can name a claim this page does not surface.
- S01primarycompany filing2026-03-02centres and regions, payer revenue, competition classes, the consolidated medical group, risks, Digital Health clearances and customers, the equity offerings and iCADbacksC01C02C07C20C21C23C29C32C38C40
- S02primarycompany filing2026-08-10same-centre revenue and costs, acquisitions including Gleamer and the Idaho joint venture, debt, the segment note and AI deploymentbacksC05C08C09C11C22C26C31C35C36
- S03primarycompany release2026-08-09Q2 2026 results, volumes, Digital Health, payer mix, cash flow, net debt and revised 2026 guidancebacksC01C04C05C08C09C10C11C15C22C26C38C39
- S04primarycompany release2026-03-02full-year adjusted EBITDA, Q4 2025 results, payments and procedures by scan type, and the original 2026 guidancebacksC03C26
- S05primarytranscript2026-08-10RadNet Q2 2026 earnings call transcript (FMP), held 10 August 2026Digital Health's split and targets, the Medicare rule's effect, the hospital rule, the Florida acquisition, margin questions and net debt
Our analysis shows a negative impact of less than $1 million to 2027 revenue.
backsC11C12C15C31C32C36C38 - S06primarycompany release2026-06-10the $250M incremental term loan, the April 2031 maturity, the margin cut to SOFR plus 2.00% and the undrawn revolverbacksC09
- S07primarycompany filing2026-04-20the leaders, ownership, bonuses, share grants and the 2022 performance optionsbacksC29C30
- S08primarycompany filing2026-08-10five years and the trailing year of revenue, operating income, net income, cash flow, capital spending, stock pay and diluted sharesbacksC06C07C27C37
- S09primaryvendor data2026-10-09Research pack for RDNT (FMP market tier, FINRA short interest), fetched 10 October 2026the 9 October close, the 52-week range and returns, analyst ratings and estimates, the valuation series, short interest and the insider seriesbacksC25C26C27C28
- S10primaryregulator2026-07-07the proposal to pay off-campus hospital departments the office rate for imaging without contrast, and CMS's price comparisons
In 2025, this service had a physician office payment rate of around $30 and an OPPS payment rate of approximately $106, making the same scan more than three times as expensive in the OPD than in the physician office. ... Medicare paid 294 percent more in an OPD than in a freestanding office.
backsC13C22 - S11primaryindustry data2025-04-14Smith-Bindman et al., 'Projected Lifetime Cancer Risks From Current Computed Tomography Imaging', JAMA Internal Medicine, 2025 (PMC11997853) JAMA Internal Medicinethe national count of CT exams and patients in 2023 and the growth since 2007
An estimated 61 510 000 patients underwent 93 000 000 CT examinations in 2023 ... CT use is 30% higher today than in 2007
backsC16C22 - S12primaryregulator2026-10-10the count of AI-enabled devices by panel and year, and rival authorisations
The AI-Enabled Medical Device List is a resource intended to identify AI-enabled medical devices that are authorized for marketing in the United States.
backsC17 - S13primaryother2026-10-10SimonMed Imaging home page SimonMed Imaginga private rival's footprint and pitch
With 170+ locations across the US, it's easy to find a convenient location near you.
backsC18 - S14primarycompany filing2026-02-13a hospital group naming freestanding imaging centres among its competitionbacksC19
- S15primarycompany filing2026-02-05a scanner maker's AI software sold to cliniciansbacksC19
- S16primaryindustry data2024-03-21the projected physician shortage to 2036
the United States will face a physician shortage of up to 86,000 physicians by 2036.
backsC23 - S17primaryindustry data2025-02-04Hernström et al., MASAI randomised trial, The Lancet Digital Health, 2025 The Lancet Digital Healththe screen-reading workload saved by AI-supported mammography
resulting in a 44·2% reduction in the screen-reading workload
backsC24 - S18primaryregulator2026-07-16CMS, CY 2027 Physician Fee Schedule proposed rule (Federal Register, 16 July 2026, doc. 2026-14327) CMSthe proposed 2027 conversion factor
the 2027 nonqualifying APM conversion factor represents a projected decrease of $0.56 (-1.68 percent) from the current conversion factor of $33.5875.
backsC14 - S19primarycompany release2026-05-10Q1 2026 results, the weather effect and the first guidance raisebacksC26C33
- S20primarycompany filing2026-10-10each open-market sale, its price and its Rule 10b5-1 flagbacksC28
- S21primarycompany release2025-11-09Q3 2025 revenue, adjusted EBITDA and volumes, the baseline for Q3 2026backsC26C34C39
- S22primarycompany filing2026-10-26deal consideration, earn-out, closing conditions and timing
$350,000,000 in cash
backsC41 - S23primarycompany release2026-10-26what ONRAD does and the revenue and adjusted EBITDA RadNet expects it to add
ONRAD is expected to add $190 million to $200 million of Revenue and $60 million to $65 million of Adjusted EBITDA
backsC41C42