RadNet's AI Draft Reporting Crosses a Threshold — and the Tape Is Paying Attention
Record Q2 revenue and adjusted EBITDA, a fresh FDA-cleared breast ultrasound AI, and a 90-day +38.6% rally position RadNet as a company-unique AI-radiology mover in a tariff-obsessed tape.
RDNT · Earnings Call · 2026-08-10
A step-change in draft reporting
RadNet's second-quarter print on Aug 10 was, on its face, a broad-based beat: “Total company revenue increased 25% to $623 million from $498 million in last year's second quarter, and total company adjusted EBITDA increased 22.7% to $99.7 million” — Howard G. Berger, CEO · 2026-08-10. But the genuinely new signal on the call was the draft reporting milestone. The FDA 510(k) clearance for DeepHealth breast ultrasound turns one of the most operator-dependent exams in radiology into an breast ultrasound that — as Kaes Wesdorp put it — “automates lesion detection, measurements, characterization and reporting” — Cornelis Wesdorp, President and CEO of Digital Health · 2026-08-10 and generates standardized BI-RADS draft reports. Critically, the breast pool is 4x the size of the thyroid program that first validated the model.
The adoption math is what makes this a step-change rather than an incremental feature. On the thyroid program, over 90% of AI-generated draft reports have been accepted by radiologists for final sign-off without further edits. Management's stated plan:
we expect close to 15% of RadNet volumes to run through AI powered autodraft solutions by year-end. Growing to over 50% by end of the second quarter of 2027.
That trajectory matters because RadNet's own labor costs — the pressure on salary called out repeatedly in the transcript — keep rising for technologists and radiologists. Draft reporting attacks both the radiologist interpretation bottleneck (30%+ faster reporting) and the sonographer scan-time (about a third faster), effectively adding capacity without adding headcount. As Dr. Berger framed it, “our primary investments have been made in those modalities, the routine modalities, X-ray, ultrasound, and mammography, which comprise 70% of our volume” — Howard G. Berger, CEO · 2026-08-10 — precisely where AI draft reporting has the widest surface area. This echoes the prior quarter's strategic framing that “the future of radiology, the future of imaging is more about software than it is hardware” — Mark Stolper, CFO · 2025-11-10.
Digital health compounding, and the tape is voting
The Digital Health segment is no longer a side show. ARR reached $106M, up 97% y/y, with 63% now from external (non-RadNet) customers and a TCV funnel that more than doubled during the first six months to $224M. The prior quarter's “funnel that has developed now toward and above the $150 million total contract value mark” — Kaes Westorpe, President and CEO of Digital Health · 2026-05-11 has now concretized into signed deals — the commercial momentum is real, with ~$21M TCV closed in Q2 alone and guidance reaffirmed at $135–145M revenue and $10–12M adjusted EBITDA for the segment.
The market is paying attention. RDNT is up roughly 38.6% over the last 90 trading days, on top of a +3,558% cumulative return since 2010, and the stock sits within 11% of its November 2024 all-time high. In a global tape that spent 2026 fixated on tariff deadlines, Iran conflict lines, and batch-zero ERCOT drama (all prominent in the 2026 global keyword trajectory), RadNet is running a company-unique AI transformation story — with none of its reporting peers touching automated graph reporting or draft reports on this scale.
Deploying capital: leverage, hospitals, and the 2027 rate backdrop
Financially, management used the strength to opportunistically upsize the balance sheet: a $250M incremental term loan (SOFR + 200bp, repriced 25bp lower on both the term loan and revolver), leaving ~$726M cash and net leverage at 1.8x. Proceeds are earmarked for acquisitions, de novos, and health-system partnerships — 157 of 442 centers are now inside JVs, and inbound hospital interest is rising as the proposed Medicare fee schedule for 2027 includes a site-neutrality provision that would cut hospital outpatient imaging reimbursement 30–50% on certain non-contrast studies. RadNet's own volume-weighted read of the 2027 proposal is "almost net neutral" (a negative of less than $1M on ~$2.4B revenue), despite the 1.68% conversion-factor cut.
The cash conversion is the quiet tell. Free cash flow (less SBC) printed $48M in the June quarter, up 266% y/y, with DSOs at a near-record 31 days — evidence that the center-level workflow investments are flowing to the bottom line rather than getting absorbed by labor comp inflation. The counterpoint sits in the GAAP P&L: operating margin swung to -5.7% in the June quarter — the same quarter adjusted EBITDA set a record — highlighting how stock-based comp and integration spend mask underlying cash profitability. Adjusted EBITDA, the metric management runs the business on, hit a quarterly record, and the full-year imaging-center EBITDA guidance was raised to $345–358M.
What to watch
The next proof points are the breast ultrasound reimbursement ramp (the same 0690T CPT code path as thyroid, with 30–40% of payers already covering that T-code) and the second wave of FDA submissions — MR spine autodrafting, vascular ultrasound, and next-gen chest X-ray. If the 50%-of-volumes-by-mid-2027 draft-reporting target holds, RadNet is essentially building an internal SaaS cascade: every deployed AI solution is simultaneously a cost lever at its own 3M+ annual exams and a sellable product to hospitals and outpatient peers. That dual-use dynamic is the rare kind of company-unique story — health system partnership interest feeds both revenue lines, and with the 90-day tape at +38.6%, the market is already beginning to price the draft-reporting transition in.