RadNet is the largest US outpatient diagnostic imaging operator (418 centers, $2.04B revenue). The bull thesis is a two-part stack: dominant imaging network with pricing power, plus a high-margin AI software layer (DeepHealth) building on top. The stock is down 19% in a week on the Gleamer acquisition announcement. All 8 analysts rate it Buy with a $92 mean target (+52% upside).

The question: is the selloff a buying opportunity, or is the market right to be nervous?

The Imaging Machine

The core business is fine. Better than fine.

$2.04B revenue (+11.5%), $299M operating cash flow (+28%), net debt 1.0x EBITDA. Same-center advanced imaging growth hit 9.6% in Q4 — above guidance. The mix shift toward PET/CT (now 28.6% of procedures, up 178bps) is structural: Alzheimer's diagnostics (Leqembi/Kisunla require amyloid PET confirmation) and prostate cancer screening are secular growth drivers with high reimbursement rates. DSO at a record low 29.5 days.

151 of 418 centers operate as hospital joint ventures — sticky, hard to replicate, and a pipeline for volume. The serial acquisition machine added 73 net new centers over two years. This is a real cash engine with genuine competitive moats in local markets.

The AI Segment: Profitability Always 18 Months Away

Digital Health revenue: $92.7M (+41.1%). ARR: $75M. Customers: 2,075. Sounds impressive until you decompose it.

The growth was "driven primarily by external expansion, not internal volume increases" — the filing's own language. iCAD ($110.7M all-stock, July 2025) added 1,500 customers overnight. Strip inorganic contribution and organic growth is far more modest.

More concerning: GAAP operating loss widened to -$32.3M from -$14.9M from -$5.2M. Three consecutive years of accelerating losses. Adj EBITDA only improved $0.9M despite 41% revenue growth. The segment is spending faster than it's growing. $20M in non-capitalized R&D is real cash burned.

Management guides iCAD integration breakeven mid-2026 and Gleamer EBITDA positive mid-2027. They've guided to profitability timelines before. Each acquisition resets the clock.

Gleamer: The Big Bet

Closed March 2 (same day as the 10-K): €215M upfront + up to €15M contingent for a Paris-based radiology AI company. 700+ customers, 44 countries, ≈$30M ARR, 100% SaaS model. Creates "largest provider of radiology clinical AI worldwide."

The strategic logic is sound — fills the routine imaging gap (X-ray, MSK) that DeepHealth's existing breast/lung/prostate portfolio doesn't cover. EMEA commercial expansion. But this is the fourth major acquisition in 12 months (iCAD, See-Mode, CIMAR UK, Gleamer), totaling ≈$360M+ in commitments. Pro-forma leverage rises to roughly 1.6-1.8x.

At ≈7x ARR for a segment that has never been GAAP profitable, Gleamer is a bet that scale unlocks margins. Maybe it does. But the pattern — acquire, integrate, promise breakeven in 18 months, acquire again — is the same one that's produced three consecutive years of widening losses.

Factor Decomposition

Factor regression (250d) shows RDNT at 72% idiosyncratic variance — below the 75% threshold. Momentum (MTUM) accounts for 22% of return variance. Healthcare sector (XLV) contributes 6%. Market beta is essentially zero.

The 22% momentum loading matters right now. RDNT ran from $45 to $86 in 12 months. The -19% selloff is partly a momentum unwind, not purely fundamental repricing. Momentum reversals can extend (Paleologo Ch.8: left-tail events, short-side rallies of 200-350% in 2009/2016).

Edge Audit

Zero edge across all factors.

$4.7B market cap. 8 analysts, unanimous Buy. 100% bullish consensus. No open market insider purchases — only stock awards ($7.5M to CEO on filing day). Forward P/E 60x on next-quarter EPS estimate of -$0.18. Short interest 11.8% with 10.8 days to cover — shorts have a thesis too.

The idio component (72%) requires healthcare services expertise: CMS reimbursement modeling, clinical AI adoption curves, FDA pathway assessment, radiology labor market dynamics. We have none of these.

The Bear Case Nobody's Making

SBC jumped 83% to $54.6M. That's real dilution — and it signals management expects to keep acquiring with stock as currency. Each acquisition adds integration complexity. The ARR-to-revenue gap ($75M ARR vs $92.7M GAAP revenue) means $18M in non-recurring implementation fees and hardware — not pure SaaS quality. Opex is growing faster than imaging revenue (+13.2% vs +10.9%). Seven centers were abandoned in 2025 ($8.6M charge vs $2.5M prior year). Equipment disposal losses: $9.8M vs $2.3M.

The honest question: is Digital Health a high-margin AI platform in the making, or a Rube Goldberg machine where each acquisition "synergizes with" the last and profitability is permanently 18 months away?

Conclusion

RDNT is a solid imaging business trading at 60x forward earnings because of an AI thesis that has never generated a GAAP profit and widens losses every year. The imaging machine is genuinely good — $299M operating cash flow, 1.0x leverage, structural PET/CT tailwind. But the AI layer is funded by that cash flow, and every year it consumes more without delivering profitability.

The -19% selloff at RSI 36 with unanimous analyst Buy ratings is a contrarian setup — for someone with informational edge on diagnostic imaging and clinical AI adoption. That's not us. 72% idio (below threshold), 0% edge, outside our coverage zone. No action.

LR: 0.9 — The filing mildly undermines the consensus bull case. Widening AI losses, SBC explosion, and the "perpetual 18-month breakeven" pattern are real concerns that 8 Buy-rated analysts may be underweighting. But "mildly bearish vs analyst consensus" from a position of zero informational edge is worth approximately nothing.

Evidence