A decade after the ‘Godfather of AI’ said radiologists were obsolete, their salaries are up to $571K and demand is growing fast

Radiology appears to be defying early AI displacement fears: U.S. active radiologists have grown about 10% over the last decade, with roughly 4,333 open listings and a 130-day average time to fill as of March. The average radiologist salary has risen to $571,000 in 2025, up 9% year over year, while radiology case loads have increased 25% between 2018 and early 2025. The article argues AI is more likely to automate specific tasks than eliminate the profession, aided by reimbursement rules and the need for licensed physicians to provide final reads.

Analysis

The market takeaway is not “AI won’t hit healthcare,” but that workflow automation often expands rather than collapses demand when the bottleneck is throughput, not demand creation. Radiology is a useful template for how generative AI can compress the admin layer while increasing the volume of billable work, which is bullish for the picks-and-shovels layer that improves utilization rather than displacing the human credentialed signer. The second-order effect is margin tension for labor-heavy service providers: productivity gains accrue slowly because reimbursement, liability, and licensure keep the human in the loop.

For NVDA, the relevant edge is not direct radiology replacement but rising compute intensity in clinical imaging, triage, and report automation embedded inside enterprise healthcare IT. The durable bull case is that every “human-in-the-loop” deployment still adds model inference, storage, and workflow integration demand, while healthcare’s regulated environment slows commoditization. That said, the article also implies a ceiling on near-term monetization: adoption can be real without translating into a step-function headcount reduction, which limits how fast investors should underwrite revenue inflection from vertical AI.

SNAP is the odd loser by analogy: the market is increasingly sensitive to management teams using AI as a pretext for labor cuts, but this article underscores that AI narratives can overstate substitution and understate augmentation. If labor market weakness proves exaggerated, ad demand may hold up better than feared, making the downside from cost-cutting headlines more tactical than structural. NFLX and NYT are effectively neutral here; the key read-through is broader skepticism toward doomsday labor narratives, which should cap valuation compression tied to “AI disruption” fears in content and media until there is actual evidence of sustained substitution.

The contrarian point is that the real risk is not immediate displacement but career-pipeline damage: if students internalize an overhyped replacement story, supply shortages persist longer and wage inflation becomes stickier. That can keep healthcare labor expensive for years, but it also means the immediate AI trade is probably in workflow vendors and enterprise software, not in shorting the end profession. The underappreciated catalyst is regulatory: as long as final-read requirements and liability remain with physicians, AI adoption will be additive to volume and compliance spend rather than destructive to the labor base.

Key Decisions for Investors

• Long NVDA on a 3-6 month horizon: favor any pullback driven by ‘AI overhype’ headlines, because regulated vertical AI should increase inference and workflow-optimization demand even without full job replacement. Risk/reward is attractive if the market underprices healthcare and enterprise AI adoption as a durable compute tailwind.
• Pair trade: long NVDA / short a basket of labor-automation beneficiaries with fragile monetization claims over 1-2 quarters. Thesis: the market is rewarding headline AI efficiency stories faster than actual P&L compression is arriving; NVDA captures real picks-and-shovels spend, while the shorts are more exposed to hype fade.
• Avoid chasing the short-SNAP-for-AI-layoff narrative for now; if anything, use strength to fade short-term cost-cutting optimism only after ad spend data confirms weakness. Time horizon: 1-3 months. Risk: management can still realize near-term margin gains even if the strategic AI thesis proves overstated.
• Initiate a small long on healthcare staffing/utilization beneficiaries or imaging workflow software names, if available, rather than shorting radiology adjacencies. The article argues the bottleneck is throughput and reimbursement, so the cleaner trade is on companies monetizing more studies per clinician, not on an outright collapse in clinician demand.
• Do not add to bearish NFLX/NYT AI-disruption bets on this headline alone; the better expression is a wait-and-see stance. Over the next 6-12 months, the burden of proof remains on AI to show it can replace integrated human judgment, not just automate discrete tasks.