January 2026
258 AI algorithms approved by the FDA in 2025 — a record year
Thomas Hagemeijer
Founder & CEO, HGM Advisory
Key takeaway
The FDA approved a record 258 AI-enabled medical device algorithms in 2025, bringing the cumulative total to 1,387. Radiology dominates with 79% of approvals, but the shift toward treatment-planning AI and growing international competition signal a maturing market.
In 2025, 258 AI algorithms were approved by the FDA, the highest number so far. 1,387 AI-enabled medical devices have been approved total, mostly in radiology (79%), followed by cardiovascular (10%) and neurology (5%). Top 20 companies led by GE HealthCare, Siemens, and Shanghai United Imaging.
A record year: 258 new approvals
The growth trajectory has been remarkable: from 29 approvals in 2017 to 258 in 2025, a 31% CAGR. The 510(k) pathway accounted for 89% of approvals.
Notably, 2025 saw the FDA issue its first guidance on ‘continuously learning’ AI algorithms — systems that update their models after deployment.
Breakdown by specialty
Radiology: 79% (204 approvals). Cardiovascular: 10% (26 approvals). Neurology: 5% (13 approvals). The most significant trend is the shift from diagnostic-only AI to treatment-planning AI — approximately 12% of new approvals included treatment recommendations.
Pathology AI remains notably underdeveloped with only 4 cumulative FDA approvals, representing a significant white space.
Top companies by FDA approvals
GE HealthCare and Siemens Healthineers lead. The most notable trend was the rise of Chinese manufacturers — Shanghai United Imaging entered the top five with 34 approvals.
| Rank | Company | Total Approvals | Primary Specialty | HQ |
|---|---|---|---|---|
| 1 | GE HealthCare | 83 | Multi-specialty | Chicago, US |
| 2 | Siemens Healthineers | 71 | Radiology / Cardiology | Erlangen, Germany |
| 3 | Philips | 52 | Radiology / Cardiology | Amsterdam, Netherlands |
| 4 | Canon Medical | 38 | Radiology | Tochigi, Japan |
| 5 | Shanghai United Imaging | 34 | Radiology | Shanghai, China |
| 6 | Aidoc | 29 | Radiology (triage) | Tel Aviv, Israel |
| 7 | Viz.ai | 24 | Neurology / Cardiology | San Francisco, US |
What this means for the market
The market is consolidating around two tiers: MedTech incumbents that embed AI into hardware, and pure-play AI companies that build platforms on top of existing infrastructure.
Reimbursement remains the bottleneck — fewer than 50 of 1,387 approved algorithms have established CMS reimbursement pathways. Closing the gap between approved and deployed algorithms is the defining challenge for AI in medicine.