The Doximity 2026 Physician Compensation Report Found That 46 Percent of Physicians Polled Are Considering Early Retirement. That Is Up 12 Points From the Year Before.

Forty-six percent. That is the share of physicians who told Doximity in June 2026 that they were considering early retirement. Doximity reported that number as 34 percent the year before.

The 2026 Doximity Physician Compensation Report, released August 25, draws on nearly 23,000 physician surveys collected in 2025 as part of a seven-year dataset. Alongside it, Doximity fielded a June 2026 poll of more than 600 physicians on career satisfaction and early retirement intent, a separate instrument from the compensation survey. That poll produced the 12-point jump. The compensation dataset found physician pay grew just 2 percent from 2024 to 2025 (its slowest rate in several years) and that the pay gap between surgical specialists and primary care physicians widened for the first time after three consecutive years of modest narrowing.

What the Survey Measured

The June 2026 poll found that 82 percent of physicians said they are overworked, a high share even as AI tools have spread through clinical settings. The compensation picture from the 23,000-survey dataset is a mix of slow growth and uneven distribution. Interventional radiology gained 10.8 percent and neurosurgery gained 10.7 percent, leading all specialties.

The surgical-to-primary-care pay gap reached 90.1 percent in 2025, up from 87.3 percent in 2024. The shift matters less in absolute terms than in direction. From 2022 through 2024, the gap had been closing slowly. That generated some optimism about rebalancing. The trend reversed in 2025. Primary care physicians are now falling behind surgical specialists again after three years of relative progress.

The report also tracked how physicians are thinking about AI. About two-thirds said staying current with AI tools would give them a meaningful earnings edge over peers who don’t. Roughly a quarter expect AI to increase their compensation within the next year.

Who Keeps the Dividend

The more revealing finding may be this: 44 percent of physicians told Doximity they believe they should be the primary financial beneficiaries when AI allows them to complete more clinical work in the same amount of time.

The question was normative: who should benefit, not a prediction about who will. In a traditional employment or hospital-employed setting, when AI reduces the time needed to complete a chart or coordinate a referral, the efficiency gain flows to the organization first. A physician’s contracted income doesn’t automatically adjust because throughput increased. The concern is structural, not cynical.

In a DPC practice, the physician sets the terms of the relationship with patients and controls the economics of the practice directly. A DPC physician who adds an AI tool that frees 40 minutes per day makes the call about what happens to those 40 minutes. That is a structurally different position from the default in a hospital-employed setting.

Why the 12-Point Jump

The publicly available findings don’t point to a single cause for the acceleration in early retirement intent. The data suggests a convergence: pay growth at its slowest point in years, a pay gap that just reversed direction after improving, and a workforce already running above 80 percent on the overwork measure. Physicians who were waiting to see if compensation conditions would improve had three years of modest data to support that hope. The 2025 data cuts against it.

Administrative burden sits underneath a lot of this. Documentation, prior authorization and insurance-related tasks are structural features of fee-for-service practice, not a volume problem that scaling up or down would fix.

The Case Against Reading This as a DPC Story

The strongest version of the skeptical read deserves stating, because this post has leaned on a claim it hasn’t fully earned.

Start with the number. The 46 percent comes from a poll of more than 600 physicians, self-selected from Doximity’s own membership, not from a probability sample of the workforce. A 12-point move in a single year is large enough that some of it could be a change in who chose to answer rather than a change in what physicians want. The 23,000-response compensation dataset is the sturdier of the two instruments, and it is not the one that produced the retirement figure. Neither instrument breaks results out by practice model, so nothing here measures DPC physicians against anyone.

Then the dividend argument. A DPC physician does control what happens to 40 minutes that an AI tool gives back. Controlling the time is not the same as capturing the money. DPC revenue is a fixed monthly membership fee, so freed minutes don’t raise income unless the physician adds members, and adding members works against the small panel that makes the model function. The employed physician loses the efficiency gain to the organization. The DPC physician often has nowhere to bank it either.

What survives is narrower than the section above implied. A DPC physician decides what the recovered time is for, whether that’s a longer visit, an earlier evening, or nothing at all. That’s a claim about who holds the decision, not a claim about pay. And intent is intent. The poll asked what physicians are considering, which is a different measurement from what they go on to do. A reader who wants evidence that owning a practice changes any of this will not find it in the 2026 report.