The Next Healthcare Disruptor Won't Be a Healthcare Company

An AI art company just walked into medical imaging. The real lesson isn't about the scanner — it's about where disruption now comes from, and whether the incumbents are ready for it.

If you'd asked anyone in medical imaging a year ago which companies might shake up their world, you'd have heard the usual names — Siemens, GE, Philips — and maybe a well-funded health startup or two. Nobody would have said Midjourney. Midjourney makes pictures from text prompts. It has no hospitals, no radiologists, no history with the FDA, and no obvious business going anywhere near a human body.

On June 18, 2026, it walked into medical imaging anyway. On a stage in San Francisco, the company showed off a machine it says can scan your whole body in sixty seconds — no radiation, no magnets, just sound and water. The headline practically wrote itself: the AI art company is coming for the MRI.

That is the part worth slowing down on. Whether this exact scanner works is almost beside the point (short version: not yet). What should make people sit up is where it came from. More and more, the company that upends a slow, regulated, century-old industry isn't one of the insiders — it's a total outsider with a famous name in some completely different field, walking in through a door nobody bothered to lock. If that can happen to medical imaging, one of the most regulated and expensive industries on earth, it can happen almost anywhere.

Next Healthcare Disruptor

So let's read this the way a sharp incumbent should: by pulling apart what's genuinely built from what's just good marketing. Start with the gap the skeptics jumped on. The scanner on stage doesn't do sixty seconds. It takes about twenty minutes, it's been used on roughly a dozen people, it uses no AI at all — the CEO said so out loud — and it has no approval to diagnose anything. It would be easy to roll your eyes, call it hype, and move on. That's exactly the comfortable mistake, because the gap between the pitch and the prototype isn't a flaw. It's a strategy.

What's actually in the machine

Start with the name, because the name is already marketing. They call it an "Ultrasonic CT" scanner. Normally "CT" means a CAT scan — X-rays fired through your body. This machine has no X-rays and no radiation at all. The "CT" refers only to the math part: stitching lots of angles together into a 3D picture of your insides. The actual signal is plain ultrasound — sound waves, the same kind used to image a baby.

Here's what the prototype really does. You lie on a platform that lowers into a shallow pool of water, surrounded by a ring of something like 358,000 tiny sound emitters firing through you from every direction. The water isn't a spa touch — sound travels into the body far better through water than through air, so the pool is what makes the scan possible at all. A bank of computers then turns the flood of returning sound into a cross-section showing muscle, fat, bone, and organs.

The three machines people keep comparing all work in completely different ways. A CT scan fires X-rays through you — fast and great for bone, but every scan is a dose of radiation, so you can't have them often. An MRI uses no radiation; it puts you inside a powerful magnet and reads the signal your tissues send back, which makes it brilliant at telling soft tissues apart — but it's slow, loud, expensive, and off-limits if you have metal in your body. Midjourney's device uses neither X-rays nor magnets, just sound and its echoes, fired from a full ring instead of the single handheld wand you'd see in a clinic.

            SIGNAL              RADIATION   SOFT TISSUE   SPEED / ACCESS
-----------------------------------------------------------------------------
CT          X-rays              yes         moderate      fast, dose-limited
MRI         magnet + radio      no          excellent     slow, costly, loud
Ultrasonic  sound waves         no          unproven      aims for fast,
CT          (ring of sensors)               (the catch)   cheap, repeatable

Which brings us to the part that matters most. Sound is cheap, harmless, and repeatable — exactly what lets Midjourney dream about scanning you every week. But ultrasound has always had two weak spots. It struggles to match an MRI at telling similar soft tissues apart, and bone bounces sound back the way a mirror bounces light, leaving dark blind spots behind it. Whether a full ring of sensors and a lot of computing power can work around those blind spots and close the gap with MRI is exactly what no independent radiologist has confirmed yet.

Now the part you can actually trust. The underlying physics is real — building images from sound bounced through a water bath has been studied in labs for years. And the hardware isn't even Midjourney's: the imaging chips are licensed from Butterfly Network, a public company whose "ultrasound-on-a-chip" squeezed a whole ultrasound machine onto a single chip, and each prototype packs forty of them. You can trust this part because Butterfly had to tell its shareholders — the deal sits in an official SEC filing, worth up to $74 million over five years. A founder on a stage can say anything; a public company filing a number with regulators is sworn to it. The Butterfly deal is the most solid thing in the whole announcement, precisely because the least flashy party had to put it on the record.

Sorting the real from the sales pitch

The quickest way to read any launch like this is to line up each claim next to where it actually stands today, and note who's vouching for it.

CLAIM (Midjourney)            WHERE IT STANDS TODAY        WHO'S VOUCHING
--------------------------------------------------------------------------
60-second scan                ~20 minutes (prototype)      Company
~60-100x faster than MRI      Unverified                   Company
~10x cheaper than MRI         Unverified                   Company
MRI-comparable images         Not independently checked    Company
Uses AI                       "Not even using any AI yet"  CEO, on record
40 Butterfly chips per unit   True                         SEC filing (sworn)
~$74M to Butterfly / 5 yrs    True                         SEC filing (sworn)
50,000 scanners deployed      Long-term ambition           Company
1 billion scans / month       Long-term ambition           Company
30% of deaths avoidable       No scientific backing        Company
Approved to diagnose          No approval                  Confirmed absent

Read down that last column and the picture is clear. Two lines are sworn to regulators. One is the CEO admitting something inconvenient. Everything else is the company describing the future as if it had already arrived. The sixty-second headline is the cleanest example — twenty minutes today, sixty seconds maybe someday, and they led with the number that isn't true yet. This isn't lying, exactly; it's the normal grammar of a product launch, where the destination gets described in the present tense. The reader's job is to keep translating it back. One simple filter does most of the work: before you repeat a number, ask who'd get in trouble if it turned out false. If the answer is nobody, it's marketing.

Why they're starting with "what are you made of"

Here's the move most of the coverage skated past. Midjourney isn't launching a diagnostic scanner. It's launching a "body composition" device — something that measures what your body is made of, not what's wrong with it. Sounds like a technicality. It's actually the whole game plan.

Measuring body composition means answering how much fat you have, where it sits, how much muscle, how dense your bones are. Diagnosing means answering a very different question — is there a tumor, a blockage, a fracture — and in most countries that's a regulated act. It needs a trained radiologist, usually a doctor's referral, and a device the regulator has approved for exactly that job.

                    BODY COMPOSITION          DIAGNOSIS
-------------------------------------------------------------------
The question        What am I made of?        What's wrong with me?
Needs a radiologist No                        Yes
Needs a referral    No                        Yes
Regulated as        General wellness (light)  Medical device (heavy)
How often you'd use Routinely                 Rarely, gated by access

The reason this matters is a quirk in the rules. The FDA lets "general wellness" gadgets — the bucket a smart scale or a fitness watch sits in — skip the heavy machinery of medical-device approval, as long as they're only for wellness. By calling its scanner a maker of "body composition maps," Midjourney parks itself squarely in that bucket, on purpose. And it isn't a cheap trick: tracking the fat packed around your organs genuinely matters, since that's a real risk factor for heart disease and diabetes. But look at what the choice buys them. They can ship now instead of in 2031. They can start charging people and collecting scans while the slow grind toward real diagnostic approval happens quietly in the background. And they can scan you as often as they like, because there's no radiation to ration. Starting with wellness isn't the modest move it looks like — it's the cleverest part of the plan.

Where this is actually headed

So why would a company that makes art software want to run a chain of body scanners? Two reasons, and neither one is the scanner.

The first is a punchline that's easy to miss: the most famous AI company to walk into medical imaging just shipped a medical device with no AI in it. The CEO said it plainly — "not even using any AI in this yet, just really cool hardware and software." The image-making models Midjourney is known for aren't in the box. For now it's pure physics and signal processing.

The second reason explains the first. The prize was never the scanner — it's the scans. Look at what Midjourney says it wants: fifty thousand machines, a billion scans every month. Take away the hardware and that's a plan to build the largest library of labeled medical images in history. Every scan is a training example. Get the device into millions of hands, not for what it earns per scan but for the data it throws off — data that trains better models, which make a better product, which pulls in even more data.

Body composition  ->  Organ sizes      ->  AI flags          ->  Full diagnosis
(what am I         (liver, spleen,      "something here        with FDA
 made of)          kidney)               looks off"            approval
 ~2027              ~2028                 ~2029-31              2031+

Seen this way, "no AI yet" stops being embarrassing and starts looking like a sequence. Build the hardware. Get it in front of people under the light-touch wellness label. Pile up scans at a scale no hospital could match. Then train the diagnostic AI the whole thing was always meant to feed. Body composition is just the bottom step, and each step buys the data and the credibility to reach the next. The staircase only stands if the images are good enough to learn from and enough people actually show up — neither of which is a given. But if it works, whoever ends up holding the world's largest pile of body scans is in a powerful position, no matter what the hardware looks like.

Why the big imaging companies shouldn't relax

So should Siemens, GE, and Philips be worried? In the short term, honestly, no — and it's not close. Look at the scale these three operate at, all from their most recent quarter. GE HealthCare pulled in $5.1 billion in revenue and is sitting on a record $21.8 billion of orders. Siemens Healthineers booked €5.7 billion (about $6.4 billion), with its imaging division alone — the MRI and CT machines — bringing in nearly €3 billion and growing 6% year on year. Philips turned over €3.9 billion, its imaging-heavy Diagnosis & Treatment unit the single biggest slice at €1.8 billion. Those numbers are the moat made visible: years of hospital purchases locked in with companies that hold decades of FDA approvals, slot into insurance billing codes, and come with whole ecosystems of trained radiologists. A nine-person team with a twenty-minute prototype isn't competing for any of that.

The medium term is where it gets interesting, because Midjourney isn't charging the moat — it's strolling around it. Routine scanning for healthy people is a market the big players barely touch, and you can't defend a market you were never in. There's a narrative risk too: if "why does a body scan need a referral, a hospital, and a four-figure bill?" starts landing with the public and insurers, the incumbents will have to defend their prices out loud. And an outsider building from scratch can wire the data pipeline into the device from day one, where the incumbents have been slow.

                       NEAR TERM (1-3 yrs)     MEDIUM TERM (4-7 yrs)
-----------------------------------------------------------------------
Hospital diagnostics   safe                     mostly safe
Wellness screening     being bypassed           real threat
The AI advantage       pressure to respond      strategic risk
Cost/access optics     early grumbling          serious pressure

This is the oldest pattern in disruption: show up where the big player isn't looking and the rules are loose, move fast because you've got nothing to protect, and only later climb toward the lucrative, regulated core. We've already watched it play out in this exact industry. The Apple Watch arrived as a fitness gadget that no medical-device maker took seriously, and in 2018 it quietly earned FDA clearance for an ECG feature that flags atrial fibrillation — turning a wellness toy into a regulated heart monitor on hundreds of millions of wrists. Nobody at the big diagnostics firms saw a watch coming, either. And the pattern is most dangerous exactly when it looks harmless: a slow prototype tested on a dozen people is what a serious competitor looks like in year one. The mistake isn't losing to Midjourney. It's glancing at that clunky prototype and deciding there's nothing to prepare for.

The real lesson

Midjourney is one company in one industry, but the shape of this should make every incumbent in a regulated, expensive, credential-heavy field — banking, energy, law, insurance, education — sit up. The thing that reshapes your industry might not be a rival you're already tracking. It might be a household name from a totally different business, pointing capital or data you didn't know existed at a side door you assumed was too small to matter. By the time an outsider like that looks like a "real" competitor, the users and the data are already theirs.

A year ago, nobody would have put an AI art company on a list of medical-imaging threats. That's exactly why it belongs there now — and why the question worth sitting with isn't whether Midjourney pulls it off, but whether the giants in your own industry are watching the doors they forgot they had.

Midjourney's Medical Scanner: The Next Healthcare Disruptor Won't Be a Healthcare Company