The AI image company unveiled a full-body scanner this week — and a spa to put it in. The real product may be the dataset, and the distance between the pitch and the proof is the story.
The company that taught a generation to type a sentence and get back a photorealistic image now wants you to step into a pool of warm water and come out with a map of your insides.
Midjourney announced a new division this week, Midjourney Medical, built around a single piece of hardware it calls the Midjourney Scanner. It is the company’s first physical product, and it has nothing to do with generating pictures of cats in spacesuits. Founder David Holz described it onstage in San Francisco as a full-body ultrasound device that is, in his words, “in many ways superior to even MRI machines,” with no radiation and none of the heavy magnets an MRI requires.
For a company synonymous with AI, the most revealing line was almost an aside. The scanner, Holz said, is “not even using any AI in this yet, just really cool hardware and software.” That should tell you where the artificial intelligence actually lives in this plan — and it isn’t in the machine.
How The Machine Is Supposed To Work
The technology is a decades-old approach called ultrasound computed tomography, or USCT. Instead of a handheld probe sweeping across your skin, a ring of transducers surrounds the body underwater and fires sound waves from every angle at once, reconstructing a full 3D volume from the echoes.
A user steps onto a platform that lowers into a shallow pool at about five centimeters per second. As you descend, the ring images you slice by slice. Midjourney says the finished scan will eventually take 60 seconds and produce a sub-millimeter map that “looks a lot like today’s MRIs.”
The physics here are real and well understood. USCT has been studied since the 1950s and recently earned FDA clearance for breast cancer screening. What no one has done before is attempt it across the entire torso and legs at consumer scale — which is the part Midjourney is betting the division on.
The Partnership That Makes It Possible
The scanner is not built on Midjourney’s own silicon. According to a securities filing from November 2025, the company signed an exclusive co-development and licensing deal with Butterfly Network, the publicly traded maker of ultrasound-on-a-chip technology.
The terms are not small. The filing describes a $15 million upfront payment and roughly $10 million a year over a five-year term, with milestone and revenue-sharing payments that could push the total toward $74 million.
That detail reframes the announcement. Midjourney is licensing a proven semiconductor platform and wrapping it in an experience — which is a marketing and distribution problem far more than a physics one.
The Gap Between The Video And The Prototype
Here is what the launch reel does not say. The current prototype reportedly takes about 20 minutes per scan, not 60 seconds. It has been used on roughly a dozen people. And it is being built by a team of about nine, led by Ahmad Abbas, who joined Midjourney in 2023 after working on hardware for Apple’s Vision Pro.
None of that makes the project fake. It makes it early. The 60-second figure is a roadmap target gated by how fast data can move from the transducers to the compute cluster — the system generates something on the order of 17 gigabytes of raw acoustic data per second. The company’s own timeline points to a custom-silicon third-generation scanner around 2028 before the experience matches the pitch.
The instinct to invoke Theranos is understandable and, in this case, mostly wrong. The underlying science is sound, the hardware partner is credible, and Holz was unusually candid about the prototype’s limits. The risk isn’t fraud. It’s the long, unglamorous distance between a working demo and a regulated medical product deployed at scale.
The Spa Is The Strategy
The cleverest move in the entire announcement is the venue. Midjourney isn’t opening clinics. It’s opening a roughly 25,000-square-foot “Midjourney Spa” near Union Square in San Francisco at the end of 2027, with hot tubs, saunas, cold plunges and ten scanners, running 24 hours a day.
This is positioning, not décor. By selling “body composition maps” inside a wellness destination, the company stays inside the FDA’s general wellness lane, which lets non-invasive measurement devices that make no disease claims operate without full medical clearance. Consumer whole-body imaging outfits like Prenuvo and Ezra have used the same path.
The scan, as Midjourney frames it, is a “side effect” of a nice afternoon. That framing keeps regulators at arm’s length while the company accumulates the one thing it actually needs: scans, by the millions.
What They’re Really Building
Read the long-range numbers and the thesis becomes obvious. Midjourney wants more than 50,000 scanners worldwide and the capacity for a billion scans a month by 2031. A billion scans is not a membership plan. It is a training set.
A longitudinal, full-body imaging dataset of that size is exactly the raw material you would want to teach a model to spot abnormalities in healthy people before symptoms appear. The scanner doesn’t use much AI today. The point is what the scanner collects so AI can be used tomorrow.
That ambition runs straight into a well-documented problem. Whole-body screening of people with no symptoms turns up incidental findings in 20% to 40% of scans, yet only a small fraction ever require treatment, as University of Michigan radiologists have noted. At a billion scans a month, even modest rates imply hundreds of millions of ambiguous results a year, each one demanding a clinical decision and producing a worried customer.
Midjourney’s announcement claims that enough early imaging could help the world avoid 30% of deaths and half of all healthcare costs. No peer-reviewed evidence is offered for either figure.
The Tell Worth Watching
Midjourney has no outside investors and calls itself a community-backed research lab, and it says it can fund the first spa on its own. That independence is genuinely unusual, and it’s why the company can announce something this strange without a board asking why.
It also means the discipline has to come from somewhere else — from regulators, from radiologists, and from whether anyone actually shows up to be scanned. The image-generation business was a product. This is an infrastructure bet dressed as a wellness brand, and the scoreboard for it won’t be a launch video. It’ll be FDA submissions, scan volume, and the first independent radiologist willing to say the pictures are as good as the company says they are.
