David Holz's medical bet turns Midjourney into a body-data company

Why it matters

Midjourney is testing whether a self-funded AI lab can build a consumer medical-data network before it has a fully regulated diagnostic product, a strategy that could reshape preventive imaging if the scanner, consent model, and clinical evidence catch up to the pitch.

David Holz is taking Midjourney into medicine with Midjourney Medical, a new division built around a full-body ultrasound scanner and an eventual San Francisco spa that would make internal imaging feel less like a hospital visit and more like a routine wellness appointment, Forbes reported Friday.

The move is not a side quest for an image-model company. It is Holz applying Midjourney's core operating instinct - build an interface that changes how people produce and consume visual data - to the body itself. Midjourney describes itself as a 60-person, community-funded research lab working across "imagination, coordination, reflection, beauty, and human flourishing." In this case, the founder's bet is that the next valuable model will not come from another scraped internet corpus. It will come from a proprietary stream of scans that do not exist yet.

That is why the most important line in the announcement is not the MRI comparison. Holz said onstage in San Francisco that the scanner is "not even using any AI in this yet," according to Forbes. The device is meant to create the data layer first. The AI comes later, if Midjourney can persuade enough people to step into warm water often enough, under consent terms robust enough, with clinical results strong enough, to make the dataset useful.

RuntimeWire reported Thursday that the product pairs an ultrasound imaging machine with a planned spa, while diagnostic use still depends on FDA clearance. The sharper point, after Midjourney's own technical post and Butterfly Network's filings, is that Holz is building a distribution system for imaging before he has a medical product doctors can prescribe.

The scanner is the first wedge

Midjourney calls the device the Midjourney Scanner. The company says a user would stand on a platform that lowers into a shallow pool while a ring of underwater sensors sends ultrasonic waves through the body from multiple angles. Midjourney says the goal is a scan that takes no more than 60 seconds and produces a 3D map of the body down to a fraction of a millimeter.

That is the pitch. The current machine is earlier than the pitch. Forbes reports that the prototype takes about 20 minutes per scan and has been used on roughly a dozen people. That gap matters because Midjourney is not merely promising better medical imaging. It is promising medical imaging with consumer frequency. A 20-minute prototype can produce demos. A 60-second scan can support the usage pattern Holz wants: repeat scans, population-scale comparisons, and eventually longitudinal AI.

The physics are not magical. The approach, ultrasound computed tomography, uses sound waves rather than radiation or the heavy magnets used in MRI. Midjourney's claim is about scale, form factor, speed, and experience: a full-body implementation that feels casual enough to repeat. In its announcement post, Midjourney frames the scanner as a way to collect more health data faster and cheaper, using the phrase "megabytes per second per dollar" to describe the kind of health-information throughput it wants.

That phrase is revealing. Holz is not talking like a hospital operator. He is talking like a systems founder trying to increase data bandwidth between humans and software.

Butterfly gives Midjourney a hardware base

Midjourney is not building the scanner from a blank silicon roadmap. A November 2025 SEC filing from Butterfly Network shows that Butterfly entered a co-development and licensing agreement with Midjourney for Butterfly's semiconductor-based ultrasound technology, software, and backend technology.

The economics make the partnership more than a vague collaboration. The filing says Midjourney agreed to pay Butterfly a $15 million one-time fee and about $10 million annually during a five-year term, with milestone and revenue-sharing payments tied to commercialization of hardware products using Butterfly chips, plus payments for chip purchases. Butterfly has described the agreement as worth up to $74 million in expected payments over five years.

For Butterfly, Midjourney validates its Ultrasound-on-Chip platform beyond conventional point-of-care ultrasound. For Midjourney, Butterfly shortens the path into physical imaging and lets Holz focus on the layer Midjourney understands best: product experience, user desire, and large-scale data infrastructure.

That does not make the project easy. It does make the strategy legible. Midjourney is licensing a core imaging platform, wrapping it in a consumer ritual, and aiming to convert the resulting scans into a proprietary medical-data asset.

The spa is the regulatory and distribution strategy

The strangest part of the announcement is also the most commercially important. Midjourney says its first spa will open in San Francisco in 2027, with hot tubs, saunas, cold plunges, and scanning rooms. Forbes reports that the planned location is roughly 25,000 square feet near Union Square, with 10 scanners operating 24 hours a day.

A spa is not just branding. It is how Midjourney tries to make frequent imaging socially normal before the scanner has broad diagnostic authority. In its own roadmap, Midjourney says it will begin by giving users "detailed body composition maps" and will submit regular test results to the FDA for expanded capabilities.

That distinction is central. The FDA's general wellness guidance covers low-risk products that promote healthy lifestyles and are unrelated to diagnosis, cure, mitigation, prevention, or treatment of disease or conditions. If Midjourney stays in the body-composition and wellness lane, it may have more room to operate as it gathers evidence. If it starts making disease-detection claims, the regulatory burden changes.

This is the line every preventive-imaging company has to walk. Ezra sells full-body MRI with radiologist review. Prenuvo has built a consumer clinic brand around whole-body MRI. Neko Health combines skin, cardiovascular, and blood markers with a physician consultation. Q Bio has long pursued a purpose-built whole-body scanning platform. Midjourney's version is different because it comes from a self-funded AI lab with a community distribution engine and no disclosed venture backers.

Midjourney's advantage is not medical incumbency. It is that people already understand the brand as a place where technology makes hidden visual structure visible. Holz is betting that trust can stretch from prompts and pixels into anatomy. That is an ambitious extension of brand permission, and it will be tested by clinical reality.

The hard problem is what happens after the scan

The upside of routine imaging is obvious: more data, earlier signals, and the possibility of measuring change over time rather than waiting for symptoms. Midjourney goes further, saying its goal by 2031 is to collect billions of scans.

The medical system's concern is just as obvious. Screening asymptomatic people can find things that matter, and it can also find ambiguous findings that trigger anxiety, follow-up procedures, expense, and harm. The American College of Radiology said in 2023 that it did not see sufficient evidence to recommend total-body screening for people without symptoms, risk factors, or family history suggesting disease or serious injury.

Midjourney can avoid calling the first version diagnostic, but users will not experience a scan of their internal organs as a neutral wellness map. They will ask what the findings mean. They will ask whether a spot is dangerous. They will ask who owns the data, who can train on it, whether a doctor has reviewed it, and what happens when an algorithm improves after their scan.

Those questions do not invalidate Holz's bet. They define it. Midjourney Medical is not competing with hospitals on day one. It is competing with the idea that medical-grade body data should be scarce, episodic, and institutionally mediated. The scanner, the spa, the Butterfly deal, and the 2031 scan-capacity target all point to the same thesis: if Midjourney can make body imaging frequent and desirable, it can build a dataset no model lab can download.

That is founder logic at its purest. Holz is taking the company's strongest muscle - turning a difficult interface into something people want to use - and moving it into a market where the consequences are higher and the proof bar is not aesthetic. In image generation, users decide whether the output is good. In medicine, the body decides, the FDA decides, and eventually doctors and patients decide together.

The opportunity is large because the current system leaves most people with too little longitudinal information about their bodies. The constraint is equally clear: a body-data company cannot become a healthcare company on vibes, visual quality, or throughput alone. It has to prove that the scans are accurate, that the interpretations are useful, and that the flood of findings improves health rather than merely increasing intervention.

Holz has made Midjourney's first hardware product a scanner. The real product, if this works, is the repeat relationship with the user's body.