Midjourney unveils self-funded ultrasonic body scanner

An image generation company building a 358,000 element ultrasound ring for a spa near Union Square is not the pivot I expected to read about this week, but Latent Space's AINews recap lays out exactly that announcement from Midjourney's David Holz.

The hardware claim is the part that grabbed me. Holz is reportedly framing the Midjourney Scanner as the 'first new whole-body medical imaging modality in 50 years,' using ultrasound rather than radiation or magnets. The prototype is described as 40 systems arranged in a 70 cm ring with 358,000 ultrasonic elements total, capturing around 17 GB per second and producing roughly 40 GB of data per body slice at about 0.5 mm resolution. A current scan takes approximately 20 minutes; the stated target is several hundred slices in 60 seconds. About 12 people have been scanned so far on a Gen 1 prototype, by a nine-person team.

The business model is the louder claim. Midjourney says it is self-funded, has no investors, and can fund the first spa itself. That location is reportedly a roughly 25,000 sq ft, four-floor spa near Union Square in San Francisco with hot tubs, saunas, cold plunges, gym facilities, and nine to ten scanners, targeting an end of 2027 opening. The company has floated a figure of around $20 billion to scale to thousands of spas globally, which is the kind of number you note but do not bank on.

The honest caveat is that none of this is a clinical diagnostic product yet. Holz's framing is 'as powerful as an MRI' and 'as casual as a trip to the spa,' but the regulatory path the company has reportedly started discussing with the FDA leads through body composition, because that is considered easier. The reporting does not tell you what clinical validation looks like, how a scan gets paid for, or how a wellness pitch avoids becoming an incidental-findings problem for everyone who walks out of one.

What is interesting beyond the spa branding is the data shape. If a small team can actually capture body data at the density and rate it claims, the bottleneck for a lot of AI-in-health work stops being access to imaging and starts being what models can do with cheap, dense, longitudinal scans. Whether or not the first spa opens on time, that is the thread worth pulling on.