Midjourney Previews 2-Petaflop Ultrasonic CT Hardware

Midjourney is moving from generative software into medical hardware. In a 20-minute video published on July 2, 2026, Midjourney engineer Marcin Plaza detailed the construction of a prototype whole-body imaging device called the Midjourney Scanner. The system, which the company calls an Ultrasonic CT, surrounds a human body with thousands of transducers to capture dense 3D internal maps.

Hardware and Compute Architecture

The underlying hardware relies on heavy sensor density rather than traditional magnetic resonance. The current prototype arranges 40 Butterfly Ultrasound-on-Chip imaging modules into a 70 cm immersion ring. Users are lowered into a fluid tank via an elevator mechanism to facilitate acoustic coupling.

This array totals roughly 358,000 individual ultrasonic sensors, capturing 17 GB of raw data per second. To reconstruct sub-millimeter 3D maps from this volume of acoustic input, the scanner utilizes over two petaflops of dedicated processing power. Midjourney targets a rapid scan cycle, though independent reports indicate a performance gap between the prototype and production timelines.

Metric	Current Prototype	Production Target
Sensor Elements	358,000	500,000
Scan Duration	Up to 20 minutes	60 seconds
Resolution	0.5 mm	Sub-millimeter

Commercial Roadmap and Licensing

Midjourney is bypassing the immediate regulatory hurdles of medical diagnostics by initially categorizing the scanner as a wellness product. The first application will focus on body composition mapping, specifically measuring muscle and fat density.

The initial public rollout will bypass traditional clinical settings entirely. Midjourney plans to deploy the first units inside a 25,000 square foot, four-floor spa facility near Union Square in San Francisco in late 2027. The facility incorporates hot tubs, cold plunges, and saunas to accommodate users exiting the wet dunk-tank environment required for the scan.

While Midjourney claims the hardware development is self-funded by its estimated $200 million annual software revenue, SEC filings reveal a five-year, $74 million co-development and licensing agreement with Butterfly Network. The company aims to scale production to 50,000 units globally by 2031.

Diagnostic Viability and Physics Constraints

Medical imaging experts point to fundamental physics limitations that the Ultrasonic CT method has not proven it can overcome. Traditional ultrasound waves cannot effectively penetrate bone or air, severely limiting the hardware’s ability to image the brain or lungs.

Radiologists have also raised concerns regarding the software layer processing raw sensor data. AI upsampling models can generate plausible, highly detailed images from sparse or noisy inputs. In a diagnostic context, generating plausible tissue structures that do not accurately represent the real patient introduces critical medical risk. The device currently holds no FDA clearance for diagnostic use, though Midjourney plans to submit test results incrementally.

If you are building health technology or evaluating AI inference in medical settings, the distinction between generative image enhancement and diagnostic accuracy remains the primary constraint. Validating structural truth against algorithmic interpolation will determine whether this dense sensor architecture scales beyond cosmetic wellness applications.