An AI model shaped by patterns observed in living rat neurons is moving from a specialist cloud provider into Amazon Web Services. Selected AWS customers will receive limited-preview access to technology from The Biological Computing Company, or TBC, beginning Tuesday.
The model is designed to improve video generation rather than replace conventional AI architecture with living tissue. TBC stimulates neural cells, records how they process encoded visual information and translates the resulting patterns into software that modifies an existing generative model.
Living Cells Inform a Software Model
TBC researchers work with rat brain cells and human stem cells placed on multi-electrode silicon arrays made by Swiss company 3Brain. The arrays deliver electrical stimulation to the neurons and capture their responses. The startup analyzes that activity for computational patterns that can be applied to AI.
Images were a practical starting point because the physical positions of electrodes affect their interaction with neurons. Visual information can be mapped onto that grid more naturally than text, according to the company. TBC then used what it learned about neural image processing to fine-tune video-generation software.
The approach differs from systems that keep living neurons inside a working computer. AWS also works with Cortical Labs, which combines lab-grown neurons and silicon and sells a biological computing device for laboratories. TBC instead uses biological activity as a research input while delivering a conventional software model to customers.
AWS startup technology director Deap Ubhi described that as a pragmatic choice: the company is working within current generative-AI standards and asking whether neural insights can make visual models more efficient.
Funding and AWS Expand the Experiment
Neuroscientists and neurosurgeons Alexander Ksendzovsky and Jon Pomeraniec founded TBC in Baltimore four years ago. The company raised $25 million earlier this year in a round led by Primary Venture Partners, then secured another $25 million after that round closed in March. WIRED reports total funding above $50 million.
TBC now has 35 employees at a San Francisco research and development laboratory. Its software was previously available through neocloud provider Bluesky Compute. AWS distribution could expose the model to a much broader group of enterprise customers, and both organizations expect access to expand beyond the initial preview.
The startup claims its model can generate video as much as five times faster than the open-source model on which it runs while substantially lowering inference cost. It has not identified that comparison model beyond calling it a frontier video generator.
Those are company claims rather than independently verified results described by WIRED. The undisclosed baseline makes it difficult to judge how broadly the reported speed improvement applies.
Longer Video Will Test the Claimed Gains
TBC chose video partly because established benchmarks make it possible to compare performance on known tasks. Investor and AI researcher Jeff Dean encouraged the company to begin there before attempting broader uses for its neural research.
The next question is whether efficiency survives more demanding workloads. Ubhi noted that customers may want videos lasting ten minutes or an hour, requiring the model to preserve events and visual consistency over time. A speed gain on shorter benchmark material may not translate into fidelity at that scale.
Biological computing also carries operational complexity absent from ordinary software development. Cells must be kept alive or preserved, monitored and connected to digital systems in a laboratory. Progress depends on both biological technique and model engineering.
The AWS preview is therefore a commercial test, not proof that biologically inspired AI is ready for general deployment. Customer use can show whether TBC’s patterns improve real production workloads, whether the cost advantage persists and whether longer output remains coherent. The model’s unusual origin is compelling; its value will be determined by reproducible performance.
