AI Models Built From Rat Brains Just Got Closer to Reality

The Biological Computing Company (TBC) is putting a “rat brain” derived AI model into a limited AWS preview for select customers starting Tuesday, aimed at video generation. Amazon and TBC say a broader enterprise rollout is expected soon.

That’s a distribution leap for a field that usually lives in wet labs and academic demos. If biology-derived efficiency tricks can ride the same cloud rails as ordinary generative models, the fringe gets a real customer funnel — and a real bar for whether the gains hold at longer videos.

We’re intrigued and skeptical in equal measure. TBC’s cofounder Alexander Ksendzovsky describes encoding images into biological material, watching how neurons process them, then building software that mimics the process. AWS’s Deap Ubhi likes the pragmatism: improve existing visual models rather than reinvent the transformer. That’s the right product instinct if the claims hold.

The company says its model is up to five times faster at video generation than the unnamed open-source frontier video model it compares against, with lower inference cost. They won’t name the baseline. Previously it ran only through Bluesky Compute; AWS scales the addressable market overnight. Jeff Dean, listed as an investor, reportedly pushed them to prove value on video benchmarks before broader claims — a sensible forcing function.

Context matters. TBC raised $25 million led by Primary Venture Partners, then another $25 million after that March close — more than $50 million total, with the second round previously unreported per WIRED. A San Francisco lab of about 35 people works with rat brain cells and human stem cells on 3Brain multi-electrode arrays. Amazon already works with Cortical Labs’ wetware-as-a-service offerings; this is not AWS’s first biology bet.

Useful AI would mean cheaper, faster video inference without a mystery tax on fidelity. Ubhi’s own caution is the one we’re watching: if customers push toward ten-minute or hour-long video, does consistency hold? Biology-inspired software that only wins on short clips is a demo. Biology-inspired software that survives long-form coherence is a category.

I’m not treating “rat brain AI” as sci-fi branding. It’s a lab pipeline translated into cloud software. The harm path if this scales carelessly is hype laundering — wetware mystique covering ordinary model tricks. The upside if it scales honestly is real efficiency for a compute-hungry modality. Demand the baselines.

WIRED reporting by Lauren Goode, September 22, 2026. Funding figures and performance claims are attributed to TBC and AWS sources in that piece; the open-source comparison model is unnamed.