We develop wetware, hardware, and software for the next frontier of power efficient, future-proof computing: biological computers.
We are building an alternative compute substrate. One that does not rely on silicon.
A living neural computing platform that operates at ultra-efficient power levels,
self-optimizes, and self-programs.
In other words, custom brains.
Even under our highly constrained bootstrapped resources, in 6 months, we have developed in-house:
- An Ultra-Low-Cost Culture System that is 300 times less expensive than the academic standard and is optimized for biocomputing performance.
- An Ultra-Low-Cost Neural Interface Device that is 100 times less expensive than commercial equivalents.
Both of which are scalable by design.
Chimerion aims to solve practical problems using our technology. We are focusing on developing smarter autonomous systems, ultra-low power datacenters, biology-powered AI, and enhancing the neural capabilities of humans.
Power Efficiency: Biological neural networks operate approximately 550,000 times more efficiently than the best silicon computing platforms. The most powerful supercomputer to-date, the El Capitan operates at 2.746 exaFLOPs / 30 Mega Watt, costing about $138,312 / day of operation. Over $50 million per year. The Human Brain operates at 1 exaflop / 20 Watt. Approximately 550,000 times more efficient than the best supercomputer.
Self-Optimization: Inherent neural plasticity of neural substrate drives continuous performance improvement and reduced algorithmic overhead. In other words, self-reprogramming.
Obsolescence Resistance: The lifetime of a typical NVIDIA GPU in an AI datacenter is approximately 18 months. Brains take decades to begin to degrade. With our biocomputing culture system, we could almost entirely avoid hardware obsolescence.