Frontier Computing

Frontier Computing, founded in 2025 by Michael Domarkas and backed by Y Combinator, operates at the intersection of biotechnology and artificial intelligence. The company is developing scalable biological neural networks to serve as a substrate for machine learning training, addressing the high costs associated with training large parameter count models on traditional hardware. The core of Frontier's offering is a proprietary tissue-culture approach that overcomes the traditional limitations of in-vitro neuron scaling. While academic efforts have generally been confined to approximately 1 million neurons, Frontier has engineered a method to surpass this, with a pathway to scale to over 100 million neurons. This is achieved by solving the vascularization problem, which limits oxygen supply to cells in larger cultures. This technology collocates memory and compute at the substrate level, which enables the training of models with significantly larger parameter counts than what is tractable with conventional GPU architectures. Founder Michael Domarkas, a former Cambridge Natural Scientist, began culturing neuronal tissues with grant funding at the age of 17. His work led to the development of a scalable neuronal culture platform, which he validated by training a biological brain to play the game 'Frogger'. Frontier Computing is now focused on growing exascale biocomputing clusters. The company aims to launch a 500 million biological neuron cluster by January 2027. The company's business model involves providing the biological substrate, interfacing, and cloud infrastructure for third-party clients to train their machine learning models. This positions Frontier as a hardware and infrastructure provider in the biocomputing space, targeting entities that require highly efficient and cost-effective training for complex, long-horizon language reasoning and other advanced AI applications. Keywords: biocomputing, biological neural networks, machine learning substrate, ML training, neuronal tissue, artificial intelligence hardware, exascale computing, biological brains, Y Combinator, organoid intelligence, wetware, substrate-level compute, AI model training, neural culture, reinforcement learning, biotechnology, deep learning hardware, large parameter models, biocomputation, living hardware