Computers made of neurons

Why this conversation

Most of the compute behind AI runs on silicon chips. At Cloudflare Connect 2026 in San Francisco, the host met two founders who are betting on something else: neurons grown on a dish. Ali Amjad and Zubair Zafar explained what a biocomputer is, why it might help robots learn from less data, and how far along they are. Alec Reynolds was recorded at the same event.

From hackathons to co-founders

Ali left a job in Canada five months before the conversation and came to San Francisco to explore startups. Along the way Ali started building in biocomputing and went to a lot of hackathons, which is where the two met. Zubair graduated about six months earlier and also came to San Francisco to build a startup, but first joined a different one. They did a couple of hackathons together. “We won a lot of them. And now we are co-founders. We’re building our own startup.”

What a biocomputer is

Ali’s definition: “the biocomputer is basically you grow neurons on a dish and connect with multi-electrode array to teach those neurons to do the work.” Think of it as an alternative to silicon, though it is still early stage, so Ali expects it to be a supplement to silicon chips. The bet behind it: humans learn faster with less data. Can that be replicated in robotics, with robots that learn from less data using biological computers?

Where they stand

They are working towards a prototype. Ali already has access to one biocomputer and makes algorithms and software on it, while they are in the process of building their own. Zubair comes in on the software, and a third co-founder, a postdoc in biology, will help build the computer itself.

Zubair found the field a couple of months before the conversation and expects it to be “really viral”, because training LLMs, VLA models and VLM models on silicon chips costs a lot more than it would on a neuron chip. The hypothesis is that biocomputers could train more models with a minimal amount of compute.

Asked how people can connect, they point to LinkedIn: “If you’re interested in biocomputing, please reach out.”

When it comes out

The episode is edited and in the release order. Like every episode it goes out on YouTube first, then on Apple Podcasts, Spotify and Amazon Music.