Human neurons in a bio-computer learn to master the game Doom

200,000 neurons: Cortical Labs' CL1—built on roughly 200,000 human neurons on a multi-electrode array—learned to play Doom within a week, demonstrating novice target-finding and firing. CL1 was commercialized by 2025 and is offered as hardware or via cloud with an open API for researchers to refine encodings, rewards and learning rules. Cortical Labs positions neuron-based processors as potentially complementary to silicon for plasticity and pattern discovery, but current performance remains rudimentary (comparable to a first-time player).

Analysis

Living neuronal substrates create a new, high-margin demand vector for specialized lab hardware, reagents and closed‑loop control software rather than for raw compute. If early adopters scale from pilot labs to commercial pilots, MEA vendors and lab-automation firms could see recurring revenue lift measurable in the tens to low hundreds of millions annually within 2–4 years, driven by high experiment cadence and consumable use‑rates.

Second‑order supply effects favor vertically integrated suppliers of human cell lines, GMP neuronal manufacturing, cryo‑logistics and precision stimulators — firms that can guarantee reproducibility and reduce culture variability will capture pricing power. Conversely, pure‑software AI providers risk being peripheral unless they bundle low‑latency interfacing and closed‑loop control; cloud hosts that enable hybrid experiments will win incremental utilization and premium storage/compute billings.

Key near‑term gating risks are non‑technical: reproducibility failures, a damaging bioethics incident, or regulatory curbs could compress investment and stop deployments for 6–24 months. Positive catalysts that would de‑risk the theme are independent replications, standardized benchmarks and early commercial pilot contracts in niche robotics or signal‑processing where silicon falters; expect meaningful signal one to two years out rather than immediate disruption to mainstream AI infrastructure.

Key Decisions for Investors

• AXGN (Axion Biosystems) — Initiate a tactical 1–2% net long position (12–24 month horizon). Thesis: direct beneficiary from rising MEA deployments and consumable repeat revenue; target ~2.0–2.5x upside if adoption accelerates. Risk controls: hard stop at -40% and reassess on failed third‑party replications.
• NVDA (NVIDIA) — Buy a 9–12 month modest call spread (bull call spread to cap premium) sized for 2–3% portfolio exposure. Thesis: hybrid experiments amplify GPU/cloud training and simulation demand while open APIs increase iteration velocity; payoff 2–4x if enterprise adoption shows momentum within 12 months. Risk: limited to premium paid by using spreads.
• DHR (Danaher) or TMO (Thermo Fisher) — Overweight (3–5% position) across 12–24 months into lab automation, reagents and GMP cell‑supply exposure. Thesis: these incumbents capture the consumables and automation tailwinds from repeated neuronal experiment cycles; expect steady, defensible revenue lift (target 15–30% outperformance vs peers). Hedge: buy 18–24 month 10–15% out‑of‑the‑money puts equal to 25% of position cost to limit downside to high‑volatility biotech shocks.
• Portfolio sizing / contrarian hedge — Avoid overallocating to early neuronal computing startups; treat the theme as optional alpha with <5% total venture/illiquid exposure. If seeking downside insurance, buy a small (<=1% portfolio) put position on a biotech index or use single‑name puts on speculative neuro hardware microcaps rather than increasing core equity exposure.