Why a startup is teaching human brain cells to play “Doom”

Cortical Labs trained a biological computer made of ~200,000 human neurons on a chip to play 'Doom' (having earlier taught cells to play 'Pong'), showcasing progress in neuron-based computing. The startup aims to scale neuron-packaged, energy-efficient 'biological computers' for conventional data-centre racks as a potential complement or alternative to silicon chips; this remains early-stage proof-of-concept with significant technical, scalability and regulatory hurdles before commercial impact.

Analysis

Biological compute platforms — if they deliver even a 5-10x energy or footprint advantage on narrowly defined pattern-recognition or low-precision inference tasks — change the marginal economics that govern where and how certain AI workloads run. That shift would be felt first in rack-level procurement and PUE math: hyperscalers could re-architect tiering for inference workloads, reducing peak-PoE and cooling CAPEX for a subset of services within a 3–10 year window. Expect early deployments to target cost-per-inference and latency-insensitive batch tasks rather than general-purpose training.

The most direct, durable winners are not the headline startups but the industrial ecosystem: contract cell manufacturers, lab-automation and reagent suppliers, GMP-grade facility builders, and cold-chain/logistics providers that enable predictable scale-up. Hyperscalers and cloud integrators also gain optionality by sponsoring pilot lines; their balance sheets buy them time to lock preferred supplier economics and IP cross-licenses. Conversely, areas that face slower secular growth include segments of semiconductor capital equipment and commodity GPU cycles that service narrowly replaceable inference loads over a multi-decade horizon.

Principal risks are biological reproducibility, interfacing latency to digital systems, and regulatory/biodefense constraints — any of which can extend commercialization timelines from years to decades. Near-term positive catalysts are reproducible third-party replication, partnerships between startups and a major hyperscaler, and visible CAPEX commitments to hybrid racks; negative catalysts include major regulatory limits or a high-profile safety/containment failure. Time-to-signal: credible replication or hyperscaler pilot announcements likely within 12–36 months; meaningful commercial density (>hundreds of racks) unlikely before 5–10 years.

The consensus technical narrative overweights novelty and underweights systems integration and supply-chain scaling. Positioning should therefore favor industrialized exposure (automation, reagents, cloud integration) and optionality instruments rather than large equity stakes in early-stage pure-play biocomputing names. That reduces binary startup risk while retaining upside to the secular re‑architecting of some AI workloads.

Key Decisions for Investors

• Buy Danaher (DHR) shares — 12–36 month horizon. Rationale: exposure to lab automation, reagents, and bioprocessing scale; asymmetric payoff if biological compute pilots accelerate. Risk/reward: limited downside to cyclic healthcare weakness (~-10–15%) vs 20–35% upside on accelerated adoption and share gains.
• Buy Thermo Fisher Scientific (TMO) 12–24 month calls (or equivalent LEAP) for leveraged exposure to contract cell-manufacturing and cold-chain solutions. Rationale: optionality on higher reagent and consumables demand with low operational correlation to semiconductor cycles. Risk/reward: option premium loss if pilot activity stalls; 3:1 upside if adoption path materializes.
• Initiate overweight Microsoft (MSFT) on the 12–24 month horizon (or buy a modest amount of 1Y call spreads). Rationale: strategic optionality for hyperscalers to integrate hybrid racks and lock commercial partnerships; captures services/hosting value without binary startup risk. Risk/reward: moderate upside (15–25%) vs typical market drawdowns.
• Allocate a small private/venture carve-out (0.5–1% NAV) to crossover rounds or co-investments in validated biocompute startups rather than public micro‑caps. Rationale: best way to capture outsized returns with negotiated governance and downside protections; keep position sizes small due to high technical/regulatory execution risk.