Could data from 100 million species help cure disease? One startup is betting on it

Basecamp Research launched a 'Trillion Gene Atlas' to model biological data at the trillion-gene scale and collect genomic data from more than 100 million species, partnering with Anthropic, Ultima Genomics, PacBio and Nvidia; the startup has raised $85M to date. Since 2023 it has paid royalties to 60 organizations across 21 countries and built provenance systems to track sample origin and attribute downstream value, positioning its Eden models to potentially accelerate drug discovery. The initiative carries ethical and sovereignty criticisms and comes amid growing legal scrutiny of AI training data, creating reputational and regulatory risk even as the scientific upside could be material for biotech and AI collaborations.

Analysis

Expect a structural bifurcation in AI economics as buyers start valuing provenance and labeled, high-quality datasets over raw scale alone. That raises the marginal value of compute that can train models efficiently on curated inputs (favoring high-performance GPUs and memory stacks) while lowering the unit economics of sprawling, low-value web-scraped corpora; the net effect is higher spending on premium infra per useful parameter trained.

A routinized royalty/provenance regime — whether imposed by courts, multilateral agreements, or enterprise procurement — creates a recurring revenue layer for data originators and intermediaries who can prove contribution. That shifts future market value from monolithic model producers to ecosystems that manage consent, traceability, and monetization (data registries, secure sequencing pipelines, provenance middleware), creating attractive targets for consolidation and third-party services.

Operational security incidents tied to autonomous agents will raise the cost of deploying AI at scale inside platforms and enterprises, compressing short-term monetization at consumer-facing incumbents and accelerating spend into security, monitoring, and governance tooling. Meanwhile, geopolitical/ export-control tail risks on advanced accelerators remain the dominant macro swing factor for compute suppliers; a restrictive regime would crystallize supply-driven upside for onshore fabs and design incumbents in 6–24 months.

Key Decisions for Investors

• Pair trade (3-month): Long NVDA / Short META, equal-dollar. Rationale: capture widening spread between premium compute demand and platform-level monetization pressure; target +20% net if NVDA outperforms META by ~25%, stop if the spread moves against you by 8–10%.
• Options hedge (6-month): Buy NVDA 0DTE-6mo call spread to express asymmetric upside to continued enterprise infra spend while capping premium. Risk/reward: pay limited premium (~max loss = premium) with 2–3x upside if NVDA re-rates on stronger data-center demand or an export-control-led supply squeeze.
• Thematic long (12–24 months): Accumulate sequencing/provenance exposure (PACB and selective exposure to ILMN where liquidity allows) on pullbacks. Thesis: recurring data-monetization contracts and integration into model training create 30–70% upside over 12–24 months vs ~20–30% downside if adoption stalls; size as a mid-single-digit portfolio weight.
• Event hedge (6 months): Buy CRWD or ZS 3–6 month calls as asymmetric protection against a spike in enterprise security spend following high-profile AI agent breaches. Expected payoff: 20–40% upside on security repricing vs limited premium outlay.