Nvidia Buys Illumina? The AI Foundry for Healthcare (newsletter.semianalysis.com)

The piece argues that a move like Nvidia acquiring Illumina would be less about an M&A headline and more about creating an “AI foundry for healthcare” that finally lets Moore’s‑Law style compute economics bend Eroom’s Law (rising drug‑development costs). By marrying accelerated compute, cloud platforms and AI with high‑throughput sequencing and omics instrumentation, the industry could unlock scalable clinical genomics, faster diagnostics and more data‑driven drug discovery—shifting bottlenecks from raw sequencing cost to data management, analysis and time‑to‑insight. Technically, the opportunity is concrete: Illumina’s NovaSeq X cuts per‑genome reagent cost to about $200 but still needs 21–48 hours to run; sequencing instruments produced data that grew from ~100 PB (2018) to 280 PB (2021), and genomics could demand ~40 exabytes/year by 2025. Today less than 10% of total genomic program costs are sequencing itself—>90% is sample collection, storage, reduction and secondary analysis—so accelerated GPUs, optimized pipelines, and LLMs for ingesting unstructured EHRs (EHR adoption rose from ~10% to 95% in 2008–2014; global healthcare data reached ~2,314 EB by 2020) are the real leverage points. Combined with breakthroughs like AlphaFold (<0.1 nm accuracy in protein folding), integrating genomics, proteomics and epigenomics on scalable compute stacks could compress time‑to‑market, enable personalized medicine at scale, and materially counteract Eroom’s Law.