AI for Life Sciences: From Compute to Clinical Impact

IgniteGTM's AI for Life Sciences event brought together infrastructure providers, cloud operators, enterprise AI leaders, investors, and diagnostics founders around a shared premise: life sciences is becoming one of AI's most consequential frontiers. The day moved from the physical foundations of AI compute to deploying AI responsibly in regulated organizations, then to the investment landscape and founders building AI-enabled diagnostics at the point of care.

Infrastructure Is A Systems Problem

The opening infrastructure discussion positioned GPUs as necessary but insufficient. Life sciences workloads depend on the whole system: moving large datasets quickly, sustaining high-bandwidth communication among nodes, and storing massive volumes of images, genomic data, and signals. Supermicro described four interconnected zones, data, HPC, AI, and platform, with networking tying them into a single operating system for scientific work. The message: infrastructure must match how scientists actually work, with AI agents amplifying scientists rather than replacing their judgment.

Biology Is Becoming Computable

AMD argued the next AI demand wave will come from "large quantitative models" representing physical and biological systems: protein folding, genotype-to-phenotype relationships, molecular chemistry, and physics.

Unlike language workloads that tolerate FP4/FP8 precision, these often require FP64. AMD described a hardware split, the Mi 455x for FP4/FP8 and agentic workloads, the Mi 430x for FP64 science, and its Helios rack-scale system (72 GPUs, ~23 TB/s interconnect). It also stressed open-source: provide the engine, let partners build on top.

AMD also emphasized its open-source strategy. Its goal is to provide the “engine” while allowing customers and ecosystem partners to build on top of it. The company described day-zero support for major open models and argued that coding agents are reducing software-porting friction: code optimized for a competing GPU platform can increasingly be translated into ROCm-compatible code in minutes.

Context is the enterprise challenge

A panel on enterprise adoption argued that the hardest issue is not the availability of compute or models. It is context: who the user is, what data they can access, and what safety limits apply.

A context layer encoding identity, provenance, and safeguards is essential infrastructure. The group distinguished data integration (accessibility) from interoperability (meaning), noting large organizations now face deferred data-strategy debt. Prompt engineering is giving way to context engineering.

Investors look beyond hype.

Capital is seeking defensible, regulated-ready opportunities: wearables and passive data, AI infrastructure for regulated environments, diagnostics with clinical value, genomic-phenotypic data fusion, and patient data sovereignty. In healthcare, hallucination and omission are clinical and regulatory risks, not just product flaws. Panelists anticipated a shift from reactive "sick care" toward prevention and longevity.

Diagnostics is where AI meets the patient

The closing panel featured robotic 3D ultrasound for breast imaging, AI-assisted cystoscopy for bladder cancer and misfolded-protein biomarkers for neurodegenerative disease. Each kept a clinician in the loop with auditable records. The reinforced lesson: governance is a design requirement, not an afterthought. Healthcare AI must be reproducible, clinically valid, auditable, and genuinely beneficial to patient care.

Core Takeaway

AI for life sciences is becoming real through convergence, compute scaling, models maturing, deployment getting practical, and diagnostics translating AI into earlier detection and better clinical decisions.

The path forward is connecting trustworthy data, fit-for-purpose infrastructure, rigorous validation, and human expertise.

Bay Area Startups Collectively Secured $13B+ in September MTD

September funding is running ahead of what was a busy August – but then, September 2025 was a record-breaking $34B. Two open-source companies closed rounds – Temporal ($550M) and Arcee AI ($150M). Arcee is focused on open-weight frontier models and their funding puts them in a position to capitalize on the growing enterprise need to control AI costs.

Exits, M&A down: Q3 M&A is down in the U.S.- 61% off by deal value and 46% down on deal count. SV M&A is close to the deal value dropoff - 63% down, $128B in Q2 and just $48B so far in Q3 - but deal numbers have stayed up, 130 in Q2 and 105 in Q3.

Startups: Looking for partners, keeping a watch for possible competitors? LinkSV's Pulse of the Valley weekday newsletter and alerts can keep you on top of new companies launching; database keyword search gives you prospects for partners and customers. Founders get the newsletter, and alerts for just $7/month ($50 value). Check it out and sign up here.

Follow LinkSV on LinkedIn to get early notice of new tech, stay on top of SV funding intelligence, and keep track of the companies, investors and executives impacting the startup ecosystem.

Early Stage:

• TAR closed a $120M Series A, builds energy infrastructure in the United States.

• Delos Data closed a $100M Series A, building faster, stronger and more efficient inference, starting from how compute, acceleration, and data is distributed and scaled.

• Rune Energy closed a $40M Series A, turns solar plants into data centers.

• Expanse closed a $5.3M Seed, delivers compute certainty for AI infrastructure, predicting exactly what every AI workload needs before it runs

• VerifAIX closed a $5M Seed, an AI-powered verification for the next generation of bug-free chips.

Growth Stage:

• Temporal closed a $550M Series E,develops and distributes the world's leading open source durable execution system.

• Instinct Company closed a $250M Series B, a personal assistant that understands what you're working on and what's important to you.

• Factory closed a $200M Series D, building software that keeps developers in control of the important high-level details while agents handle the coding.

• Arcee AI closed a $150M Series B, builds frontier open-weight foundation models.

• Planted Solar closed a $31.8M Series B, the autonomous power deployment company.

Accelerating OpenFold3 with AMD and Vultr

AI is opening new doors in drug discovery, but researchers still have to get increasingly complex models into production before they can use them.

OpenFold3 can predict the 3D structures of proteins, antibodies, DNA, RNA, protein complexes, and small molecule ligands. Running those workloads means dealing with GPUs, Kubernetes resources, networking, secure access, and model serving.

Researchers want to spend their time studying biology, not assembling the infrastructure underneath it.

That is what AMD and Vultr are working to simplify.

What they built
AMD Inference Microservices for OpenFold3 is available through the Vultr Kubernetes Engine Marketplace, running on AMD Instinct MI325X GPUs. The preconfigured environment handles much of the Kubernetes deployment, networking, HTTPS configuration, and model serving required to get OpenFold3 running.

MindWalk independently deployed and validated the stack, demonstrating a path for running demanding protein structure prediction workflows on standardized, Kubernetes native infrastructure.

Why this, why now?
Protein structure models are getting more capable at modeling the complex interactions researchers care about in drug discovery, antibody research, protein engineering, and disease research.

Making those models easier to deploy shortens the distance between access to powerful AI infrastructure and actually using it for scientific work.

Explore the full OpenFold3 use case from Vultr to see how the stack was deployed and validated.

Your Feedback Matters!

Your feedback is crucial in helping us refine our content and maintain the newsletter's value for you and your fellow readers. We welcome your suggestions on how we can improve our offering. [email protected] 

Logan Lemery
Head of Content // Team Ignite

Some teams never seem to stop moving. They're on Attio, the agentic CRM.

It’s your always-on revenue engine: agents and workflows build pipeline, chase every buying signal, and move deals forward alongside your team.

Teams like Parallel, Turbopuffer, and Wordsmith build on Attio. Are you one of them?

Try Attio now