Nvidia CEO, Jensen Huang, asserts that accelerated computing has a missing word: application acceleration. The “vertically integrated” and “horizontally open” chip maker is set on building the infrastructure that delivers AI into real world use.
“Accelerated computing is not a chip problem,” said Huang when he took the stage for his annual NVIDIA GTC keynote in San Jose in March. “The only way for us to accelerate applications and bring tremendous speed up and cost reduction is through domain specific acceleration.”
That mission has hit drug discovery, where approval timelines exceed a decade and clinical trial failure rates approach 90%.
A new wave of platforms from Amazon Web Services (AWS), OpenAI, and Anthropic have customized general-purpose assistants into AI-powered workflows for science research. The trend points to the growing role of cloud infrastructure and agentic AI in unifying fragmented tools, streamlining data management, and making domain expertise more accessible.
Lab-in-the-loop
In April, AWS introduced Amazon Bio Discovery, a “lab-in-the-loop” workflow that combines access to more than 40 open-source and proprietary biological foundation models with AI agents that guide experimental design. The platform also integrates CRO partners, including Twist Bioscience, Ginkgo Bioworks, and A-Alpha Bio, for lab validation. The launch was announced at the AWS Life Sciences Symposium at the Javits Center in New York.
In collaboration with Memorial Sloan Kettering Cancer Center, Amazon Bio Discovery designed nanobodies with nanomolar affinities by generating nearly 300,000 candidates that were narrowed to the top 100,000 for wet lab testing in weeks, a noticeable reduction from the up to one year timeline typical of traditional methods.
Dan Sheeran, vice president and general manager, healthcare and life science at AWS, explains that while biological AI models have driven breakthroughs in areas, like protein design, their reliance on coding expertise and complex compute infrastructure remains a significant barrier to broader accessibility.
“Choosing the right model for a given task is itself a significant challenge. Computational biologists, the specialists who bridge AI and biology, are in short supply,” Sheeran told GEN Edge. “The result is a collaboration bottleneck, not because the science isn’t available, but because the tooling doesn’t support how these teams need to work together.”
David Younger, PhD, co-founder and CEO of A-Alpha Bio, adds that the partnership with AWS highlights a “fundamental gap” in AI-powered drug discovery, the lack of high-quality, experimental data at scale to evaluate protein design models. In silico candidates designed using Amazon Bio Discovery can be rapidly validated in the lab with A-Alpha’s AlphaSeq platform, which quantitatively measures protein-protein interactions by the hundreds to millions.
“The convergence of technology and life sciences isn’t just about faster compute or better algorithms,” Younger told GEN Edge. “It’s about connecting those advances to real-world, experimental observations.”
Amazon Bio Discovery is built on the same AWS infrastructure that is currently adopted by 19 of the top 20 global pharmaceutical companies. Each organization’s data is isolated within its application environment, and all proprietary data, models, and designs remain customer-owned.
Rosalind reasons
Two days after Amazon Bio Discovery’s launch, OpenAI announced GPT-Rosalind, a specialized reasoning model that supports evidence synthesis, hypothesis generation, and experimental planning for research across biology, drug discovery, and translational medicine. The platform includes a freely accessible life sciences research plugin for Codex that connects to over 50 public multiomics databases, literature repositories, and computational biology tools.
The model is available through a trusted-access program for qualified enterprise customers in the U.S. Amgen, Moderna, the Allen Institute, and Thermo Fisher Scientific are among GPT‑Rosalind’s customers.
“Research organizations are actively looking for systems that are built for scientific workflows, not adapted from general-purpose models, and life sciences remains one of the most important areas where better tools could meaningfully accelerate progress,” wrote OpenAI in an email to GEN Edge when describing the motivation for building GPT-Rosalind.
Named after Rosalind Franklin, PhD, whose work was critical in the discovery of the DNA double helix, the model scored 0.751 on BixBench, a benchmark that evaluates large language model (LLM) performance in bioinformatics and computational biology tasks. The score was a modest lead ahead of GPT-5.4, xAI’s Grok 4.2, and Google’s Gemini 3.1 Pro.
On LABBench2, a benchmark spanning literature retrieval, database access, sequence manipulation, and protocol design, GPT-Rosalind outperformed GPT-5.4 on six out of 11 tasks. The largest improvement was shown on CloningQA, which requires end-to-end design of DNA constructs and enzyme reagents for molecular cloning workflows.
GPT-Rosalind is one step in OpenAI’s growing momentum across pharma and healthcare. In recent weeks, the company introduced ChatGPT for Clinicians to support clinical workflows, such as documentation and medical research, alongside partnerships with Novo Nordisk to enhance workforce AI readiness and improve manufacturing and supply chain efficiency, and Massive Bio to expand access to clinical trials.
Inference inflection
Anthropic is forging its own path into life sciences, having recently drawn attention for acquiring Coefficient Bio, a roughly 10-person AI drug discovery start-up founded by former Genentech scientists, for $400 million.
The OpenAI competitor has also been building Claude for Life Sciences, the AI assistant specialized for researchers, clinical coordinators, and regulatory affairs managers, since last fall.
In an October blog post, Anthropic reported that the customized platform powered by Claude Sonnet 4.5 scored 0.83 in Protocol QA, a benchmark that tests the model’s understanding of laboratory protocols. The score outperformed the human baseline of 0.79 and Sonnet 4’s performance of 0.74. Claude for Life Sciences also incorporates several connectors to scientific platforms, including Benchling’s digital notebooks, PubMed literature, and 10x Genomics tools for single cell and spatial analysis.
“We want to give scientists the same experience as software engineers of having a brainstorming partner to work with and to delegate tasks,” said Eric Kauderer-Abrams, PhD, head of biology and life sciences at Anthropic, in a video accompanying the product launch.
In January, Anthropic expanded the platform to Claude for Healthcare, a complementary set of tools that allow healthcare providers, payers, and health tech companies and startups to use Claude for medical purposes through HIPAA-ready products.
When reflecting on these life science releases, Enke Bashllari, PhD, founder and managing director at Arkitekt Ventures, says the three are “playing different games.” OpenAI is selling the “sharpest reasoning engine” with limited access, while AWS is building infrastructure and lab integration. Anthropic is betting on breadth of workflow and making acquisitions to close the specialization gap.
“For startups, the question isn’t which platform wins. It’s which layer you build on,” wrote Bashllari on LinkedIn.
Chris Leiter, founder and general partner at Atria Ventures, believes the shift to bioconsumerism will be the “most significant period of disruption for life sciences in the modern era.”
“Medicine is the use case that justifies the entire buildout,” wrote Leiter on LinkedIn. “The public skepticism starts to erode when the output is a drug that reaches a patient five years early, or a diagnostic that catches a cancer no doctor would have seen.”
As models increasingly move beyond isolated predictions into complex reasoning across biological systems, the question is no longer whether to adopt, but how quickly the industry can adapt to a new scientific discovery paradigm.
Huang says it best, “we are now in the beginning of a new platform shift. The inference inflection has arrived.”
The post Big Tech Targets Drug Discovery with Wave of Life Science Platforms appeared first on GEN – Genetic Engineering and Biotechnology News.
