Designing the Best AI Infrastructure for Drug Discovery

Drug discovery is a foundational part of business for a pharmaceutical company, one that can unlock new revenue streams and safeguard an organization’s future. However, finding and developing new drugs has traditionally been a costly, time-consuming process involving a lot of regulatory approvals. According to Deloitte, in 2024, research and development (R&D) for a new drug cost an average of $2.23 billion USD.[1] It takes twelve to fourteen years to develop a new drug, so reducing this timeline by even one year could save $159 million per drug. Over a major pharmaceutical pipeline, savings could reach greater than $15 billion per year.

While AI isn’t new to the pharmaceutical sector, it now promises to make drug discovery faster and more cost-effective than ever. It’s enabling companies to identify novel proteins and genes to target for treatment, generating new drug molecules, and streamlining manufacturing processes. And we’ve only begun to tap into its potential.

To take advantage of the AI opportunity, pharmaceutical companies need the right AI infrastructure in the right places. AI workloads for drug discovery involve huge volumes of data and require high-performance computing (HPC) and low latency data transfer. That means pharmaceutical companies need to be strategic about choosing infrastructure that meets their AI requirements, now and in the future.

Rising to meet the challenge of the pharma patent cliff

A patent cliff is a drop in revenue that occurs when patent protection for a product expires. In the pharmaceutical industry, when the exclusive rights to a drug expire, other manufacturers can produce and sell versions of the medicine more cheaply, leading to a fast drop in sales. Since the biggest revenue opportunity for a drug manufacturer is during the period when they have exclusivity, anything that speeds up the drug discovery timeline can mitigate revenue risk.

It’s predicted that there’s more than $200 billion USD in annual revenue at risk between now and 2030 due to the next patent cliff in pharmaceuticals.[2] If companies don’t produce new drugs faster, the upcoming patent expirations will lead to falling revenue.

AI and machine learning offer enormous promise in the face of this challenge. By speeding up innovation and drastically cutting down drug discovery timelines, AI can help pharmaceutical companies maintain a competitive edge.

Strategic considerations for drug discovery workloads

Organizations in the life sciences and healthcare industry are investing heavily in AI and supporting technologies as they look to leverage a treasure trove of health and genomic data. But implementing the right technologies can be complex. From initial ideation to full realization of a new drug on the market, pharmaceutical companies need to be strategic about deploying infrastructure that optimizes drug discovery workloads for cost, latency, ecosystem access and connectivity. It’s not just about benchmarking GPU performance; you need to think through the whole AI solution, from query to outcome.

Here are 4 important areas to think about when planning AI infrastructure for drug discovery:

1. Who are your users?

Research scientists are the primary end users of AI applications for drug discovery. Pharmaceutical researchers need fast, seamless access to data and AI applications. That means infrastructure needs to be within reach of where the research is taking place.

2. Where’s your data coming from, and is it subject to data residency requirements?

The data used in drug discovery includes genetic data, health data and data from large repositories such as the UK Biobank or U.S. National Institutes of Health (NIH). Drug discovery involves a large volume of data, most of which is sensitive information. Access to all this data enhances accuracy for drug discovery applications, but your infrastructure needs to be close to where data is generated to ensure low latency for AI applications. In addition, a lot of health data is subject to regulations requiring data to stay in the country where it’s generated.

3. Who are your partners and service providers?

Pharmaceutical companies work with a variety of partners and service providers, from technology providers to biobanks, research universities and contract research organizations (CROs). Pharma companies need to exchange data with these partners quickly and securely to advance drug discovery initiatives.

4. Where are AI applications and infrastructure hosted?

Pharmaceutical companies may choose to host their AI applications in the cloud, in on-premises infrastructure or in a colocation facility. Where they place compute can directly affect costs and application performance, and thus impact researcher experience. Global companies typically have R&D hubs in multiple regions, so they also need to ensure connectivity between locations.

Hosting AI applications in the public cloud can get expensive, but many drug discovery platforms are cloud-native, so having access to software in the cloud is a must. On-premises infrastructure can be challenging to scale for AI and often isn’t designed to accommodate the advanced power and cooling capabilities AI workloads require. Colocation facilities, on the other hand, offer AI-ready data centers, as well as access to clouds and service providers in a neutral environment.

Taking AI-driven drug discovery to the next level

As pharma companies build out their AI strategy for drug discovery and prepare to move AI initiatives from ideation to production, several approaches can improve outcomes:

• Optimize your use of cloud: Using the cloud strategically, as part of a hybrid infrastructure model for AI, can reduce overall cloud spend while still allowing you to take advantage of born-on-cloud drug discovery applications. Meanwhile, you can use private infrastructure for data storage and data protection.
• Position yourself at the heart of research hubs and healthcare ecosystems: Research and collaboration are vital in the pharmaceutical business. To get the most from those relationships, you need to be physically close to research hubs such as Boston-Cambridge in the U.S.; Zurich, Switzerland; and Tokyo. You also need to be close to your research partners, data providers and other key players.
• Leverage AI-ready data centers: Why not take advantage of data centers already designed for AI workloads, especially if your on-prem data centers aren’t? Having the right power and cooling capabilities for HPC is a game changer, driving greater efficiency, performance and cost effectiveness.
• Connect with clouds and technology service providers in a vendor-neutral place: No company can do AI alone. In addition to clouds, network service providers and other IT services play a huge role in any AI solution, so it’s important to make ecosystem access a priority from the start.

Why consider drug discovery at Equinix?

Equinix has the capabilities and ecosystem to drive innovation for pharmaceutical companies. You can easily connect with thousands of clouds, networks, data and model providers, and other IT services at Equinix. Our AI-ready data centers are designed to help you optimize drug discovery applications for low latency and high performance. And our global footprint allows you to put infrastructure where you need it to meet your R&D and data residency requirements.

Equinix has helped numerous companies in the pharmaceutical industry innovate faster and operate more efficiently. That’s why Singapore-based Nanyang Biologics chose Equinix as the home for their Drug-Target Interaction Graph Neural Network (DTIGN) for AI-driven drug discovery. Using Equinix’s HPC environment led to a 68% acceleration in drug discovery and 90% reduction in R&D costs, while providing a gateway to a global AI ecosystem. They’re now working with HPE and Equinix to offer an as-a-Service proprietary AI-driven platform for drug discovery.

Learn more about the advantages of deploying infrastructure for drug discovery at Equinix by downloading our solution brief.

[1] Be brave, be bold: Measuring the return for pharmaceutical innovation, Deloitte, March 25, 2025.

[2] Big pharma’s looming threat: a patent cliff of ‘tectonic magnitude’, BioPharma Dive, February 21, 2023.