Qubit Pharmaceuticals validates drug candidates 5x faster with Equinix

Quantum-level simulation without hyperscaler cost surprises

Qubit Pharmaceuticals designs breakthrough drugs by simulating molecules at the quantum level, work that demands millions of GPU-intensive calculations. To get predictable costs and consistent performance for that demand, Qubit partnered with Axians and Equinix to deploy a dedicated 192-GPU cluster inside a carrier-neutral Paris data center.

Rather than screening drug candidates through brute-force laboratory testing, Qubit runs parallel simulations at quantum-level precision on its Atlas platform to identify and refine candidate molecules digitally. Qubit realized an 83% acceleration in target-to-hit time on one lead program versus traditional drug discovery methods, with only 6% of designed compounds ever requiring physical synthesis.

A mission to decode nature’s complexity

Qubit Pharmaceuticals is a pioneer in molecular discovery. Its flagship technology, Atlas, is a cloud native scientific computation platform built to digitally replicate and accelerate the work of a traditional chemistry lab. “Our mission is to find new molecules that can be groundbreaking treatments for diseases that are currently considered incurable with existing methods,” said Romain Delassus, Chief Product and IT Officer at Qubit.

At the center of that mission is a deceptively simple-sounding task: predicting how a candidate molecule will interact with a target inside the human body and with the human body in general. “Atlas is a simulation platform,” Delassus explained. “It lets you design chemical experiments and draft the various stages of an experiment, and then Atlas orchestrates all the computation behind it.” Krystel El Hage, Head of the RNA Platform and Drug Discovery Program Lead at Qubit, relies on Atlas every day to design small molecules that can bind with high affinity to a disease-causing target. “It’s the interface that holds everything together: simulation, data, AI,” she said. 

Because nature is dynamic rather than static, Qubit can’t rely on a single snapshot of a molecule. “We need to analyze what we call a molecular dynamic, a kind of film of what’s going on between a molecule and its target. Those films need a few million frames to provide sufficient information, and each frame requires a quantum-level calculation,” said Delassus.

When the cloud couldn’t keep up

“GPU compute can be found at hyperscalers, but it’s very costly, and hyperscalers are always pushing new technology. We wanted to optimize our algorithms on specific hardware to get the full benefit of it,” said Delassus.

The silent tax of unreliable infrastructure

For Qubit’s scientists, the cost of inconsistent network performance in cloud infrastructure showed up not in a spreadsheet but in lost momentum. “The main infrastructure problems we saw were hardware failures or simulations being cut in the middle just because of a connection issue,” noted El Hage. “Sometimes a simulation cuts short two nanoseconds before the expected end which is something that already took five days.” That lost momentum carried a real financial cost. Because Qubit works with external partners to synthesize the compounds that its simulations identify, a stalled simulation could leave those partners idle.

Building a dedicated, connected compute environment

To solve the problem at its root, Qubit forged an end-to-end partnership with Axians and Equinix. Rather than renting variable hyperscale capacity, Qubit deployed its own dedicated 192-GPU cluster inside a secure, carrier-neutral Equinix data center near Paris. Just as important is how that cluster connects to the rest of Qubit’s technology stack. Atlas itself runs in the cloud, on AWS, while the heaviest simulation workloads run on the dedicated infrastructure at Equinix, with the two environments linked by a private connection rather than the public internet.

Faster, more confident drug discovery

Since deploying its Equinix cluster, Qubit has been able to go from idea to validated result five times faster, according to Delassus, a shift that shows up most concretely in the results of individual drug discovery programs.

“We managed to go from target to hit with an acceleration of 83%, and from hit to lead with an acceleration of 66%, compared to industrial standards,” added El Hage. “We also had a 75% reduction in cost for this program.”

A foundation for partnership and growth

The infrastructure decision has implications well beyond any single research program. Having previously run global HPC operations for Eli Lilly, Andrea De Souza, Chief Corporate Development Officer at Qubit, understands the scale and reliability that industrial drug discovery demands. Drawing on that experience, De Souza sees the Equinix partnership as meeting the same bar. “Our partnership with Equinix provides us with a compute ecosystem that enables Qubit to operate at an industrial level,” she said.

For De Souza, the partnership has a strategic dimension as much as a technical one. With infrastructure risk largely removed from the equation, Qubit’s team can focus on what differentiates the company commercially.

Preparing for the quantum era

Qubit’s ambitions extend beyond today’s GPU clusters. The company is already preparing its software stack, through a product called Hyperion, to run on quantum-native hardware as soon as it becomes commercially viable.

“In five years’ time, real quantum hardware will be here,” said Delassus. “As we did with Equinix, we need the best hardware to provide the most accurate results to our chemists, so we’re already preparing to port our algorithms onto quantum-native hardware. As soon as that hardware is deployable by a company like us, we’ll extend our cluster and host it inside the Equinix data center.”

Benefits

Predictable, high-utilization compute

Qubit’s dedicated 192-GPU cluster runs at extraordinary intensity. Last year, the company orchestrated 1.6 million GPU hours of simulation, an average of roughly 170 GPUs running continuously, around the clock.

Direct, private connectivity to the cloud

A private connection between Qubit’s Equinix-hosted cluster and its AWS environment keeps Atlas’ cloud flexibility intact while giving the heaviest workloads dedicated, predictable performance.

Sovereign, self-operated infrastructure

Because the cluster is Qubit’s own hardware inside a French data center, the company can offer customers full data ownership, local regulatory compliance and IP security.