Lattice Graph × Shell

Lattice Graph is a computational materials-discovery platform built around a knowledge graph that spans millions of compositions. Every candidate material that enters the pipeline is validated by multiple independent machine-learning interatomic potentials — MACE, CHGNet, MatterSim, and ORB — and must reach consensus on both phonon and thermodynamic stability before it advances. That multi-potential consensus requirement is not a formality: it filters out the large fraction of computationally attractive candidates that fail in practice, so the leads that reach a partner's bench have already survived a demanding in silico gauntlet. First-principles DFT calculations then sharpen the property predictions on the survivors, and targeted simulations resolve application-specific questions around dielectric retention, thermal conductivity, and electrochemical compatibility. Beyond stability screening, Lattice Graph runs composition- and claim-level screening across more than 300,000 materials patents, giving partners a clear picture of the intellectual-property landscape before committing resources to a candidate. The platform also carries a large atlas of labeled negative results — failed experiments documented and indexed — so that the same dead-ends are not re-entered from a different angle. The combination of rigorous multi-potential stability gating, freedom-to-operate intelligence, and structured failure data collapses the discovery timeline from years to months and substantially de-risks the decision to take a new fluid or material into formulation development.

Shell's move into data-center immersion-cooling fluids — including its collaboration with Keppel — puts the company squarely in one of the fastest-moving segments of the specialty-fluids market. The withdrawal of 3M Novec and the broader regulatory pressure on fluorinated chemistries have created a gap that the market is racing to fill: the immersion-cooling segment is tracking from roughly two billion dollars today toward eleven billion, and the companies that arrive with a qualified, spec-anchored PFAS-free product first will capture the formulation and supply relationships that prove durable. Shell has the global distribution, the lubricants and fluids infrastructure, and the customer relationships in energy and data-center operations to win that race — but only if the underlying chemistry is defensible and the product portfolio is broad enough to cover both single-phase and two-phase use cases. That is where Lattice Graph fits. Shell's differentiation in immersion fluids cannot rest on reformulating incumbent chemistries; it has to be built on a compositional foundation that is genuinely novel, demonstrably stable, and clear of the crowded patent thicket that formed around the fluorinated-fluid era. Our knowledge graph and computational screening pipeline were purpose-built for exactly this kind of challenge: identify PFAS-free candidates with the right dielectric constant, flash-point safety margin, and thermal performance window, confirm their stability under realistic cycling conditions, and map the freedom-to-operate landscape before a single liter is synthesized. The result is a package of validated leads and process assets that Shell can bring into its existing fluids development workflow rather than building computational infrastructure from scratch.

Shell's move into data-center immersion-cooling fluids — including its collaboration with Keppel — puts the company squarely in one of the fastest-moving segments of the specialty-fluids market. The withdrawal of 3M Novec and the broader regulatory pressure on fluorinated chemistries have created a gap that the market is racing to fill: the immersion-cooling segment is tracking from roughly two billion dollars today toward eleven billion, and the companies that arrive with a qualified, spec-anchored PFAS-free product first will capture the formulation and supply relationships that prove durable. Shell has the global distribution, the lubricants and fluids infrastructure, and the customer relationships in energy and data-center operations to win that race — but only if the underlying chemistry is defensible and the product portfolio is broad enough to cover both single-phase and two-phase use cases. That is where Lattice Graph fits. Shell's differentiation in immersion fluids cannot rest on reformulating incumbent chemistries; it has to be built on a compositional foundation that is genuinely novel, demonstrably stable, and clear of the crowded patent thicket that formed around the fluorinated-fluid era. Our knowledge graph and computational screening pipeline were purpose-built for exactly this kind of challenge: identify PFAS-free candidates with the right dielectric constant, flash-point safety margin, and thermal performance window, confirm their stability under realistic cycling conditions, and map the freedom-to-operate landscape before a single liter is synthesized. The result is a package of validated leads and process assets that Shell can bring into its existing fluids development workflow rather than building computational infrastructure from scratch.

The primary portfolio relevant to Shell is PFAS-free dielectric and process fluids. This portfolio was developed in direct response to the fluorinated-fluid market dislocation and encompasses the full stack from bulk fluid candidates through purification and release processes to closed-loop reuse systems — exactly the range Shell needs to take a new immersion-cooling product from candidate selection to a commercially qualified product. The leads within this portfolio have passed multi-potential stability consensus and carry documented performance specs covering dielectric retention, corrosion-inhibitor compatibility, and reuse-cycle durability, giving Shell a credible technical data package to share with data-center operators and hyperscaler procurement teams. A secondary but relevant area is the high-power thermal-interface materials portfolio. As Shell's immersion-fluid line expands, adjacent thermal-management products — gap fillers, phase-change materials, conformable pads — become natural line extensions or bundle components for data-center customers that want a single fluids-and-thermal supplier. The computational methods used to screen fluid candidates translate directly to thermal-interface discovery: the same stability and conductivity calculations that qualify a bulk coolant also benchmark a candidate thermal-interface compound. Shell can use this adjacency to deepen customer relationships beyond the coolant tank itself.

The Lattice Graph platform surface that Shell's fluids and technology teams would use day to day centers on candidate evaluation and portfolio management. Researchers can submit a composition or a class of compositions and receive a stability assessment grounded in multi-potential consensus — the same MACE, CHGNet, MatterSim, and ORB agreement requirement that gates internal discovery — along with computed properties relevant to dielectric performance and thermal stability. That workflow integrates with Shell's existing stage-gate process by delivering a standardized computational data package at the candidate-selection stage rather than after bench synthesis. The freedom-to-operate screening tool is available as both an interactive application and a programmatic interface, so Shell's IP and R&D teams can run landscape analyses as part of normal portfolio reviews. Patent coverage can be queried at the composition level or the process-claim level, and results are returned with the claim-level citations needed to support legal assessment. The negative-results database is similarly queryable, allowing Shell's formulators to cross-check a candidate fluid against documented failure modes before committing to a development track.

A typical engagement with Shell would begin with a focused scoping session in which Shell's fluids development and IP teams define the target performance envelope — dielectric constant range, boiling point or flash-point constraints, viscosity window, and materials compatibility requirements for the server hardware Shell is targeting. Lattice Graph would then run a knowledge-graph query and computational screening pass across the PFAS-free dielectric and process fluids portfolio, delivering a ranked candidate shortlist with stability certificates, property predictions, and preliminary freedom-to-operate assessments within four to six weeks. That package gives Shell's team a prioritized bench synthesis list with the computational due diligence already completed. From there, engagement can expand into co-development — Lattice Graph providing on-demand computational support as Shell's lab results come in, updating the knowledge graph with Shell's proprietary experimental data under a data-use agreement, and running targeted simulations on formulation variants as the product moves toward qualification. Licensing of one or more portfolio assets, including the immersion-cooling system specification and the purification platform, can be structured in parallel. Pricing is calibrated to the scope of computational access, the number of portfolio assets licensed, and the degree of exclusivity Shell requires in the immersion-cooling segment — a conversation that makes sense to open now, before competitive licensing discussions advance.