Frequently Asked Questions

As chip power density increases, many systems are reaching a point where junction temperature, not compute capability, limits sustained performance.

This typically shows up during long-duration workloads, where thermal mass is no longer sufficient and total thermal resistance from junction to coolant dominates. Even aggressive airflow or colder coolant often delivers diminishing returns because the primary bottleneck sits inside the package and cold-plate stack.

When optimization efforts stop yielding meaningful improvements to the junction, the thermal stack itself has become a structural constraint. At that point, maintaining acceptable junction temperatures requires reducing thermal resistance at the source, not just tuning system-level parameters.

For a deeper discussion of why this happens and how to recognize it early, read our blog article.

A 20–40% reduction in total thermal resistance can materially change system design options by lowering junction temperature at the same power, enabling higher sustainable power at fixed junction limits, and increasing margin under real inlet conditions.

The impact is greatest when dominant resistances sit close to the die and incremental airflow or coolant improvements have reached diminishing returns. In that regime, resistance reduction changes system behavior, not just efficiency.

For a deeper discussion, read our blog article.

Thermal throttling is often the first visible sign that total thermal resistance is governing sustained performance in AI and HPC systems.

When performance loss correlates with junction temperature and incremental airflow or coolant changes deliver diminishing returns, the dominant constraint is typically near-die thermal resistance rather than software or power delivery.

For a deeper discussion, read our blog article.

At high power density, dominant thermal bottlenecks often sit near the die rather than at the rack or facility level. This creates system-level constraints such as throttling, inlet sensitivity, derating, and escalating fan and pump power. Liquid metal cooling reduces total thermal resistance close to the heat source, restoring margin where air and water cooling deliver diminishing returns.

The relevant comparison for any thermal engineer is total junction-to-coolant thermal resistance, not individual component efficiency. Standard water-glycol cold plates are limited by a resistive thermal boundary layer caused by water’s low conductivity, while vapor chambers hit a “spreading limit” at high heat flux.

Pumped liquid-metal loops leverage 100x higher thermal conductivity to efficiently draw heat into the bulk fluid and provide active heat transport. This reduces the dominant thermal resistance at the socket, restoring system-level margin and allowing for higher power densities that traditional passive or water-based stacks cannot support.

For a deeper discussion, read our blog article.

Other solutions have limits. Single-phase water has heat-flux and power limits due to the potential for uncontrolled phase change. Vapor chambers have a related problem called dry out caused by high heat fluxes. Liquid metal has none of these problems. High boiling point and high vapor pressure = pure, single-phase cooling with no flux limits or dry-out.

While conventional thermal management alternatives face rigid physical thresholds, liquid metal architectures expand the operational envelope by entirely eliminating the critical failure points of legacy systems. Traditional single-phase water blocks operate under strict heat flux and power limits due to the risk of triggering an uncontrolled phase-change cycle. Similarly, standard vapor chambers suffer from localized dry-out under intense processing loads. Liquid metal experiences none of these physical constraints. Because of its extraordinarily high boiling point, this technology guarantees continuous single-phase cooling with no heat-flux limits or operational dry-out risks.

For a deeper discussion, read our blog article.

Two-phase cooling provides very high heat transfer but introduces additional system complexity and operational considerations.

Lowering total thermal resistance can reduce junction-to-coolant temperature delta enough to support higher sustained loads within a single-phase architecture, delaying or reducing the need to transition to two-phase cooling.

For a deeper discussion, read our blog article.

Inlet temperature sensitivity is largely governed by total thermal resistance. When resistance is high, small changes in inlet air or coolant temperature produce large junction temperature changes.

Reducing thermal resistance lowers this amplification, allowing systems to tolerate warmer inlet conditions and improving operational robustness.

Moving to liquid metal is more than a fluid swap; it’s about enabling architectural scaling that traditional cooling can no longer support. By treating reclaimed thermal headroom as a resource, engineers can trade lower junction temperatures for higher performance, increased density, or lower facility costs. This shift transforms cooling from a high-maintenance utility into a permanent, maintenance-free asset that outlasts the silicon it protects.

The primary constraint when integrating liquid metal is material compatibility: aluminum must be strictly avoided as it undergoes rapid embrittlement. For long-term reliability, copper components should be nickel-plated to prevent intermetallic formation, while stainless steel and titanium remain excellent choices for structural loop components.

Mechanically, designers must account for the fluid’s high density, roughly 6.5x that of water, and ensure proper surface wetting to eliminate contact resistance. Despite these constraints, the fluid’s water-like viscosity and low operating pressure allow it to fit within standard mechanical envelopes while delivering 100x the thermal conductivity of traditional coolants.

Liquid metal is a “Thermal Bridge,” not a total facility replacement. It integrates into air-cooled systems by replacing vapor chambers with high-efficiency pumped loops (LAAC), and into water-cooled facilities by serving as a high-conductivity link between the die and the standard facility water loop.

Because the fluid’s viscosity is nearly identical to PG25, it fits within existing mechanical envelopes without requiring a redesign of flow-control logic. This allows engineers to address the socket-level heat-flux crisis without changing their broader infrastructure.

Yes. Liquid metal loops and MHD pumps are designed for the 1U/2U and OCP mechanical envelopes. By replacing standard OCP vapor chambers with our LM cooling loops, we’ve demonstrated a 35% reduction in system thermal resistance without changing the server footprint.

The fluid’s high conductivity also enables more compact cold plate designs, ensuring compatibility with the most space-constrained server designs.

Scaling liquid metal systems relies on a modular, hybrid architecture that keeps the fluid localized to the high-heat-flux sockets, where its 100-fold conductivity is most effective. By using a sealed liquid-metal loop at the node level and a secondary loop (air or water) for long-distance transport, we avoid the weight and volume issues associated with a rack-wide manifold. Our strategy utilizes both high-performance mechanical pumps and solid-state MHD pumps to ensure reliable circulation, supported by an industrial supply chain capable of delivering consistent performance for mass-produced, rack-scale deployments.

Long-term reliability is engineered through a strict ‘Zero-Aluminum’ material policy, utilizing nickel-plated copper or stainless steel interfaces. Nickel serves as a permanent diffusion barrier, preventing the formation of brittle intermetallics and ensuring structural integrity over a 20-year lifecycle. Mechanically, the system utilizes MHD (Magnetohydrodynamic) pumps, which have no moving parts, seals, or bearings. By eliminating these common failure points, the system transitions from a high-maintenance consumable to a permanent, ‘fit-and-forget’ infrastructure asset.

Liquid metal loops are factory-filled and hermetically sealed, making them maintenance-free throughout their lifecycles. A clean-fill vacuum-purge process during manufacturing prevents bulk oxide formation before the loop is sealed. The fluid is inorganic and non-volatile — it does not degrade, evaporate, or support biological growth.

Systems use either high-performance, long-life mechanical pumps or solid-state MHD pumps with zero moving parts. Both designs eliminate the need for chemical monitoring or filter changes, making the result a set-and-forget cooling solution with the reliability profile of a solid-state hardware component.

While two-phase cooling offers high heat transfer, it introduces significant operational complexity due to the non-linear nature of phase-change physics. Managing vapor quality and pressure regimes requires sophisticated control logic to prevent dry-out conditions and handle transient loads.

Liquid-metal cooling achieves comparable high-flux heat removal while maintaining a stable, single-phase architecture. By providing linear and predictable thermal resistance, liquid metal eliminates the containment risks and control challenges associated with boiling. For many AI and HPC applications, this “collapses” the thermal bottleneck and restores performance margin without the mechanical overhead of a two-phase system.

The gallium supply chain is stabilized by strategic reshoring efforts in Australia and the U.S., which are on track to diversify 25% of global production. Unlike consumable coolants, liquid metal is a permanent, 20-year asset that is 100% recyclable, shielding operators from price volatility and regional supply constraints.

Our systems are engineered to leverage the high thermal performance of gallium alloys while ensuring total chemical and mechanical stability. Rather than treating material reactivity as an unknown risk, we have standardized a high-durability material stack designed for the 20-year lifecycle of a data center.

By utilizing nickel-plated copper and high-grade stainless steel interfaces, we create a permanent diffusion barrier that prevents intermetallic formation. This disciplined approach to material selection, combined with rigorous accelerated life testing and thermal cycling validation, ensures that our cooling loops remain a stable, maintenance-free asset within the compute stack.

Yes, the liquid metal-based alloys used are non-toxic, non-volatile, and environmentally friendly. With a boiling point above 2,000°C, there is zero risk of vapor inhalation. It does have some restrictions on how it is packaged for transport specifically due to its reactivity with aluminum, it is safe for human handling with standard PPE and does not require the hazardous waste disposal protocols typical of treated facility water or refrigerants.

The electrical risk posed by a liquid metal leak is managed using the same “defense-in-depth” principles as in standard water cooling. Our systems are factory-sealed hardware assets that operate at low pressure, minimizing seal stress. While the fluid is conductive, its high surface tension causes it to “bead” into discrete droplets rather than “wicking” or spreading like water-glycol. This unique behavior makes the fluid easier to contain and clean, significantly reducing the risk of permanent hardware damage compared to traditional liquid-cooling leaks.

While liquid metal carries a higher material cost than water, its system-level economics are superior when managing high-heat-flux chips. By delivering 4.5x higher heat removal efficiency, LM allows engineers to avoid massive facility infrastructure overhauls and delay the transition to costly two-phase cooling.

Furthermore, the use of MHD pumps, which have zero moving parts, reduces long-term maintenance OPEX. In high-density compute environments, the “cost” of the cooling system is best measured by the thermal headroom it creates, allowing chips to run at sustained peak performance without throttling.

Liquid metal is a sustainable, enduring asset that does not degrade, requires no toxic biocides, and is 100% recyclable at the end of the hardware’s life. Unlike glycol-based coolants or PFAS-heavy refrigerants, Liquid-metal is non-toxic and non-volatile. Its superior heat-removal efficiency also reduces total data center energy consumption and physical footprint, supporting a lower PUE and a circular material economy.