Evaluating
Crusoe Spark?

Understand what you are comparing

Crusoe’s October 21, 2026 announcement describes Spark Liquid as a modular AI factory for Crusoe Cloud. It specifies closed-loop direct liquid cooling, up to 3.5 MW of compute per unit, and planned production in 2027. These are Crusoe’s published specifications, not QIT specifications.

QIT builds liquid-cooled GPU containers that install on a prepared site in 15 hours. Our configuration and delivery scope are defined for the individual project. A useful comparison starts with the infrastructure and operating model you actually need.

Source: Crusoe Spark Liquid announcement, published October 21, 2026. Reviewed October 30, 2026. Product plans may change.

Spark, Spark Liquid, Edge Zones and DFM

Spark is Crusoe’s prefabricated modular AI infrastructure platform. Its June 2025 launch described integrated power, cooling, monitoring, fire suppression and GPU racks, with delivery as fast as three months.

Spark Liquid is the liquid-cooled generation described above. Edge Zones is the cloud deployment offering built around Spark: Crusoe describes managing the site, power integration and operations while providing access to Crusoe Cloud. See Crusoe Edge Zones.

Digital Flare Mitigation refers to Crusoe’s earlier oilfield-gas computing business. In March 2025, Crusoe announced its planned sale to NYDIG as part of the bitcoin mining operation. It should not be treated as another name for Spark Liquid or today’s GPU cloud offering.

Compare the published facts

Different scope boundaries can produce misleading headline comparisons. A cloud capacity offering, a delivered enclosure and a fully commissioned customer site are different purchase decisions.

Questions to put to both suppliers

Start by describing the same workload, location, available power and target operating date. Then request a response against the same boundaries.

• What exactly will you own, lease or consume as a service?
• Who procures and installs the compute equipment?
• Which electrical, cooling and network systems sit outside the quoted scope?
• What event starts the delivery clock, and which dependencies can move it?
• How is full-system acceptance defined and documented?
• Who operates the infrastructure and responds to equipment faults?
• Which expansion, support and replacement costs should enter the comparison?

A clear answer to these questions is more useful than comparing a module’s headline capacity with another project’s target completion time.

When to speak with QIT

Speak with QIT when you are scoping a physical GPU container, have a site or power opportunity to evaluate, and want liquid-cooled GPU capacity installed in 15 hours.

We begin with the inputs needed to define the configuration. You can review our GPU shipping container guide, the proposed delivery stages and the cooling questions before starting a conversation.

QIT is independent of Crusoe and NVIDIA. This comparison does not claim a partnership, certification, equivalent capacity or superior measured performance.

What the Snyder agreement illustrates

Energy Vault’s announced Snyder project separates infrastructure ownership from cloud operation through a purchase and leaseback of Spark units. That specific agreement shows why “cloud operator” and “module owner” are not always the same party; it does not establish general retail availability of Spark hardware.

Source: Energy Vault, July 27, 2026.

QIT’s assessment is that the useful comparison starts with the handover boundary: site infrastructure, compute hardware and operations. Read powered land versus powered shell before comparing a project proposal with a cloud capacity offer.