Torus funding, valuation and investors

Product & use cases

Illustrative buyer journey: (1) a GPU/cloud operator or large business with an industrial facility asks Torus to assess transformer spare headroom, interconnection constraints, fiber, land and local utility contract; (2) operator evaluates electricity tariffs, on/off-peak demand, curtailment, backup and power-quality requirements with Torus Reserve flywheels and batteries; (3) chooses Strategic shell (no GPUs), Partner turnkey bare metal (GPUs included) or Direct single-tenant inference; (4) Torus prefabricates enclosure, cooling, power, network and optional GPUs at Utah GigaOne, performs site prep then electrical commissioning under local permitting/utility rules; (5) workloads run with Torus-managed operations, security, load shaping, maintenance, customer monitoring and cost accounting. Vendor cites 12–16 weeks contract-to-live for suitable sites, not a universal independently benchmarked SLA. Key contract terms and published pricing are unavailable. The output is operating compute capacity, not a trained AI model. This is a documentation-based scenario, not hands-on deployment.

• AI cloud operators obtain smaller inference capacity blocks near usable existing industrial power connections instead of commissioning a new greenfield data center.
• Industrial energy sites combine battery/flywheel demand shaping and local grid services with optional hosted GPU compute.
• Enterprises procure single-tenant inference resources with power/cooling/cybersecurity operations from one supplier, subject to provider security and utility review.

Key facts

• November 11, 2026: company announces $500M growth capital led by Magnetar; financing structure and valuation undisclosed. Independent DCD describes Series C.
• Official Foundry launch: 1–20 MW inference sites using existing transformer capacity with Torus Reserve storage; company targets 12–16 weeks from signed contract to operational use.
• Three offerings: powered shell without GPUs, bare-metal units with GPUs, and single-tenant enterprise inference; specific prices and contracts undisclosed.
• Issuer identifies Hydra Host and another AI cloud as early customers, says thousands of GPUs running; aims for more than 47,000 by end-2027 (future goal).
• 2024 $67M package combined equity, note conversions and loan facility; 2025 $200M investment form unspecified. Company $767M cumulative includes differing capital types, not verified pure equity.

Competitive landscape

Edge: Integrates energy control (flywheel and battery), prefabricated power and AI computing with site commissioning and operations. Potential advantage over greenfield builds is time to available utility power; needs verification for each transformer, utility tariff and host-site interconnection. Compared with independent modular enclosure vendors, Torus bundles power-shaping and O&M. Company claims are not independently measured superiority.

Foundry is a manufactured infrastructure and energy-services model, not a pure GPU rental software platform. Key economics require contracted electricity tariff, local transformer availability, uptime/SLA, GPU capex, depreciation, financing terms, cooling, fiber, local permitting, site host agreement, utilization and O&M. Claims that a site never needs new grid infrastructure or affects neighbors require utility diligence. Manufacturer says thousands of GPUs deployed and 500MW IT manufacturing line capability, but neither independently proves profitable deployment or completed 500MW of installed Foundry capacity. First-party source originally corrected 2025 accounts of some customer 'adoptions' to 'designing', so distinguish pilots/plans from operating deployments. $500M new capital and $767M cumulative do not establish all-equity financing nor recurring revenue.

• Crusoe adjacent

  AI compute and powered-data-center provider; often develops large energy-sited infrastructure and GPU cloud, versus Torus's prefabricated sites at existing transformer headroom.

• Compass Datacenters alternative

  Develops large custom data-center campuses for hyperscale and enterprise needs; different project scale and utility lead times.

• Vertiv alternative

  Provides integrated power, thermal, modular data-center infrastructure and services, but not identical to Torus's bundled GPU and grid-flexibility operations.

• Existing colocation and GPU clouds incumbent

  Can procure currently available GPU capacity without building a Foundry site; trade-offs depend on location, tariff, SLA and control requirements.

Market / IPO context