About XCT Digital
XCT Digital is the AI-compute arm of the XCT family of companies (sister to XCT Energy). Our mission is to create value for the existing data center community by unlocking capacity from infrastructure they already own.
XCT engineering performs thorough site assessments of each facility's electrical network to show the operator exactly what is achievable and how it can be unlocked safely — without inducing risk to the asset. XCT then harvests power from the MEP main switchgear and uses it to power its compute pods. The existing IT loads, their infrastructure, and their cooling systems are not touched.
This enables XCT to monetize — for the data center operator — capacity they didn't underwrite the potential of, in a way that creates zero downside for the operations of the existing asset.
- 0 downside to existing operations
- 0 impact to existing IT or cooling
- 100% operator-aligned economics
The Market Paradox
The AI compute market is supply-constrained on electrical capacity, not silicon. Greenfield data centers face multi-year utility interconnect queues. Meanwhile, most operators carry stranded electrical headroom inside their existing assets — capacity that was underwritten conservatively, derated for harmonics or K-factor, or simply never measured against its real performance envelope.
The inference market is growing faster than training, and inference workloads are far more sensitive to time-to-deploy than to greenfield campus economics. By 2027, the AI workload mix tips decisively toward inference, and the operators who can field compute now, inside existing facilities, win the duration-sensitive revenue.
Technology — ECBS (Electrical Current Balancing System, US Patent 12,375,324 B2)
XCT's patented ECBS recovers stranded electrical headroom inside existing data center power trains:
- K-factor derating reversal — reclaims capacity lost to conservative harmonic derating on existing transformers (IEEE C57.110).
- Harmonic cancellation — actively cancels harmonics rather than oversizing equipment to tolerate them. Holds THDi below 5% at the PCC, inside IEEE 519 limits.
- Transformer capacity recovery — exposes additional usable kVA on installed transformers without replacing them.
- Power conditioning — clean, conditioned power delivered into the pod's GPU rack with no downstream effect on the host's IT loads.
Reference equipment per cluster: 2,500 kVA transformer + 2,500 kW backup genset. The genset references the pod (not the cluster) for backup and redundancy, isolating risk.
Operational reliability (IEEE-grounded)
- 5–15°C winding temperature drop on transformers and motor windings.
- ~2× insulation life per IEEE C57.91 Arrhenius rule (every 10°C drop ≈ doubles insulation life).
- 0.98+ true power factor — lower line current and I²R heating in conductors, switchgear, and upstream transformers (IEEE 141).
Grid
XCT does not draw on new utility interconnect capacity. Power is harvested from the host data center's MEP main switchgear — capacity that is already energized and on site but structurally stranded due to conservative derating, harmonic loading, or under-utilization. This bypasses the multi-year interconnect queue that constrains greenfield expansion.
Of the ~100 GW of existing US data center capacity (~99% leased / occupied), XCT estimates 10+ GW of stranded electrical capacity is recoverable inside the existing footprint.
400 kW GPU Pods
Self-contained 400 kW pods with integrated cooling and power conditioning. Deployed inside existing data center infrastructure in approximately 6 months from contract. Clusters of pods share redundant backup infrastructure, but are designed so a single pod failure cannot affect the host's existing IT load.
Operator Value
For the host data center operator, XCT is structurally upside-only:
- Zero capex — XCT funds, builds, and operates the pods.
- Zero downtime — installation and operation do not touch existing IT loads, infrastructure, or cooling.
- Zero asset risk — engineering assessments demonstrate exactly what can be unlocked safely before anything is built.
- Revenue share on capacity that was previously stranded and unmonetized.
- Faster time-to-revenue than any greenfield expansion or interconnect upgrade the operator could pursue independently.
Worked Example — 4 × 1 MW IT Data Hall (May 2026 sizing)
A representative site with four 2,500 kVA pad-mount transformer lineups, each currently carrying ~950 kW IT peak with ~43% spare capacity under existing harmonic and PUE conditions.
Per-transformer unlock potential — site without XCT Digital vs site with XCT Digital.
| Lineup |
Transformer (kVA) |
Spare Cap |
IT Peak (kW) |
MEP Peak (kW) |
Ave PUE |
Spare Unlock (kVA) |
Mech Unlock (kW) |
Total Unlock (kW) |
Pod IT (kW) |
| P1 | 2,500 | 43.3% | 950 | 454 | 1.43 | 507 | 24.9 | 532 | 400 |
| P2 | 2,500 | 45.4% | 950 | 401 | 1.38 | 560 | 21.9 | 582 | 400 |
| P3 | 2,500 | 43.6% | 950 | 445 | 1.38 | 516 | 29.7 | 545 | 400 |
| P4 | 2,500 | 42.0% | 950 | 486 | 1.41 | 500 | 32.4 | 532 | 400 |
| Total | 10,000 | — | 3,800 | 1,786 | — | 2,083 | 109 | — | 1,600 |
Operator financial impact: Before XCT, 3.8 MW IT yields $6.84M annual revenue and $114M asset value at a 6% cap rate. After XCT, 5.4 MW IT (existing 3.8 MW + 1.6 MW from four XCT pods) yields $8.76M annual revenue and $146M asset value. Net to operator: +$1.92M/yr incremental rent, +$32M asset value uplift, +28.1% IT rent uplift, zero capex. Illustrative; actual results depend on site assessment.
Insights — Supply Chain & Demand Tracker
XCT tracks the structural bottlenecks slowing new data center campuses alongside the published compute demand from neocloud providers that is going unmet. The constraint is power and physical infrastructure, not silicon. Figures reflect published third-party reporting and company disclosures (Wood Mackenzie, IEA, PJM Interconnection, Moody's, Nvidia and the outlets cited) as of early 2026.
Long-lead equipment & campus delays
- Power transformers ~128 weeks; generator step-up units ~144 weeks (some orders to four years); GSU demand up 274% since 2019 — Wood Mackenzie, Q2 2025.
- Medium-voltage switchgear at 40–60+ weeks is the new build bottleneck per EPC teams — POWER Magazine, 2025.
- Transformer prices up 77%+ since 2019, reaching 2.6× pre-pandemic levels in real terms, with lead times nearing four years — pv magazine USA, 2026.
- Project cancellations rose to 25 in 2025 (from 6 in 2024); 188+ local opposition groups across 40 states — Construction Dive, 2025.
- AI sites averaged 7+ years to energize after approval; ~2,300 GW sits in U.S. interconnection queues — Data Center Knowledge, 2025.
- ~30% transformer shortfall modeled across the U.S. fleet in 2025; grain-oriented electrical steel has a single U.S. producer — Wood Mackenzie, 2025.
Published demand going unmet — neocloud backlog tracker
The major GPU-cloud providers hold a combined contracted revenue backlog approaching $100B that the industry cannot serve fast enough. As Nvidia CFO Colette Kress put it, “the clouds are sold out and our GPU installed base is fully utilized.”
- CoreWeave — $66.8B remaining performance obligations; 850 MW active, 3.1 GW contracted.
- Nebius — $20B+ backlog (Microsoft + Meta); ~220 MW active, 2.5 GW contracted.
- Crusoe — 4.5 GW development pipeline anchored by the OpenAI Stargate program.
- Nscale — $23B Microsoft agreement; 200K GB300 GPUs contracted.
XCT's answer: rather than wait 5–7 years for new interconnect and long-lead gear, XCT taps stranded capacity already energized inside existing data centers and fields compute in ~6 months.
Offtake — Compute Customers
XCT contracts the compute output of the pods to hyperscalers and enterprise AI tenants on multi-year offtake structures. The operator does not have to source AI demand themselves; XCT brings the customer.
Clusters
XCT pods are organized into clusters that share redundant backup infrastructure (transformer + genset references), but each pod is electrically and thermally isolated. A single pod failure cannot affect the host's existing IT load or any other pod in the cluster.
Measurement & Verification
XCT operates the XPortal dashboard, which gives the host operator continuous, transparent measurement and verification of pod-level power draw, capacity envelope, demonstrated (lack of) impact on the existing facility, and real-time revenue share accounting. The headroom is managed, not assumed.
Team & Ecosystem
XCT Digital is led by the operating team of XCT Energy, with deep experience in electrical engineering, data center MEP, and large industrial power systems. Engineering, equipment, and offtake partners are listed on the team section of the site.
Frequently Asked Questions
What does XCT Digital do?
XCT Digital deploys 400 kW GPU compute pods inside existing data centers, powered by stranded electrical capacity recovered from the host's existing infrastructure using patented ECBS technology. The host operator earns revenue share on capacity that was previously unmonetized, at zero capex and with zero impact to their existing IT loads or cooling.
How does XCT avoid the multi-year utility interconnect wait?
XCT does not require new interconnect capacity. It harvests power from the host data center's MEP main switchgear — capacity that is already energized and on site but is structurally stranded due to conservative derating, harmonic loading, or under-utilization. Pods are typically deployed within approximately six months.
What is the impact on the existing IT loads and cooling?
Zero. The XCT pod is a self-contained system with its own power conditioning and cooling. Existing IT loads, their distribution infrastructure, and their cooling systems are not touched. Continuous measurement and verification via the XPortal dashboard gives the operator transparent proof of zero impact.
What is ECBS?
ECBS stands for Electrical Current Balancing System. It is XCT's patented technology stack (US Patent 12,375,324 B2, inventor Gregory A. Dockery) that reverses K-factor derating, cancels harmonics (THDi below 5% at the PCC per IEEE 519), corrects power factor, and recovers usable transformer and generator capacity inside existing data center power trains, while continuously monitoring the entire MEP network so loads stay optimized and the host's existing capacity plus its targeted safety headroom are always maintained.