FAQ
Frequently Asked Questions
Everything you need to know about GPU lease rates, datacenter financing, power economics, and AI infrastructure investment.
GPU Economics
What factors affect GPU lease rates?
GPU lease rates vary by GPU type (H100, B200, A100, L40S), region, provider, availability, power consumption, and market demand. Key drivers include the training-inference mix in your workload, contract duration, volume commitments, and whether you need dedicated or shared capacity. Our GLRI index tracks these factors across 20+ providers to help you benchmark current market rates.
How is GPU residual value calculated?
Residual value depends on GPU age, market demand, condition, technological obsolescence, and successor product timelines. For example, H100 residual curves differ significantly from A100 due to their training performance advantage. We track historical trends, secondary market transactions, and regional variations to provide accurate retention estimates for financial modeling.
What is the GLRI and how is it calculated?
The GPU Lease Rate Index (GLRI) is our proprietary benchmark that tracks GPU infrastructure pricing across the market. We aggregate lease rate data from major providers (hyperscalers, GPU clouds, colocation operators) and normalize for contract terms, commitment levels, and included services. The index is updated daily and segmented by GPU model, region, and tier.
Should I lease or buy GPUs?
The lease vs. buy decision depends on your workload predictability, capital structure, depreciation exposure tolerance, and market timing views. Leasing offers flexibility and transfers depreciation risk but typically costs more over long durations. Our Lease vs. Own calculator helps model financial scenarios based on your specific assumptions.
Power & Grid
How do ERCOT curtailment windows affect AI datacenters?
ERCOT manages the Texas electricity grid. During high-demand periods (typically summer afternoons), datacenters may face curtailment requests, reduced power availability, or significantly higher real-time prices. Understanding curtailment patterns—tracked by our CSS index—helps operators structure power contracts, schedule batch workloads, and negotiate appropriate reliability tiers.
What is the Time-to-Power Score (TTPS)?
TTPS quantifies interconnection queue risk by analyzing how long datacenter projects wait for grid connection approval. The score incorporates queue position, ISO backlog trends, transmission upgrade requirements, and historical approval velocities. Higher TTPS scores indicate faster expected energization, which correlates with lower development risk.
What are interconnection queues and why do they matter?
Interconnection queues are waiting lists maintained by grid operators (ERCOT, PJM, MISO, CAISO) for projects seeking to connect to the transmission system. Queue position affects when your datacenter can receive power—delays can extend 3-5+ years in congested regions. Our TTPS index tracks queue metrics across major ISOs to help developers assess site viability.
What is the Curtailment Stress Score (CSS)?
CSS measures grid volatility and curtailment risk for datacenter operations. It incorporates historical curtailment frequency, duration, price volatility, and forecast reliability by zone. Higher CSS scores indicate greater operational risk, which should factor into site selection, power contract structure, and backup power planning.
Colocation & Sites
What is colocation and why does it matter for AI workloads?
Colocation is renting rack space, power, and cooling in a shared datacenter facility. For AI workloads, colo matters because GPU clusters require high power density (40+ kW/rack), liquid cooling infrastructure, and low-latency network connectivity. Not all colo facilities can support AI-scale deployments—our site readiness scores help identify capable facilities.
How do I evaluate datacenter site readiness?
Site readiness encompasses power availability (MW capacity), interconnection timeline, cooling infrastructure (liquid cooling capability), network connectivity (carrier density, IX proximity), and regulatory environment. Our PAY (Power-Adjusted Yield) index synthesizes these factors into a single investment-grade metric for site comparison.
What power density do AI workloads require?
Modern GPU clusters require 40-100+ kW per rack, compared to 10-15 kW for traditional enterprise IT. This high density demands liquid cooling (direct-to-chip or immersion) and purpose-built power distribution. Older colo facilities often cannot retrofit economically, which constrains available supply for AI deployments.
Investment & Finance
How do government incentives affect datacenter ROI?
Federal and state incentives—including tax abatements, ITC for sustainability, economic development grants, and expedited permitting—can significantly impact total cost of ownership, often reducing effective rates by 10-25%. Our analysis tracks available incentives by jurisdiction and models their impact on project economics.
What is the PAY (Power-Adjusted Yield) index?
PAY synthesizes multiple site-level metrics into a single investment-grade score. It incorporates power cost, interconnection risk, curtailment exposure, cooling economics, and regional incentives to produce a risk-adjusted yield estimate. Higher PAY scores indicate more attractive investment opportunities given comparable capital costs.
How should I model GPU depreciation for financial projections?
GPU depreciation follows a curve that depends on technological obsolescence, secondary market liquidity, and successor product timelines. Unlike linear depreciation in accounting, actual residual values often decline faster initially then stabilize. Our residual value calculator uses physics-based decay models calibrated to historical transaction data.
What due diligence is required for datacenter investments?
Infrastructure PE due diligence should cover: power availability and interconnection timeline, land and entitlement status, cooling system design, network connectivity, offtake agreements, operator track record, and regulatory/incentive environment. Our platform provides the market data layer that underwrites these analyses.
Platform & Data
Where does GPU Lease Index source its data?
We aggregate data from public filings (SEC, ISO/RTOs, utility commissions), equipment vendor APIs, published price lists, and direct operator surveys. Each data stream follows defined validation and normalization protocols. We prioritize primary sources and maintain transparent methodology documentation.
How often is data updated?
Update frequency varies by data type: GPU pricing refreshes daily, interconnection queue data updates weekly, power market data streams in near real-time, and residual value curves recalibrate monthly. Major market events trigger interim updates within 2-4 weeks.
Is the platform free to use?
We offer both free and PRO tiers. Free access includes core indices (GLRI overview, TTPS scores), educational playbooks, and basic calculators. PRO unlocks granular data feeds, advanced calculators, real-time alerts, and API access for institutional users.
How can I contact the team?
For general inquiries, reach us through the platform contact form. Enterprise and API access requests are reviewed within 24 hours. We also maintain an active presence on Twitter/X where we share market commentary and index updates.