Frequently Asked Questions

Our standard hosting agreements start at a 12-month minimum term for colocation services, which allows us to allocate dedicated power and space at the most competitive rates. For managed hosting and larger deployments (1 MW+), we offer flexible term structures ranging from 12 to 60 months.

Shorter-term arrangements may be available for evaluation purposes or seasonal mining operations, typically at a modest premium. Contact our sales team to discuss your specific requirements and we will structure a commitment that aligns with your operational timeline.

Deployment timelines depend on the scale of your installation and the readiness of our target facility:

• Small deployments (1-50 units): Typically 5-10 business days from contract signing, assuming available rack space and power.
• Medium deployments (50-500 units): 2-4 weeks, including custom rack configuration, power provisioning, and network setup.
• Large deployments (500+ units / 1 MW+): 4-8 weeks for full infrastructure buildout, including dedicated power circuits and cooling.

We maintain pre-provisioned capacity at several facilities to accelerate onboarding. If speed is critical, let us know upfront and we will prioritize available hot-standby space.

Colocation means you own the hardware and we provide the facility: power, cooling, physical security, and network connectivity. You retain full control over your equipment, including remote management access. This is ideal for operators who have in-house technical teams and want maximum flexibility.

Managed hosting adds an operations layer on top of colocation. We handle hardware monitoring, firmware updates, troubleshooting, reboots, and replacements. You focus on strategy while our on-site technicians handle the day-to-day. Managed hosting is popular with investors and organizations that want exposure to mining or compute without the operational overhead.

Both options include our standard SLA for power uptime and physical security. The key difference is who handles the hands-on equipment management.

Our response depends on your service level:

• Colocation clients: Our monitoring systems detect offline units and alert you immediately. You can ship replacement parts or authorize our team to perform basic troubleshooting (power cycle, cable reseat) at no extra charge. Board-level repair requires either your own technician on-site or our managed repair service.
• Managed hosting clients: Our technicians diagnose the issue and attempt repair. If a hashboard, PSU, or control board replacement is needed, we source parts from our inventory (or yours, if pre-shipped) and complete the repair on-site. Typical turnaround for common parts is 24-48 hours.

We maintain a parts inventory for popular ASIC models (Bitmain S-series, MicroBT M-series) to minimize downtime. For GPU servers, we stock common components like RAM, SSDs, and power supplies. Replacement GPUs are handled on a case-by-case basis.

Absolutely. The majority of our colocation clients ship their own hardware directly to our facilities. Here is how it works:

• Coordinate in advance: Notify us at least 48 hours before shipment with tracking numbers, expected unit count, and any special handling instructions.
• Shipping address: We provide a dedicated receiving address and dock hours for each facility. Most sites accept deliveries Monday through Friday, 8 AM to 5 PM local time.
• Receiving and racking: Our team inspects shipments for visible damage, inventories serial numbers, and racks equipment per your deployment plan. Standard racking is included in your hosting fee.

We also partner with logistics providers for bulk shipments and international imports. If you are purchasing new equipment, we can often arrange direct-from-manufacturer delivery to our facility to save time and shipping costs.

Yes. Through our partner network, we help clients source ASIC miners, GPU servers, and networking equipment at competitive prices. Our procurement services include:

• New equipment: We have direct relationships with major manufacturers including Bitmain, MicroBT, and leading GPU server OEMs. Bulk orders often qualify for volume discounts.
• Certified pre-owned: We maintain a marketplace of tested, certified used equipment for budget-conscious deployments. All pre-owned units are benchmarked and carry a 90-day warranty.
• Turnkey packages: Hardware + hosting bundled into a single monthly payment, ideal for investors entering mining or compute for the first time.

We handle import logistics, customs clearance for international orders, and can arrange financing for large purchases. Contact our sales team for current inventory and pricing.

Every hosting client gets access to our monitoring dashboard, which provides real-time visibility into your deployment:

• Unit status: Online/offline status, hashrate (for miners), temperature, and power consumption per device.
• Power metrics: Total kW consumption, daily/monthly power costs, and PUE tracking for your allocated space.
• Alerts: Configurable notifications via email or webhook for unit failures, temperature exceedances, power anomalies, and scheduled maintenance.
• Historical data: 90-day rolling data for trend analysis, performance benchmarking, and SLA verification.

Managed hosting clients receive additional weekly performance reports and proactive recommendations for optimizing their fleet. API access is available for programmatic integration with your own monitoring tools.

Yes, and we design our agreements to accommodate growth. Expansion options include:

• Same-site scaling: If your current facility has available power and space, we can provision additional capacity within 1-2 weeks.
• Multi-site deployment: Distribute your fleet across multiple locations for redundancy, geographic diversity, or to access different power rates.
• Power upgrades: Move from standard colocation to dedicated transformer circuits for MW-scale deployments.

We proactively track your utilization and reach out when you are approaching capacity limits. Most clients start with a modest deployment and scale up within 3-6 months as they validate performance and ROI.

Our standard hosting SLA guarantees 99.5% power uptime measured on a monthly basis, excluding scheduled maintenance windows (announced 72 hours in advance). This translates to a maximum of approximately 3.6 hours of unplanned downtime per month.

For mission-critical deployments, we offer enhanced SLAs up to 99.9% uptime with dedicated redundant power feeds, on-site generator backup, and priority incident response. Enhanced SLAs carry a premium but include financial credits for any breach.

Our historical uptime across all facilities exceeds 99.7%, with most outages attributable to utility grid events rather than facility infrastructure.

Yes. While ASIC miners and GPU servers are our primary focus, our facilities are equipped to host a wide range of compute hardware:

• Standard servers: 1U-4U rack-mount servers for web hosting, databases, and general compute workloads.
• Storage nodes: High-density storage servers for Filecoin, Chia, or enterprise backup/archive solutions.
• Network equipment: Routers, switches, and firewalls for clients building out private network infrastructure.
• Specialized hardware: FPGA arrays, custom ASICs for non-Bitcoin applications, and research equipment.

If your hardware has unusual power, cooling, or environmental requirements, reach out to our engineering team for a custom assessment.

Getting started is straightforward:

• Step 1 - Consultation: Book a free 20-minute call with our team to discuss your goals, budget, and timeline.
• Step 2 - Hardware selection: Choose from our inventory of new and certified pre-owned ASICs, or ship your own equipment.
• Step 3 - Facility selection: We recommend the best location based on your priorities (lowest power cost, proximity, climate, etc.).
• Step 4 - Deployment: Sign the hosting agreement, and we rack, connect, and configure your miners. Most deployments are hashing within days of equipment arrival.

We also offer turnkey mining packages that bundle hardware, hosting, and management into a single monthly fee -- perfect for first-time miners who want a hands-off experience.

We support all major SHA-256 ASIC miners currently in production, including:

• Bitmain Antminer: S21, S21 Pro, S21 XP, S21 Hydro, S19 XP, S19j Pro+, and legacy S19/S17 models.
• MicroBT Whatsminer: M60S, M60, M56S, M53S, M50S, M50, and legacy M30S++ models.
• Canaan AvalonMiner: A1566, A1466, and compatible models.

Our infrastructure accommodates both air-cooled and liquid/immersion-cooled units. If you are considering a newer or less common ASIC model, contact us to confirm compatibility with our power distribution and cooling systems. We continuously update our infrastructure to support the latest generation hardware.

Mining profitability depends on several key variables:

• Hashrate: Your miner's computational power (measured in TH/s).
• Power consumption: How many watts your miner draws at the wall.
• Electricity cost: Your all-in rate per kWh (including hosting fees).
• Network difficulty: The current Bitcoin mining difficulty, which adjusts approximately every two weeks.
• Bitcoin price: The market value of BTC at the time of mining.
• Pool fees: Typically 1-2% of mining rewards.

Use our Mining Profit Calculator in the Tools section for a detailed projection. As a rule of thumb, electricity typically represents 60-80% of operational costs. This is why our competitive power rates (starting at $0.055/kWh) are a significant advantage for long-term profitability.

Remember that profitability is dynamic -- Bitcoin price appreciation and difficulty adjustments can dramatically change the equation over time.

We do not mandate a specific mining pool -- you are free to point your hashrate wherever you choose. That said, popular options among our clients include:

• Foundry USA: Largest US-based pool, strong compliance, OFAC-compliant.
• Braiins Pool (formerly Slush Pool): Oldest pool, transparent scoring, excellent developer tools.
• ViaBTC: Global pool with competitive fees and multiple payout methods.
• Ocean: Non-custodial pool focused on decentralization and miner sovereignty.

For managed hosting clients, we can configure pool settings during deployment. For colocation clients, pool configuration is entirely under your control via your miner's web interface or firmware. We recommend choosing a pool based on fees, payout frequency, geographic proximity, and your stance on mining centralization.

Yes. For clients who want Bitcoin mining exposure without purchasing hardware, we offer:

• Turnkey mining packages: We provide the hardware, hosting, and management. You pay a flat monthly rate and receive the mining output minus operational costs.
• Hashrate leasing: Rent a portion of our hosted fleet's hashrate on monthly or quarterly terms. Useful for testing mining economics before committing to hardware ownership.

These options are particularly popular with institutional investors and funds that want Bitcoin production exposure with predictable operational expenses. All arrangements are fully transparent with real-time dashboard access to your allocated hashrate and earnings.

Power efficiency is measured in joules per terahash (J/TH) and is the single most important metric for mining profitability. Current-generation models:

• Bitmain S21 XP Hydro: ~12 J/TH (liquid-cooled, 270 TH/s)
• Bitmain S21 Pro: ~15 J/TH (air-cooled, 234 TH/s)
• Bitmain S21: ~17.5 J/TH (air-cooled, 200 TH/s)
• MicroBT M60S: ~18.5 J/TH (air-cooled, 186 TH/s)
• Previous gen S19 XP: ~21.5 J/TH (air-cooled, 140 TH/s)

Lower J/TH means less electricity per unit of mining output. The jump from S19 to S21 generation represented a roughly 20-30% efficiency improvement. Combined with our competitive power rates, current-generation hardware can remain profitable even during market downturns.

Yes, with some guidelines:

• Underclocking (downclocking): Fully supported and encouraged during periods of low profitability. Reducing hashrate by 10-20% can improve efficiency (J/TH) by 15-25%, extending profitable operation. Many of our managed clients use dynamic tuning based on Bitcoin price and difficulty.
• Overclocking: Permitted as long as your total power draw stays within your allocated circuit capacity. Overclocking increases heat output and may require enhanced cooling arrangements. You are responsible for any hardware damage resulting from overclocking.

For managed hosting clients, our team can implement firmware-level auto-tuning that dynamically adjusts hashrate based on profitability thresholds you define. This maximizes uptime and efficiency without manual intervention.

Our facilities are purpose-built for high-density compute, so noise and heat are managed by design:

• Noise: ASIC miners are housed in dedicated halls isolated from office and common areas. Sound levels in mining halls typically range from 80-95 dB -- all personnel in these areas use hearing protection per OSHA standards.
• Heat: Our cooling systems are engineered for the thermal loads of dense mining deployments. Air-cooled facilities use high-volume exhaust fans, evaporative cooling, and free-air cooling (in cooler climates). Immersion facilities use dielectric fluid with heat exchangers.

You do not need to worry about noise or heat management -- that is our responsibility. Our facilities maintain ambient intake temperatures between 60-80 degrees F year-round to ensure optimal equipment performance and longevity.

Yes. AI/ML infrastructure is a core service line at Rax. We offer:

• GPU colocation: Bring your own GPU servers (NVIDIA A100, H100, L40S, or AMD MI300X) and we provide power, cooling, and high-bandwidth networking.
• Managed GPU clusters: We provision and manage multi-node GPU clusters for training runs, inference serving, and fine-tuning workloads. Includes NCCL optimization, InfiniBand configuration, and job scheduling.
• Reserved capacity: Guaranteed GPU availability for organizations with ongoing training pipelines. Avoid the spot-instance lottery.

Our AI/HPC facilities feature enhanced cooling, low-latency InfiniBand networking, and power density up to 50 kW per rack -- purpose-built for the thermal and electrical demands of modern GPU hardware.

We support and can procure the following GPU platforms:

• NVIDIA H100 SXM5: The gold standard for large-scale training. 80 GB HBM3, 3.35 TB/s memory bandwidth, NVLink 4.0 for multi-GPU scaling.
• NVIDIA H200: Next-generation with 141 GB HBM3e for larger model context windows and reduced memory bottlenecks.
• NVIDIA A100: Proven workhorse for training and inference. Available in 40 GB and 80 GB configurations.
• NVIDIA L40S: Cost-effective for inference, fine-tuning, and mixed AI/graphics workloads. 48 GB GDDR6X.
• AMD MI300X: Competitive alternative with 192 GB HBM3 -- excellent for memory-bound workloads and large language models.

GPU technology evolves rapidly. Contact us for the latest availability and pricing -- we continuously onboard new platforms as they reach production maturity.

Absolutely. Multi-node GPU training is one of our specialties. Our infrastructure supports:

• InfiniBand HDR/NDR: 200-400 Gbps low-latency interconnects for NCCL-optimized distributed training across hundreds of GPUs.
• NVIDIA DGX / HGX configurations: 8-GPU nodes with NVSwitch for intra-node communication, scaled across nodes via InfiniBand fabric.
• Shared storage: High-performance parallel file systems (Lustre, GPFS, or NFS over RDMA) for training data and checkpoints.
• Cluster management: SLURM or Kubernetes-based job scheduling, with our team handling infrastructure while you focus on model development.

We have experience supporting training runs on clusters from 8 to 512+ GPUs. Our network architecture is designed to maintain linear scaling efficiency even at large node counts.

Yes. Inference has different requirements than training, and we optimize accordingly:

• Dedicated inference nodes: Right-sized GPU servers (L40S, A100, or H100 depending on model size) with low-latency networking and guaranteed availability.
• Load-balanced clusters: Multi-node inference with automatic scaling, health checks, and failover for production AI applications.
• Edge-optimized: Geographically distributed inference endpoints for latency-sensitive applications.

We support common serving frameworks including vLLM, TensorRT-LLM, Triton Inference Server, and custom solutions. Our team can assist with optimizing throughput, latency, and cost-per-token for your specific model and traffic patterns.

AI infrastructure pricing at Rax is built on transparency and flexibility:

• Colocation (BYO hardware): Priced per kW of allocated power, similar to our mining hosting but with enhanced networking and cooling included. Rates start from $0.07-0.10/kWh depending on location and commitment term.
• Managed GPU clusters: Priced per GPU-hour or as a monthly reserved rate. Includes hardware, power, cooling, networking, and management. Significantly cheaper than major cloud providers for sustained workloads.
• Hybrid models: Burst to our infrastructure during training peaks while maintaining a smaller reserved footprint for steady-state inference.

For large-scale commitments (100+ GPUs), we offer custom pricing with substantial volume discounts. Our goal is to provide cloud-like convenience at colocation-level economics.

The primary advantages of Rax over AWS, Azure, GCP, or CoreWeave for AI workloads:

• Cost: 40-60% lower total cost for sustained workloads. Cloud is optimized for variable demand; if your GPUs run 24/7, dedicated infrastructure is dramatically more economical.
• Availability: No spot instance interruptions, no capacity shortages. Reserved means reserved -- your GPUs are always there.
• Performance: Bare-metal access eliminates virtualization overhead. Direct InfiniBand connectivity without software-defined networking bottlenecks.
• Data sovereignty: Your data stays on your hardware in a known physical location. No multi-tenant risks or cloud provider data policies to navigate.

The tradeoff is less elasticity -- scaling up takes days instead of minutes. For organizations with predictable GPU demand (which describes most serious AI operations), the economics overwhelmingly favor dedicated infrastructure.

Yes. Our GPU cluster management service covers the full operational stack:

• Hardware management: Monitoring, thermal management, component replacement, and firmware updates across all nodes.
• Networking: InfiniBand fabric configuration, topology optimization, NCCL tuning, and network health monitoring.
• Software stack: CUDA drivers, container runtimes (Docker, Singularity), SLURM/Kubernetes orchestration, and storage configuration.
• Job scheduling: Multi-tenant scheduling with fair-share policies, priority queues, and preemption controls.

Our managed service is designed for organizations that want to focus on model development and research rather than infrastructure operations. We handle everything below the application layer so your data scientists and ML engineers can focus on what they do best.

GPU servers have significantly higher power density than traditional servers:

• Single GPU workstation (1-2 GPUs): 500-1,500 W
• 4-GPU server (e.g., Lambda, Supermicro): 2,000-4,000 W
• 8-GPU server (e.g., DGX A100/H100): 6,500-10,200 W
• Full rack of 8-GPU nodes: 40-80+ kW per rack

Our AI/HPC facilities are engineered for these power densities with dedicated high-amperage circuits, PDUs rated for 40-60+ kW per rack, and cooling systems matched to the thermal output. Standard enterprise data centers typically max out at 8-15 kW per rack, which is insufficient for modern GPU deployments.

When planning your deployment, provide us with your hardware specifications and we will confirm power allocation and cooling capacity at your target facility.

Yes. While AI/ML is our primary GPU use case, our infrastructure supports any GPU-accelerated workload:

• 3D rendering: Blender, V-Ray, OctaneRender, and other GPU renderers for film, architecture, and product visualization.
• Scientific visualization: ParaView, VMD, and domain-specific visualization tools for research data.
• Video transcoding: NVENC-based encoding pipelines for media companies processing large video libraries.
• Simulation: Computational fluid dynamics, molecular dynamics, and physics simulation on GPU clusters.

The NVIDIA L40S is particularly well-suited for mixed rendering and AI workloads. For pure rendering, we can also provide NVIDIA RTX 6000 Ada configurations. Contact us with your specific application requirements.

High-performance storage is critical for GPU workloads. We offer:

• Local NVMe: Each GPU server is configured with high-speed NVMe SSDs for low-latency data access during training and inference. Typical configurations: 2-8 TB per node.
• Shared parallel storage: Lustre or BeeGFS clusters providing hundreds of GB/s aggregate throughput for large training datasets. Essential for multi-node training where every GPU needs fast access to the same data.
• Object storage: S3-compatible storage for checkpoints, model artifacts, and long-term data retention. Lower cost per TB for data that does not need real-time access.

We size storage based on your workload profile. Training runs with large datasets (multi-TB) typically need parallel shared storage, while inference workloads often perform well with local NVMe alone.

Yes. We offer Kubernetes-based GPU orchestration as a managed service:

• NVIDIA GPU Operator: Automated driver management, GPU device plugin, and monitoring for Kubernetes clusters.
• Multi-instance GPU (MIG): Partition A100/H100 GPUs into smaller instances for inference or development workloads, maximizing utilization.
• Fractional GPU scheduling: Share GPUs across multiple pods for lightweight workloads that do not require a full GPU.
• Persistent volume claims: Integrated with our storage infrastructure for seamless data access across pods and nodes.

For teams already using Kubernetes for AI/ML (with Kubeflow, Ray, or custom operators), we provide a familiar interface with bare-metal performance. We handle the cluster lifecycle, upgrades, and infrastructure, while you manage your workloads.

Rax operates and partners with data center facilities across the United States, organized into three strategic regions:

• Northwest: Washington, Oregon, Montana, Idaho -- leveraging abundant hydroelectric power and cool climates for naturally low cooling costs.
• Southwest: Texas, Oklahoma, New Mexico, Nevada, Arizona, Utah, Colorado -- accessing competitive natural gas and solar energy resources.
• Midwest: Ohio, Indiana, Kentucky, North Dakota, South Dakota, Iowa, Nebraska, Kansas, Missouri, Wisconsin, Minnesota, Illinois, Michigan, Wyoming -- benefiting from diverse energy markets and strategic grid positions.

Each region offers distinct advantages in terms of power cost, climate, regulatory environment, and grid reliability. Visit our Data Center Facilities page for detailed information about each location, including available power, current rates, and facility specifications.

Yes, we welcome facility tours for prospective and existing clients. Here is how to arrange a visit:

• Schedule in advance: Contact your account manager or book through our website at least one week before your desired visit date.
• Identification required: All visitors must present valid government-issued photo ID. Non-US visitors should bring their passport.
• Escort policy: All visitors are escorted by Rax personnel at all times within the facility for security and safety reasons.
• Photography: Limited photography is permitted in common areas and your own equipment space. Photography of other clients' equipment or security systems is not allowed.

Tours typically take 60-90 minutes and include the operations center, power distribution, cooling systems, and the hosting floor. We recommend visiting the specific facility where you plan to deploy for the most relevant experience.

We deploy multiple cooling strategies depending on climate, equipment type, and density:

• Air cooling: High-volume exhaust fans with evaporative pre-cooling. Our standard for ASIC mining in moderate climates. Cost-effective and proven at scale.
• Free-air cooling: In cooler climates (Northwest, Northern Midwest), outside air provides direct cooling for 6-9 months per year, dramatically reducing cooling energy costs.
• Liquid cooling (rear-door heat exchangers): For high-density GPU racks exceeding 30 kW. Chilled water loops with rear-door or in-row cooling units.
• Immersion cooling: Single-phase and two-phase immersion solutions for maximum density and efficiency. Ideal for overclocked ASICs and ultra-dense GPU deployments. PUE as low as 1.02-1.05.

Our engineering team recommends the optimal cooling solution based on your hardware, density requirements, and the target facility's climate. Immersion cooling carries higher upfront costs but delivers the lowest long-term operating expenses and best equipment longevity.

Our facilities maintain industry-standard certifications and comply with relevant regulations:

• SOC 2 Type II: Certified facilities with audited controls for security, availability, and confidentiality.
• OSHA compliance: All facilities meet Occupational Safety and Health Administration standards for workplace safety.
• Local electrical and building codes: All power infrastructure inspected and certified by local authorities having jurisdiction (AHJ).
• Environmental compliance: Facilities comply with EPA regulations for generator emissions, coolant handling, and waste disposal.

For clients with specific compliance requirements (HIPAA, PCI-DSS, FedRAMP), we can discuss which facilities and configurations meet your needs. Not all locations carry the same certifications, so compliance-sensitive deployments should be discussed during the planning phase.

Redundancy varies by facility tier, but our standard infrastructure includes:

• Power: Dual utility feeds where available, on-site diesel generators with automatic transfer switches (ATS), and N+1 UPS systems for graceful failover. Generator fuel reserves for 24-72 hours of sustained operation.
• Cooling: N+1 cooling redundancy ensures that a single cooling unit failure does not impact operations. Critical facilities have independent cooling loops.
• Networking: Redundant uplinks from multiple ISPs with automatic failover. Core switching is fully redundant with sub-second convergence.
• Fire suppression: Pre-action dry pipe sprinklers, VESDA early warning smoke detection, and clean agent (FM-200 or Novec) suppression in critical areas.

For mission-critical deployments, we offer Tier III-equivalent configurations with 2N power redundancy and maintenance-concurrent capabilities. Discuss your specific uptime requirements during planning.

Power Usage Effectiveness (PUE) measures how efficiently a data center uses energy. A PUE of 1.0 means all power goes to computing; higher numbers mean more energy is spent on cooling and overhead. Our PUE varies by facility and cooling method:

• Air-cooled mining facilities: PUE of 1.10-1.25, depending on climate and season. Free-air cooling in northern climates achieves the lower end.
• Liquid-cooled GPU facilities: PUE of 1.08-1.15. Rear-door heat exchangers are highly efficient at removing concentrated heat loads.
• Immersion-cooled facilities: PUE of 1.02-1.05. The best efficiency available, as immersion eliminates fans entirely and captures nearly all waste heat.

For context, the global data center industry average PUE is approximately 1.55. Our facilities consistently operate well below this benchmark, which translates directly to lower energy costs for our clients.

Our electricity rates are among the most competitive in the industry, starting at $0.055/kWh for large-scale deployments in our most cost-effective locations. Rates vary by:

• Location: Each facility accesses different utility rates and energy markets. Northwest hydro locations tend to be cheapest; Southwest solar-rich locations offer competitive blended rates.
• Commitment term: Longer terms (36-60 months) unlock lower rates. 12-month contracts are available at standard pricing.
• Scale: MW-level deployments receive volume discounts. The more power you commit to, the better your per-kWh rate.
• Service level: Basic colocation carries the lowest rate. Managed hosting includes a management premium. AI/HPC colocation includes networking and enhanced cooling in the rate.

All rates are fully loaded (include facility overhead, cooling, and maintenance). There are no hidden fees for power delivery or distribution. Request a quote for your specific deployment scenario.

We position ourselves as best-value hosting -- not necessarily the cheapest headline rate, but the best total cost of ownership:

• vs. Major cloud (AWS/Azure/GCP): 40-60% lower for sustained GPU workloads. Cloud charges $2-3+/GPU-hour for H100s; our managed GPU hosting runs $1.00-1.50/GPU-hour equivalent.
• vs. Budget mining hosts: Comparable or slightly higher rates, but with significantly better uptime, monitoring, and support. Cheap hosting with 90% uptime costs more than premium hosting with 99.5%+ uptime when you factor in lost revenue.
• vs. Enterprise colocation (Equinix, Digital Realty): 30-50% lower for high-density compute. Enterprise colo is optimized for low-density IT, not the 40-80 kW/rack densities of mining and AI.

Our competitive advantage comes from purpose-built facilities, strategic location in low-cost energy markets, and operational efficiency from specializing in high-density compute rather than general-purpose IT hosting.

Yes. We partner with equipment financing providers to offer flexible payment options:

• Equipment leasing: 12-36 month leases with fixed monthly payments. At the end of the term, you can purchase the equipment at fair market value, return it, or renew the lease with upgraded hardware.
• Deferred payment: For large orders (100+ ASICs or $500K+ GPU clusters), we can arrange deferred start payments where hosting begins immediately and hardware payments start 30-60 days later.
• Revenue-sharing: For mining deployments, some arrangements allow hardware costs to be repaid from mining proceeds, reducing upfront capital requirements.

Financing is subject to credit approval and varies by order size. We aim to remove capital barriers so organizations of all sizes can access industrial-grade compute infrastructure.

We accept a range of payment methods to accommodate different client preferences:

• Wire transfer (ACH/SWIFT): Our preferred method for monthly hosting invoices and large equipment purchases. No processing fees.
• Credit card: Accepted for smaller invoices and initial deposits. A 3% processing surcharge applies.
• Bitcoin (BTC): We accept Bitcoin for hosting payments and equipment purchases. Invoices are denominated in USD with BTC conversion at time of payment.
• Stablecoin (USDC/USDT): Accepted for clients who prefer cryptocurrency payments with price stability.

Invoices are issued monthly in arrears for hosting services and net-30 for equipment purchases. We offer a 2% discount for annual prepayment of hosting fees.

We pride ourselves on transparent pricing with no surprises:

• Standard setup fee: A one-time setup fee covers initial racking, cabling, network configuration, and testing. For deployments under 50 units, this is typically $25-50 per device. Larger deployments often have setup fees waived or reduced.
• No hidden power markups: Your quoted $/kWh rate is fully loaded. There is no separate charge for power distribution, transformer usage, or cooling.
• No bandwidth overage fees: Standard monitoring and management traffic is included. Dedicated high-bandwidth connections (10 Gbps+) are quoted separately.
• No early termination penalties: If you need to reduce or exit, we require 60 days written notice but do not charge punitive cancellation fees on standard contracts.

Your hosting agreement will clearly itemize all costs. If something is not in the agreement, you will not be charged for it. Period.

Yes, and the discounts are significant at scale:

• 100+ units / 500 kW+: Preferential per-kWh rates, typically 5-10% below standard pricing.
• 1 MW+: Custom enterprise pricing with dedicated infrastructure. Power rates can reach as low as $0.045-0.055/kWh depending on location and term.
• 5 MW+: Strategic partnership tier with the most aggressive rates, dedicated account management, priority maintenance, and custom SLA terms.
• 30 MW+ (facility-scale): Joint venture and build-to-suit options where we develop infrastructure to your specifications.

Volume discounts apply to both hosting rates and hardware procurement. The more you deploy with us, the more cost-effective every unit becomes. Contact our enterprise sales team for a custom proposal at your target scale.

Several of our Southwest facilities integrate solar energy into their power mix:

• On-site solar arrays: Select facilities have dedicated solar installations that offset a portion of grid consumption during peak sunlight hours.
• Solar PPA agreements: We negotiate Power Purchase Agreements with local solar producers to secure below-market rates for renewable energy, which flows through the grid to our facilities.
• Curtailment mining: We strategically co-locate with solar farms that experience curtailment (excess production that would otherwise be wasted). This surplus energy is available at deeply discounted rates.

Solar integration allows us to offer competitive rates while reducing the carbon footprint of compute operations. The intermittent nature of solar is managed through grid backstop -- your equipment never goes offline due to cloud cover. It simply means that a percentage of your energy comes from renewable sources.

Power requirements vary significantly by ASIC generation and model:

• Current generation (S21/M60 class): 3,000-3,600 W per unit at wall. A standard 20A/240V circuit supports one unit.
• Previous generation (S19 XP/M50S class): 2,800-3,250 W per unit. Slightly lower power but also lower hashrate.
• Hydro/immersion models (S21 Hydro): 5,000-5,500 W per unit, but with proportionally higher hashrate. Requires liquid cooling infrastructure.

When planning your deployment, calculate your total load as: number of units x watts per unit x 1.05 (the 5% accounts for networking, fans, and overhead). For example, 100 Antminer S21 units at 3,500 W each = 367.5 kW total load.

Our facilities provide 208V or 240V power to equipment racks via C13/C19 or direct wiring, depending on the deployment model. We size circuits with 80% utilization per NEC guidelines.

Yes, and this is one of our strategic advantages. Stranded energy refers to power generation capacity that cannot reach the grid cost-effectively, while curtailed energy is grid-connected generation that exceeds demand and would otherwise be wasted.

• Natural gas stranded sites: We co-locate compute containers at natural gas wellheads where pipeline access is limited. The gas would otherwise be flared (burned with zero economic value). Converting it to electricity for mining is environmentally superior to flaring and provides the cheapest energy available.
• Wind/solar curtailment: During periods of excess renewable generation, we absorb surplus power at deeply discounted rates. This improves grid stability and generator economics.
• Demand response: Several facilities participate in utility demand response programs, curtailing our load during peak grid stress in exchange for rate credits.

Our ability to flexibly consume power -- ramping up during surplus and down during scarcity -- makes us a beneficial grid participant and unlocks energy pricing unavailable to traditional data centers.

Our power distribution follows North American electrical standards:

• Utility feed: Medium voltage (typically 12.47 kV or 34.5 kV) stepped down through on-site transformers.
• Distribution: 480V 3-phase to main distribution panels, then stepped down to 208V or 240V at the rack/unit level.
• At the device: 208V or 240V single-phase for individual ASIC miners; 208V 3-phase for GPU servers and high-power equipment.
• Connectors: C13/C19 IEC power cords standard for rack-mounted equipment. Direct-wire options available for containerized deployments.

All electrical infrastructure is installed and maintained by licensed electricians and inspected per local electrical codes (NEC in the US). Circuit breakers, disconnect switches, and emergency power-off (EPO) systems are standard at all facilities.

Power continuity is a top priority. Our outage response includes multiple layers:

• UPS systems: Battery backup provides 5-15 minutes of power to allow controlled shutdown or generator startup. Critical networking and monitoring equipment has dedicated UPS coverage.
• Diesel generators: Automatic transfer switches (ATS) bring generators online within 15-30 seconds of a utility failure. Generators are tested weekly under load.
• Fuel reserves: On-site fuel storage for 24-72 hours of continuous generator operation. Fuel delivery contracts ensure resupply within 12 hours for extended outages.
• Dual utility feeds: Where available, facilities are connected to two independent utility feeds from different substations, providing automatic failover at the grid level.

For mining operations (which are tolerant of brief interruptions), some clients opt for lower-cost facilities without generator backup, accepting that utility outages mean temporary offline periods. This can save 10-15% on hosting costs.

We take sustainability seriously while being transparent about the energy intensity of our operations:

• Renewable energy mix: Across our portfolio, approximately 40-50% of our total energy consumption comes from renewable sources (hydro, wind, solar), with our Northwest facilities exceeding 80% renewable.
• Waste heat recovery: Select facilities redirect waste heat for building heating, agricultural applications, or district heating systems.
• Stranded gas utilization: By converting flared gas to electricity, we reduce the environmental impact of methane emissions (methane is 80x more potent as a greenhouse gas than CO2).
• Efficiency focus: Low PUE ratios (1.02-1.25) minimize the energy overhead of cooling and infrastructure.

We publish annual sustainability metrics and are working toward a goal of 60% renewable energy across all facilities by 2028. For clients who require carbon-neutral operations, we can direct your deployment to our highest-renewable locations.

We offer tiered networking to match different workload requirements:

• Standard (mining): Shared 1 Gbps uplink with NAT. Sufficient for pool connections, monitoring, and remote management. Included in standard hosting rates.
• Dedicated (AI/HPC): 1-100 Gbps dedicated uplinks with public IP addresses. Required for inference serving, data ingestion, and applications needing guaranteed bandwidth.
• InfiniBand (training): HDR (200 Gbps) and NDR (400 Gbps) InfiniBand fabric for GPU-to-GPU communication in training clusters. Not routable to the internet -- this is a dedicated compute interconnect.
• Private cross-connects: Direct connections to cloud providers (AWS Direct Connect, Azure ExpressRoute, GCP Partner Interconnect) for hybrid deployments.

Networking for mining operations is straightforward and included. For AI/HPC workloads, we design the network architecture as part of your deployment planning to ensure optimal performance.

Yes, with the following options:

• Mining operations: Typically use NAT behind our shared IP space. Miners connect outbound to pools and do not need public IPs. This is included at no extra charge.
• Dedicated IPs: Available for clients who need direct inbound access to their equipment (remote management, inference APIs, custom applications). Allocated from our PI space in blocks of /29 (5 usable) to /24 (253 usable).
• Bring your own IP (BYOIP): If you have your own IP address space, we can announce it via BGP from our network. Requires a valid LOA and IRR entries.

IPv4 addresses are a scarce resource and carry a small monthly fee per IP. IPv6 is available at no additional charge. For most GPU hosting and inference workloads, we recommend a small dedicated IP block for management and API endpoints.

Network latency varies by facility location and destination:

• To major mining pools (Foundry, Braiins): 10-40 ms from most facilities. Pool latency has minimal impact on mining revenue -- the difference between 10 ms and 100 ms is negligible for block reward probability.
• To major cloud regions (AWS us-east, us-west): 5-30 ms depending on facility proximity. Relevant for hybrid cloud/colo architectures.
• To internet exchanges (Equinix Ashburn, CoreSite LA): 10-50 ms. Important for inference serving and latency-sensitive applications.

For mining, latency is a non-issue -- focus on power cost. For AI inference serving end users, facility location matters more. We can help you select the right facility based on your latency requirements to specific endpoints.

Yes. For clients with multi-site deployments or hybrid infrastructure needs:

• Site-to-site VPN: Encrypted tunnels between your on-premises network and our facilities using IPSec or WireGuard. We can host the VPN endpoint on dedicated network equipment.
• Inter-facility VLAN: Layer 2 connectivity between Rax facilities for clients with deployments at multiple locations. Traffic stays on our private backbone without traversing the public internet.
• Private peering: Direct peering arrangements with major networks and cloud providers at select facilities.

For AI/HPC clients with strict data isolation requirements, we can provision fully isolated network segments with dedicated switching and firewalling. Your traffic never touches shared infrastructure.

Our network security includes multi-layered DDoS mitigation:

• Upstream filtering: Our transit providers implement basic volumetric DDoS filtering at the network edge, blocking common attack patterns before they reach our infrastructure.
• On-site mitigation: Dedicated DDoS scrubbing appliances at key facilities can absorb and filter sophisticated attacks targeting specific IP addresses.
• Black hole routing: In extreme cases, we can null-route targeted IPs at the BGP level to prevent collateral impact on other clients. This takes the targeted IP offline but protects the rest of the network.
• Third-party integration: For clients running public-facing applications, we recommend and integrate with services like Cloudflare or Akamai for application-layer (L7) DDoS protection.

For mining operations, DDoS risk is minimal since miners make outbound connections to pools. For inference and API hosting, we recommend a layered defense strategy designed during deployment planning.

Physical security is a top priority at every Rax facility:

• Perimeter: Fenced property with controlled vehicle and pedestrian access. Security cameras cover all exterior approaches 24/7 with 90-day retention.
• Building access: Electronic keycard/badge access with biometric verification (fingerprint or facial recognition) at facility entry points.
• Hosting floor: Separate access controls for the hosting floor. Only authorized personnel and escorted visitors can enter.
• Rack-level: Lockable racks and cages available for clients requiring physical isolation of their equipment. You control who has keys to your space.
• Personnel: Background checks on all employees with facility access. Visitor logs maintained with ID verification.

Select high-security facilities include 24/7 on-site security guards, man-trap entry vestibules, and intrusion detection systems. We will match the security posture to your requirements and the value of the equipment being hosted.

Our cybersecurity posture covers the infrastructure layer (network and systems we manage):

• Network segmentation: Client traffic is isolated using VLANs and access control lists. No cross-client traffic is possible without explicit configuration.
• Firewall and IDS: Stateful firewalls and intrusion detection systems monitor for anomalous traffic patterns and known attack signatures.
• Management access: All administrative access to infrastructure uses MFA, encrypted connections (SSH with key-based auth), and is logged with audit trails.
• Patch management: Network equipment and management systems are patched on a regular cycle. Critical vulnerabilities are addressed within 48 hours.

Important note: For colocation clients, cybersecurity of your equipment (OS hardening, application security, miner firmware) is your responsibility. For managed hosting clients, we include firmware management and basic security hardening as part of the service. We are happy to recommend security best practices for your specific deployment.

Here is how insurance works for equipment at our facilities:

• Our coverage: Rax carries comprehensive general liability and property insurance covering our facilities, including coverage for damage caused by our negligence, facility-level incidents (fire, flood, structural failure), and third-party liability.
• Your equipment: Client-owned equipment is not covered by our insurance policy. We strongly recommend that you maintain your own inland marine or property insurance for your hosted hardware.
• Optional coverage: We can connect you with insurance providers who specialize in technology equipment and cryptocurrency mining hardware. Premiums vary based on equipment value, location, and coverage scope.

Our hosting agreement includes a liability framework that clearly defines responsibilities. In the event of equipment damage due to a facility issue (e.g., cooling failure causing overheating), our insurance covers documented damages up to the limits specified in your agreement.

We treat client data privacy seriously, even though most of our workloads (mining, AI training) do not involve personal data:

• No data access: We do not access, inspect, or log any data on client-owned equipment. Our management access is limited to power, network, and hardware-level operations.
• Data destruction: When you decommission equipment, we offer secure data destruction services (NIST 800-88 compliant) including cryptographic erasure and physical drive destruction.
• Confidentiality: Client identities, deployment details, and commercial terms are treated as confidential. We do not disclose your presence at our facilities without your explicit consent.
• Law enforcement: We comply with valid legal process (subpoenas, court orders) but will notify you before disclosing any information unless legally prohibited from doing so.

For clients processing regulated data (healthcare, financial), we can discuss additional controls and compliance documentation as needed.

Fire safety is engineered into every facility:

• Detection: VESDA (Very Early Smoke Detection Apparatus) systems provide the earliest possible warning by continuously sampling air for smoke particles. Supplemented by heat detectors and manual pull stations.
• Suppression: Pre-action dry pipe sprinkler systems are standard. Water only enters the pipes when both a sprinkler head activates AND the detection system confirms a fire, virtually eliminating false-discharge water damage.
• Clean agent: Critical infrastructure areas (network rooms, control centers) use FM-200 or Novec 1230 clean agent systems that suppress fire without water or residue.
• Prevention: Hot work permits, strict no-smoking policies, combustible material controls, and regular electrical inspections minimize fire risk.

All fire suppression systems are inspected quarterly and maintained by certified fire protection contractors. Our staff receives annual fire safety training including evacuation procedures and extinguisher use.