Question 1

What is colocation?

Accepted Answer

Colocation means renting rack space, power, and cooling in someone else's data centre instead of building your own. The colocation provider owns and operates the building, power infrastructure, cooling systems, and physical security. The customer owns and manages their own hardware, but benefits from enterprise-grade infrastructure without the capital expenditure of building a facility. Colocation is priced per kilowatt of power per month, typically ranging from $90-220/kW/month depending on market, redundancy level, and contract term. It is the dominant model for enterprise IT, financial services, content delivery, and managed service providers.

Question 2

What is PUE (Power Usage Effectiveness)?

Accepted Answer

PUE is the standard metric for measuring data centre energy efficiency. It is calculated by dividing the total facility power consumption by the IT equipment power consumption. A PUE of 1.0 would mean all power goes directly to computing with zero overhead - theoretically perfect but physically impossible. A PUE of 2.0 means the facility uses as much power for cooling, lighting, and other systems as it does for IT. Modern best-in-class hyperscale facilities achieve PUE of 1.08-1.15. Good colocation facilities operate at 1.2-1.4. The industry average is approximately 1.58. Every 0.1 improvement in PUE saves roughly $50,000/year per MW at US average power rates.

Question 3

How much does colocation cost?

Accepted Answer

The short answer: $90-220/kW/month depending on where and what tier. Base rates range from $90/kW/month in emerging markets to $220/kW/month in constrained markets like Silicon Valley and New York. A typical enterprise deployment of 100 kW in a primary US market costs $12,000-16,000/month for the base lease, plus $500-800/month for metered power, $150-500/month per cross-connect (typically 4-20 needed), and $75-150/hour for smart hands support. All-in, a 1 MW deployment in Northern Virginia runs approximately $150,000-200,000/month. Key cost factors include power density, redundancy level (Tier III vs IV), contract length (longer terms yield 10-20% discounts), and interconnection requirements.

Question 4

What are data centre tiers?

Accepted Answer

Tiers I through IV rate how redundant a data centre is - how many things can break before your servers go offline. Tier I (99.671% uptime) has no redundancy - any maintenance requires a full shutdown. Tier II (99.741%) adds redundant UPS and generators but a single distribution path. Tier III (99.982%) has multiple distribution paths allowing any component to be maintained without downtime - this is the standard for commercial colocation. Tier IV (99.995%) is fully fault-tolerant with simultaneously active dual paths, meaning any single failure has zero impact on operations. Construction costs increase roughly 30-50% per tier level, from $6-8M/MW for Tier I to $16-22M/MW for Tier IV.

Question 5

What is liquid cooling for data centres?

Accepted Answer

Liquid cooling pipes coolant directly to the chips instead of blowing air over them. It removes 5-10x more heat per rack than air cooling. Two primary methods exist: direct-to-chip liquid cooling (DLC), where cold plates mounted on CPUs and GPUs circulate coolant to absorb heat at the source; and immersion cooling, where entire servers are submerged in non-conductive fluid. DLC supports rack densities of 80-150+ kW (vs 15-20 kW with air) and achieves PUE of 1.1-1.3. Immersion cooling pushes this further to 100-250+ kW per rack equivalent and PUE of 1.02-1.10. Liquid cooling has become essential for AI workloads using NVIDIA H100/H200 and Blackwell GPUs, which generate far more heat than traditional air cooling can handle. The market is projected to reach $32 billion by 2028.

Question 6

What is an Internet Exchange Point (IXP)?

Accepted Answer

An IXP is where networks plug into each other directly instead of paying a middleman to carry their traffic. IXPs reduce latency, lower transit costs, and increase network resilience. Major IXPs like DE-CIX Frankfurt, AMS-IX Amsterdam, and LINX London handle terabits of traffic per second. For data centre operators and tenants, proximity to an IXP is a key site selection factor because it enables direct peering relationships and reduces the cost of moving data. Carrier-neutral colocation facilities near IXPs command premium pricing because of the interconnection density they offer.

Question 7

How much power does a data centre use?

Accepted Answer

Data centre power consumption ranges from under 1 MW for a small enterprise facility to 200+ MW for a hyperscale campus. A single standard server rack consumes 5-10 kW. An AI training rack with NVIDIA GPUs consumes 40-120+ kW. A typical 10 MW colocation facility powers roughly 1,000-2,000 standard racks or 80-250 high-density AI racks. Globally, data centres consumed approximately 460 TWh of electricity in 2025 - roughly 2% of total global electricity generation. This is projected to reach 1,000 TWh by 2030, driven primarily by AI workloads. Power availability has become the primary constraint on new data centre development, with transformer lead times extending to 36 months and utility interconnection queues in some markets stretching to 2030+.

Question 8

What is a carrier-neutral data centre?

Accepted Answer

A carrier-neutral facility lets you pick from dozens of network providers instead of being locked into one. More carriers means better pricing and more redundancy. This means tenants can choose from dozens or hundreds of network providers, negotiate competitive rates, and establish direct cross-connects to peers and cloud on-ramps. The alternative - a carrier-specific facility - locks tenants into a single provider's network, limiting connectivity options and pricing leverage. Major carrier-neutral operators include Equinix, Digital Realty, and CoreSite. Carrier neutrality is especially valuable in primary interconnection markets like Ashburn, Chicago, and Frankfurt where network density drives real business value.

Question 9

How do I choose a data centre location?

Accepted Answer

Six factors determine data centre location suitability: power infrastructure (substation proximity, transmission line access, utility capacity), connectivity (IXP proximity, fibre density, carrier diversity), natural disaster risk (flood zones, seismic hazard, severe weather exposure), market maturity (existing DC capacity, operator ecosystem, tenant demand), regulatory environment (zoning, permitting timeline, tax incentives, community sentiment), and cost structure (power rates, construction costs, labour, taxes). Power is the most heavily weighted factor because without adequate electrical infrastructure, nothing else matters. Use our free Viability Score tool to get an instant assessment of any US address across these dimensions.

Question 10

What is N+1 and 2N redundancy?

Accepted Answer

N means exactly what you need with zero backup. N+1 means one spare. 2N means everything is doubled. The notation tells you how many things can fail before you lose power or cooling. N means the exact number needed with no backups - if one unit fails, the system fails. N+1 means one additional backup unit exists: in a system needing 4 cooling units, N+1 provides 5. If any single unit fails, the remaining 4 handle the full load. 2N means the entire system is duplicated: two complete, independent sets of infrastructure. If an entire power feed fails, the second one carries the full load. 2N+1 adds one extra unit on top of the duplicated system. Higher redundancy directly correlates with higher uptime and higher cost. N+1 is standard for Tier III facilities; 2N is standard for Tier IV.

Question 11

How much does it cost to build a data centre?

Accepted Answer

Data centre construction costs average $11.3M per megawatt in the US as of 2026, but range from $8M/MW in low-cost markets like Texas, Ohio, and Nevada to $14-18M/MW in premium markets like Northern Virginia and Silicon Valley. A 10 MW facility costs roughly $80-180M depending on location and specification. Key cost components: shell and core construction ($3-5M/MW), mechanical/cooling systems ($2-4M/MW), electrical/power distribution ($3-5M/MW), fire suppression and controls ($0.5-1M/MW). On top of construction, budget for land ($50K-500K/acre), utility interconnection ($1-10M+), permitting ($200K-2M), and design/engineering (5-8% of construction). Monthly construction starts in the US averaged $10.1 billion through early 2026 - a 600% increase over two years.

Question 12

What is edge computing and edge data centres?

Accepted Answer

Edge computing processes data closer to where it is generated rather than sending it to a centralised cloud or enterprise data centre. Edge data centres are smaller facilities (typically 0.5-5 MW) deployed in metropolitan areas to reduce latency for applications that require near-real-time response: autonomous vehicles, IoT processing, content delivery, 5G network functions, and augmented reality. Edge facilities are typically Tier II or III, optimised for low latency rather than massive scale. The edge market is projected to reach $317 billion by 2028, driven by 5G rollouts and real-time AI inference at the network perimeter. Key edge operators include EdgeConneX, DataBank, and Vapor IO. Edge complements rather than replaces centralised hyperscale - most architectures use both tiers.

Question 13

What is a data centre REIT?

Accepted Answer

A data centre REIT (Real Estate Investment Trust) owns and operates data centre properties while passing at least 90% of taxable income to shareholders as dividends. Equinix is the largest at ~$85 billion market cap, followed by Digital Realty at ~$55 billion. Together, publicly traded DC REITs control over 3,000 facilities worldwide and generate $30+ billion in annual revenue. They offer investors liquid exposure to the data centre sector without building or operating facilities directly. REIT structures dominate the colocation market because the asset class delivers predictable, long-term contracted revenue with 95%+ tenant retention rates.

Question 14

What is dark fibre?

Accepted Answer

Dark fibre is unlit, unused optical fibre cable leased as raw infrastructure — the buyer provides their own optical equipment to light it. Lit fibre comes as a managed service with the provider's equipment already active. Enterprises, hyperscalers, and carriers buy dark fibre for maximum control over capacity, latency, and security. Typical lease costs run $5-15 per metre per year on long-term IRUs (Indefeasible Rights of Use), usually 15-25 year terms. Dark fibre is essential for connecting data centres within a campus or metro area where sub-millisecond latency and unlimited bandwidth scaling matter.

Question 15

What is a cross-connect?

Accepted Answer

A cross-connect is a physical cable — copper or fibre — that directly links two tenants, or a tenant to a carrier, inside the same data centre. It bypasses the public internet entirely, delivering lower latency, higher security, and dedicated bandwidth. Cross-connects cost $150-500/month each depending on the facility and media type (single-mode fibre is standard). A typical enterprise tenant maintains 4-20 cross-connects for carrier diversity, cloud on-ramps, and peering. Cross-connect revenue is high-margin recurring income for data centre operators and a primary reason tenants cluster in carrier-dense facilities.

Question 16

What does carrier hotel mean?

Accepted Answer

A carrier hotel is a data centre where dozens or hundreds of telecommunications carriers, ISPs, and enterprises converge to interconnect. The term dates to the late 1990s when telecom companies began concentrating in specific buildings for peering and transit. Iconic carrier hotels include One Wilshire in Los Angeles (over 350 carriers), 60 Hudson Street in New York (a former Western Union telegraph building), and Telehouse London Docklands. These buildings remain critical internet infrastructure because network effects are self-reinforcing — more carriers attract more tenants, which attract more carriers. Space in premier carrier hotels commands 2-3x the price of equivalent colocation elsewhere.

Question 17

What is a meet-me room?

Accepted Answer

A meet-me room (MMR) is the dedicated space inside a data centre where carriers and tenants physically interconnect via cross-connects and patch panels. It is the nerve centre of any carrier-neutral facility. MMRs are either open (shared racks accessible to all carriers) or caged (individual carriers get locked enclosures). Access is tightly controlled with biometric security and 24/7 CCTV. Operators charge $500-2,000/month per rack unit in the MMR, plus cross-connect fees. The number of carriers present in a facility's MMR is one of the strongest indicators of its interconnection value and long-term tenant stickiness.

Question 18

How long does it take to build a data centre?

Accepted Answer

Traditional purpose-built data centres take 18-36 months from groundbreaking to live load, depending on market permitting timelines and utility interconnection queues. Modular/prefabricated designs compress this to 6-12 months because major components arrive factory-tested and ready to connect. Hyperscale campuses with multiple buildings and on-site substations take 3-5 years for full buildout, though operators typically energise buildings in phases. The biggest bottleneck in 2026 is not construction — it is power. Transformer lead times run 18-36 months, and utility interconnection queues in top markets like Northern Virginia and Dublin now stretch 3-5 years. Smart operators pre-order electrical gear before they even break ground.

Question 19

What is the Uptime Institute?

Accepted Answer

The Uptime Institute is the global authority on data centre reliability, best known for its Tier Classification System (Tiers I-IV). Founded in 1993, it has certified over 3,000 facilities across 120+ countries. Uptime offers three distinct certifications: Tier Certification of Design Documents (paper review), Tier Certification of Constructed Facility (on-site verification), and Tier Certification of Operational Sustainability (ongoing operational assessment renewed every three years). A common misconception is that facilities are "Tier III" just because they have N+1 redundancy — without formal Uptime certification, that claim is marketing, not verified fact. Certification costs $50,000-150,000+ depending on tier level and facility size.

Question 20

What is a powered shell data centre?

Accepted Answer

A powered shell is a data centre building delivered with the structure, utility power feed, and core electrical infrastructure in place, but without cooling, racks, or IT fit-out. It sits between a bare shell (just walls and roof) and a fully fitted facility (turnkey with all mechanical, electrical, and IT infrastructure). Hyperscalers and large colocation operators prefer powered shells because they install their own proprietary cooling and power distribution designs. Lease structures are typically 15-25 year triple-net (NNN), with the tenant responsible for fit-out capital ($4-8M/MW) and ongoing maintenance. Powered shell deals dominate the wholesale market, with rents running $100-180 per kW/month before tenant fit-out costs.

Question 21

What are data centre SLAs?

Accepted Answer

Data centre SLAs (Service Level Agreements) are contractual guarantees covering uptime, power availability, temperature, humidity, and incident response times. Standard colocation SLAs guarantee 99.99% uptime (52 minutes of allowed downtime per year) for Tier III facilities and 99.999% (5.3 minutes) for Tier IV. When the provider misses the SLA, tenants receive service credits — typically 5-30% of monthly fees per incident, capped at one month's charges. The fine print matters: most SLAs exclude scheduled maintenance windows, force majeure events, and tenant-caused outages. Negotiate for uncapped penalties, root-cause analysis requirements within 48 hours, and the right to terminate without penalty after repeated SLA breaches.

Question 22

What is a hyperscaler?

Accepted Answer

A hyperscaler is a cloud or internet company that operates at massive scale — hundreds of thousands of servers across dozens of data centres worldwide. The five dominant hyperscalers are AWS, Microsoft Azure, Google Cloud, Meta, and Oracle, with combined annual data centre capex exceeding $250 billion in 2026. Qualifying as a hyperscaler generally means spending $2+ billion per year on data centre infrastructure and operating at multi-gigawatt scale. Hyperscalers both build their own facilities and lease wholesale colocation from third parties, consuming 60-70% of all new data centre capacity globally. Their procurement decisions — where they build, who they lease from, what power they require — shape the entire data centre industry.

Question 23

What is data centre interconnection?

Accepted Answer

Interconnection is the business of connecting tenants, carriers, cloud providers, and networks inside and between data centres. It goes beyond basic connectivity — it creates an ecosystem where proximity to other tenants and networks generates compounding value. Direct cloud on-ramps (private connections to AWS, Azure, Google) reduce latency by 30-60% compared to public internet paths. Interconnection revenue now represents 25-35% of revenue for major colocation operators like Equinix, growing 8-12% annually. The three primary interconnection products are cross-connects (within a facility), metro connects (between facilities in a metro), and fabric services (software-defined virtual connections to hundreds of endpoints).

Question 24

What is wholesale colocation?

Accepted Answer

Wholesale colocation means leasing large, dedicated data hall space — typically 1 MW or more — with the tenant managing their own fit-out within the provider's shell. Retail colocation, by contrast, sells individual racks or cages in shared halls with the provider managing all infrastructure. Wholesale pricing runs $90-140/kW/month, 20-40% cheaper than retail, because the tenant takes on more operational responsibility and commits to longer terms (5-15 years). Wholesale tenants include hyperscalers, large enterprises, SaaS companies, and managed service providers who need control over their environment. The wholesale market is growing faster than retail, driven by AI-related demand for 5-50 MW blocks of contiguous capacity.

Question 25

What is a data centre campus?

Accepted Answer

A data centre campus is a multi-building development on a single site, typically ranging from 50 MW to 500+ MW of total IT capacity. Campuses share common infrastructure — substations, fibre paths, water supply, security perimeters, and operations teams — creating economies of scale impossible with standalone facilities. Major campus clusters include Ashburn, Virginia (2,000+ MW across dozens of campuses), Prineville, Oregon (Meta and Apple mega-campuses), and Amsterdam's Schiphol-Rijk corridor. Building in phases allows operators to match capital deployment to demand, with each new building taking 12-18 months once site infrastructure is in place. Campus sites require 20-100+ acres and proximity to high-voltage transmission (115 kV+).

Question 26

How many data centres are there in the world?

Accepted Answer

There are approximately 11,000+ commercial colocation and hyperscale data centres globally as of 2026, with the total growing at 15% per year. The United States dominates with roughly 40% of global capacity, followed by Europe (25%), Asia-Pacific (25%), and the rest of the world (10%). The top five markets by power capacity are Northern Virginia (4,000+ MW), Dallas-Fort Worth, Phoenix, Chicago, and London. Including enterprise-owned facilities, the total number exceeds 50,000 worldwide. However, the trend is toward consolidation — fewer, larger facilities replacing many small server rooms. Over the past decade, the number of enterprise-owned data centres has declined 30% while total capacity has tripled.

Question 27

What is data sovereignty?

Accepted Answer

Data sovereignty refers to laws requiring that data generated in a country must be stored and processed within that country's borders. The EU's GDPR is the most prominent example, restricting transfers of personal data outside the European Economic Area without adequate safeguards. Other strict regimes include China's Cybersecurity Law, Russia's Federal Law No. 242-FZ, and India's Digital Personal Data Protection Act. For data centre operators, sovereignty requirements drive demand for in-country facilities even in markets that would otherwise lack scale justification. This has accelerated data centre construction in secondary European markets (Warsaw, Milan, Madrid) and emerging regions (Saudi Arabia, Indonesia, Brazil) where local data residency is now legally mandated.

Question 28

What is a UPS system in a data centre?

Accepted Answer

A UPS (Uninterruptible Power Supply) bridges the gap between a power outage and generator startup — typically 10-30 seconds. Without a UPS, even a momentary grid flicker would crash every server in the facility. Three main types exist: static battery UPS (most common, uses lithium-ion or VRLA batteries, 5-15 minute runtime), rotary UPS (spinning flywheel stores kinetic energy, 15-30 seconds of ride-through), and diesel rotary UPS (DRUPS, combines flywheel and diesel generator in one unit). UPS systems are sized in kVA and typically rated at N+1 or 2N redundancy. A 10 MW data centre requires $3-6M in UPS infrastructure alone, representing 15-20% of total electrical system cost.

Question 29

What is hot aisle/cold aisle containment?

Accepted Answer

Hot aisle/cold aisle arranges server racks so that all rack fronts face one aisle (cold) and all rack backs face the adjacent aisle (hot), preventing cooled and exhaust air from mixing. Without containment, hot exhaust recirculates and forces cooling systems to work 20-40% harder. Containment adds physical barriers — curtains, doors, or rigid panels — to fully seal either the hot or cold aisle. Cold aisle containment is more common and encloses the supply air path; hot aisle containment captures exhaust and routes it directly to cooling return plenums. Proper containment improves PUE by 0.1-0.3 points and is a prerequisite for rack densities above 10 kW. It costs $500-2,000 per rack to retrofit but pays for itself in energy savings within 12-18 months.

Question 30

What is a data centre PPA?

Accepted Answer

A PPA (Power Purchase Agreement) is a long-term contract where a data centre operator agrees to buy electricity directly from a renewable energy generator — typically a wind or solar farm — at a fixed price for 10-25 years. Hyperscalers are the biggest PPA buyers: Microsoft has contracted 19+ GW of renewable energy, Amazon 25+ GW, and Google 12+ GW. PPAs serve two purposes — they lock in predictable power costs (typically $30-60/MWh) and satisfy corporate sustainability commitments. Two structures dominate: physical PPAs (power is delivered directly via the grid) and virtual/financial PPAs (a contract-for-differences hedge where physical delivery is separate). For data centre operators, signing a PPA of 50 MW+ often accelerates utility interconnection because it demonstrates credit-worthy, long-term demand.

Frequently Asked Questions