When AWS or Azure to Colocation Actually Saves Money

Cloud repatriation went from heretical to mainstream in eighteen months. 37signals publicly documented saving over ten million dollars across five years moving Basecamp and HEY workloads off AWS to dedicated infrastructure. Dropbox, GEICO, X, and a growing list of enterprises have followed with similar documented economics. Andreessen Horowitz published research arguing cloud costs at scale represent a trillion-dollar drag on technology company valuations.

The conversation has moved past whether cloud repatriation makes financial sense to which specific workloads belong where. The honest answer requires actual math, not ideology.

Consider this your independent cloud repatriation review — written by an advisor with no financial stake in which infrastructure model your organization chooses.

Bottom Line: Cloud repatriation produces material savings for stable predictable workloads running at consistent utilization but costs more for variable elastic workloads with significant demand variance. The crossover point typically arrives at 50-60 percent sustained utilization for compute workloads and 40-50 percent for storage-heavy workloads. Enterprise organizations with established cloud spend exceeding two million dollars annually almost universally find specific workload categories that repatriate cost-effectively. The savings range from 40-65 percent on repatriated workload total cost of ownership over a 5-year horizon. The strategic outcome for most enterprises is hybrid architecture rather than complete cloud exit. Metro Colo Advisory evaluates cloud repatriation economics for your specific workload portfolio at no cost.

Key Takeaways:

• Cloud repatriation typically saves 40 to 65 percent TCO over 5 years for workloads at 50 to 85 percent sustained utilization
• The cloud-to-colocation crossover point arrives at 50 to 60 percent utilization for compute and 40 to 50 percent for storage-heavy workloads
• Data egress fees often produce larger repatriation savings than CIOs anticipate, particularly for workloads with significant integration touchpoints
• Variable customer-facing workloads with high traffic variance typically remain better in cloud where elasticity is genuinely needed
• The most common enterprise outcome is hybrid architecture (30 to 60 percent of workloads repatriated) rather than complete cloud exit

Get My Free Cloud Repatriation Analysis →

Why Cloud Repatriation Stopped Being Controversial

For roughly fifteen years from the launch of AWS in 2006 through 2021, the dominant enterprise IT narrative positioned cloud migration as the default direction. Workloads moved from on-premise to cloud. The exceptions were workloads that could not move for compliance or technical reasons. The economic question was assumed to be settled in cloud’s favor.

The narrative shifted as enterprise cloud bills exceeded original projections by significant margins. Multiple factors converged to drive the reassessment. Cloud workload utilization patterns stabilized as enterprises moved past initial migration and into production operations. The elasticity premium that justified cloud pricing during variable workload phases became expensive overhead during stable production phases. Data egress fees compounded as cloud workloads developed dependencies that required moving data between cloud regions, between cloud providers, and between cloud and integrated business systems. The actual cost of running stable production workloads in cloud became visible in ways that initial migration economics had not predicted.

The Andreessen Horowitz research argued that cloud spend at scale represents a meaningful drag on enterprise economics — particularly for software companies where infrastructure cost directly affects gross margins. Their analysis suggested that approximately fifty percent of cloud spend at the public software company level could be more efficiently deployed in dedicated infrastructure for the specific workloads that justified the move.

The 37signals case studies provided concrete public documentation of the savings. Multi-million dollar annual savings on workloads that had been considered settled in cloud. The transparency around the cost analysis, infrastructure choices, and operational outcomes made cloud repatriation a legitimate strategic conversation rather than a contrarian position. Industry analyst IDC research now documents repatriation as a mainstream strategic option rather than a contrarian position.

What changed is not the underlying economics. Cloud has always been expensive at high utilization. What changed is enterprise willingness to do the math honestly and act on the results.

The Workload Utilization Math That Determines the Answer

The cloud-versus-colocation decision is fundamentally a utilization math problem. Cloud pricing assumes customers will use infrastructure intermittently and pay premium hourly rates for the flexibility. Workloads with utilization patterns matching that assumption benefit from cloud economics. Workloads with stable predictable utilization patterns pay the elasticity premium without consuming the elasticity benefit.

The honest framework for evaluating which workloads belong where requires segmenting your workload portfolio by utilization pattern rather than treating all workloads identically.

Workload Type vs Cloud Repatriation Economics

Workload Type	Utilization Pattern	Repatriation Economics	Recommended Infrastructure
Stable production applications	60-85% sustained utilization	Strong savings at 50-65% TCO reduction	Dedicated colocation
Predictable batch processing	40-70% scheduled utilization	Moderate savings at 30-50% TCO reduction	Dedicated colocation with reserved capacity
Database and storage-heavy workloads	70-90% sustained utilization	Strong savings at 55-70% TCO reduction	Dedicated colocation
Development and testing environments	20-40% intermittent utilization	Mixed — depends on scale	Cloud or hybrid
Variable customer-facing applications	High variance, unpredictable peaks	Negative — cloud wins	Cloud
AI inference at production scale	60-80% sustained utilization	Strong savings at 45-60% TCO reduction	Dedicated high density colocation
Disaster recovery infrastructure	Low utilization until activated	Mixed — depends on RTO requirements	Hybrid or dedicated based on tier
Compliance-bound regulated workloads	Variable but constrained	Strong savings plus compliance benefits	Dedicated colocation

For organizations specifically evaluating disaster recovery colocation as part of broader cloud repatriation initiatives, the DR facility selection involves additional considerations beyond pure cost economics including geographic separation requirements, recovery time objectives, and matching compliance posture between primary and DR environments.

Production applications running stable utilization represent the cleanest economic case for cloud repatriation. Workloads that ran fine for years on dedicated infrastructure before cloud migration and continue running predictably in cloud are typically paying the elasticity premium without using elasticity. The repatriation conversation for these workloads is increasingly straightforward.

Database and storage-heavy workloads represent the second strongest case. Cloud storage and database services carry significant premiums versus dedicated equivalents — particularly for workloads with high I/O throughput or large persistent storage footprints. Egress fees on database workloads with integration touchpoints across multiple systems compound the cost gap further.

Variable customer-facing applications with high traffic variance and unpredictable demand peaks represent the cleanest case for staying in cloud. The flexibility to scale to zero during low demand and scale aggressively during peaks genuinely matters for these workloads. Paying premium hourly rates is the price of operational flexibility that variable workloads actually need.

Development and testing environments depend on scale. Small development environments at low utilization typically belong in cloud. Large development infrastructure at significant scale frequently produces meaningful savings under cloud repatriation.

Why Most Cloud Bills Are Larger Than CIOs Expected

The disconnect between projected and actual cloud spend has several specific causes that compound over time. Understanding the drivers helps identify which cost categories drive the largest savings under repatriation.

Compute instance pricing premiums. Cloud compute instances are priced assuming intermittent use. The hourly rate includes a meaningful premium versus dedicated infrastructure equivalent compute. For workloads running 24/7 at significant utilization, the cumulative premium becomes substantial. A workload that runs 8,760 hours per year pays the full hourly premium 8,760 times.

Data egress fees that compound over time. Cloud providers charge for data movement out of their networks. Enterprise workloads develop integration patterns that move data between cloud regions, between cloud providers, and between cloud and on-premise systems. The egress costs that seemed manageable at low data volumes scale linearly as data movement increases.

Storage premium pricing. Cloud storage carries premiums versus dedicated equivalents — particularly for workloads with significant persistent storage footprint or high I/O throughput requirements. Database workloads, log retention workloads, and content management workloads frequently develop large storage cost lines that exceed original projections.

Reserved capacity commitments that under-deliver. Cloud reserved instance pricing offers discounts for committing to capacity over 1-3 year terms. The discount sounds meaningful in negotiation but the underlying pricing still includes elasticity premiums. Reserved capacity discounts of 30-50 percent off on-demand pricing still leave significant premiums versus dedicated infrastructure.

Service tier markup compounding. Cloud providers offer managed services that abstract operational complexity at meaningful premiums versus self-managed equivalents. Managed database services, managed Kubernetes, managed storage all carry markups that compound as workloads grow. Some of these services are worth the premium for the operational simplification. Many are not at scale.

Cross-region data transfer costs. Workloads that span multiple cloud regions for redundancy or geographic distribution incur cross-region data transfer fees that frequently surprise enterprise customers. The fees seemed negligible during initial architecture but scale meaningfully as cross-region traffic increases.

Premium support pricing tiers. Enterprise cloud customers frequently subscribe to premium support tiers that cost meaningful percentages of total cloud spend. The support value varies by organization but the cost is real.

The cumulative effect of these factors typically produces cloud bills 30-60 percent higher than initial projections at the 24-36 month mark of enterprise cloud deployments. Cloud repatriation economics become attractive when the cumulative premium reaches levels where dedicated infrastructure savings justify the migration effort.

The Honest 5-Year TCO Comparison

The financial case for cloud repatriation requires concrete numbers across realistic workload profiles. The comparison below reflects current market pricing for typical enterprise workload profiles considering repatriation.

Cloud vs Dedicated Colocation — 5-Year Total Cost of Ownership

Cost Category	AWS or Azure Cloud	Dedicated Colocation
Compute capacity (100 vCPU equivalent stable workload)	$180,000-$240,000 annually	Hardware amortized over 4-year refresh
Annual compute infrastructure cost	$180,000-$240,000	$60,000-$90,000
Storage (50TB persistent)	$36,000-$60,000 annually	$12,000-$24,000 annually
Data egress and transfer fees	$48,000-$120,000 annually	Minimal (cross-connect only)
Network connectivity	Included in cloud pricing	$18,000-$30,000 annually
Power and cooling	Included in cloud pricing	$24,000-$48,000 annually
Colocation space (2-4 racks)	Not applicable	$24,000-$48,000 annually
Hardware capital (amortized)	Not applicable	$35,000-$60,000 annually
Premium support subscription	$24,000-$60,000 annually	$0-$18,000 annually
Total annual operational cost	$288,000-$480,000	$173,000-$318,000
5-year total cost of ownership	$1,440,000-$2,400,000	$865,000-$1,590,000
5-year savings range	Baseline	$575,000-$810,000 (40-50% TCO reduction)
Marginal cost economics after year 2	Scales linearly with usage	Cost stable, marginal usage near zero

Get My Free Cloud Repatriation Analysis →

The comparison reflects realistic mid-market workload profiles running stable production utilization. Workloads with higher utilization, larger storage footprints, or significant data egress patterns produce larger comparative savings under repatriation. Workloads with high variance or lower utilization produce smaller comparative savings or favor remaining in cloud.

Several specific dimensions deserve closer attention in the comparison:

• Hardware refresh cycle economics. Dedicated colocation deployments refresh hardware on cycles that match workload performance requirements rather than continuously paying premium pricing for current-generation infrastructure. Refresh cycles typically run 3-5 years for general enterprise compute and 2-3 years for high-performance workloads. The refresh cycle economics genuinely matter — dedicated infrastructure that runs 4 years before refresh delivers significantly better economics than infrastructure refreshed annually.
• Data egress savings often exceed expectations. The data egress line item produces larger savings than CIOs typically anticipate. Workloads that have grown organically frequently develop data movement patterns that the original architecture did not predict. Eliminating egress costs entirely under colocation produces line item savings that compound across the deployment lifetime.
• Operational complexity considerations. The cloud comparison sometimes understates the operational cost dimension. Cloud managed services reduce operational burden compared to self-managed dedicated infrastructure. Repatriation may shift some operational cost from cloud subscription to internal staff time or third-party operational support. The honest TCO comparison includes operational dimensions, not just infrastructure subscription cost.
• Reliability and uptime considerations. Cloud SLAs and dedicated colocation SLAs differ in structure and operational characteristics. Enterprises with strong operational capabilities frequently achieve better reliability outcomes on dedicated infrastructure than on cloud. Enterprises without strong operational capabilities may achieve better outcomes on cloud despite higher cost. The reliability dimension matters in the TCO conversation but is workload and organization specific.

When Cloud Still Wins — The Honest Counter-Case

The cloud repatriation case is strong for specific workload categories but does not extend universally. CIOs evaluating repatriation should resist treating the decision as ideological in either direction. Several workload categories genuinely produce better outcomes on cloud than on dedicated infrastructure.

Variable customer-facing applications with high traffic variance. Consumer applications, e-commerce workloads with seasonal spikes, and viral content platforms benefit from cloud elasticity. The ability to scale to zero during low demand and scale aggressively during peaks justifies premium hourly pricing. Dedicated infrastructure sized for peak demand would cost dramatically more than cloud elastic infrastructure for these workloads.

Early-stage workloads with uncertain trajectories. Workloads still defining their utilization patterns benefit from cloud flexibility. Provisioning dedicated infrastructure before utilization patterns stabilize creates significant risk of over-provisioning or under-provisioning. Most enterprise workloads should start in cloud and consider repatriation only after utilization patterns stabilize.

Geographically distributed workloads requiring presence in multiple regions. Cloud providers operate dozens of regions worldwide. Replicating that geographic footprint through dedicated colocation requires meaningful operational investment. For workloads genuinely requiring presence across many regions, cloud provides geographic flexibility that dedicated infrastructure cannot match cost-effectively.

Workloads dependent on cloud-native services without dedicated equivalents. Some cloud services have no operational equivalent in dedicated infrastructure. Serverless compute platforms, managed AI services, specialized analytics platforms, and certain integration services exist primarily as cloud-native offerings. Workloads dependent on these services face significant re-architecture costs under repatriation that may not justify the savings.

Small organizations without operational infrastructure expertise. Cloud abstracts significant operational complexity that smaller organizations cannot easily replicate internally. Organizations under approximately fifty technical employees frequently find that the operational simplification cloud provides justifies the cost premium. Repatriation economics improve as organizations grow operational capability.

Workloads with strong cloud provider integration value. Some enterprise workloads benefit from deep integration with cloud provider ecosystems including marketplace integrations, partner relationships, or specific compliance certifications that exist only in cloud configurations. The integration value can outweigh the cost premium for these specific cases.

The honest enterprise outcome for most organizations is hybrid architecture rather than complete cloud exit. Specific workload categories repatriate to dedicated infrastructure while other workload categories remain in cloud. The decision matrix runs workload by workload rather than universal direction.

The Cloud Repatriation Decision Framework

The honest cloud repatriation analysis follows a structured framework that evaluates each workload against specific criteria rather than treating the decision as a single binary choice across the entire portfolio.

Cloud Repatriation Decision Framework by Workload Category

Decision Criteria	Repatriate to Colocation	Stay in Cloud	Build Hybrid Architecture
Workload utilization pattern	50-85% sustained	Variable, spiky	Mix of stable and variable
Data movement volume	High egress costs	Minimal egress	Significant inter-workload movement
Compliance requirements	Strict, dedicated infrastructure beneficial	Standard, cloud sufficient	Mixed across workload types
Predictability of demand	High predictability	Low predictability	Mixed predictability
Storage footprint	Large persistent storage	Small or transient	Mixed across workloads
Operational team capability	Strong internal infrastructure expertise	Limited infrastructure expertise	Adequate for hybrid management
Cloud-native service dependency	Limited dependency on cloud-only services	Heavy dependency	Workloads use different services
Geographic distribution requirements	Centralized or limited regions	Many global regions required	Mixed by workload

The honest answer for most enterprises crosses categories. Some workloads repatriate cleanly. Other workloads belong firmly in cloud. The remaining workloads exist somewhere in between with answers that depend on specific business context.

The strategic outcome typically takes one of three forms. Complete cloud exit applies to organizations where most or all workloads share characteristics favoring dedicated infrastructure and where operational capability supports running infrastructure internally. This outcome is rare for enterprises but common for specific company profiles.

Partial repatriation applies to organizations that identify specific workload categories for repatriation while maintaining cloud presence for variable workloads, geographic distribution, or cloud-native service dependencies. This is the most common outcome for enterprise repatriation initiatives. Typically 30-60 percent of cloud workloads move to dedicated infrastructure while the remainder continues in cloud.

Hybrid cloud colocation architecture applies to organizations building deliberate architectures that span dedicated infrastructure and cloud environments with significant data movement between them. ServiceFabric, Open Cloud Exchange, and similar interconnection platforms enable hybrid architectures where workloads move between dedicated and cloud environments based on operational requirements rather than locked into single infrastructure models.

The Repatriation Migration Reality

The operational complexity of moving workloads from cloud to dedicated infrastructure varies dramatically by workload profile. Some repatriations are straightforward. Others require significant re-architecture effort. The honest migration planning factors in workload-specific complexity rather than treating all repatriations identically.

• Phase 1: Workload portfolio analysis (4-8 weeks). Document current cloud spend by workload, identify utilization patterns, evaluate cloud-native service dependencies, and segment the portfolio into repatriation candidates, hybrid candidates, and cloud-only workloads. The data center migration framework covers the broader migration approach that applies to repatriation initiatives.
• Phase 2: Facility selection and contract negotiation (6-10 weeks). Evaluate colocation facilities against the workload portfolio requirements including geographic considerations, compliance requirements, hybrid cloud connectivity needs, and capacity planning for repatriated workloads. The colocation site selection framework covers the complete facility evaluation process. Running an independent provider comparison across all qualifying facilities generates the competitive pressure that produces materially better contract terms. The colocation contract guide covers the term framework that applies to repatriation infrastructure contracts.
• Phase 3: Infrastructure procurement and deployment (8-16 weeks). Hardware procurement for repatriation deployments depends significantly on workload type. General enterprise compute typically runs 6-10 weeks for procurement. High-performance workloads may run 12-16 weeks. Network infrastructure, storage systems, and integration components add to the procurement timeline.
• Phase 4: Workload migration in waves (12-24 weeks). Repatriation typically proceeds in waves rather than single cutover events. Lower-risk workloads migrate first to validate operational procedures. Higher-risk workloads migrate later with established migration patterns. Each wave includes pre-migration testing, migration execution, post-migration validation, and cloud termination for the migrated workloads.
• Phase 5: Hybrid architecture optimization (8-16 weeks). For organizations building hybrid architectures rather than complete cloud exit, the post-migration phase includes optimization of the hybrid model including data movement patterns, integration architecture, and operational procedures spanning both environments.

Total timeline for enterprise repatriation initiatives typically runs 6-12 months for partial repatriation and 12-18 months for complete cloud exit. The timeline is meaningfully longer than initial cloud migrations because workloads have developed operational dependencies and architectural patterns during cloud deployment that require careful unwinding.

Compliance and Repatriation — The Often-Overlooked Driver

Cloud repatriation conversations frequently focus on cost optimization without recognizing the compliance dimension. For regulated industry workloads, the repatriation case is often stronger on compliance grounds than on pure cost grounds.

Healthcare workloads running protected health information face the recently finalized 2026 HIPAA Security Rule update that strengthens BAA requirements, mandates network segmentation, and requires comprehensive encryption. Cloud configurations meeting these requirements are achievable but require careful attention to specific service configurations and may not include all cloud services the organization currently uses. Dedicated colocation with HIPAA BAA coverage provides cleaner compliance posture for workloads where the entire deployment can be HIPAA-compliant rather than specific services within a broader cloud environment. See our HIPAA colocation guide for the complete framework.

Financial services workloads face increasingly specific regulatory requirements from FINRA, the SEC, and state financial regulators. Cloud configurations supporting these requirements exist but require ongoing attention to compliance scope as cloud services evolve. Dedicated colocation provides more stable compliance posture for workloads that benefit from infrastructure consistency.

Government and defense workloads frequently require FedRAMP authorization, FISMA compliance, or other government-specific certifications. The cloud configurations meeting these requirements are narrower subsets of cloud capability than commercial deployments. Dedicated colocation at FedRAMP-authorized facilities provides full government workload capability without scope limitations.

Life sciences research workloads processing clinical trial data, genomic data, or research subject data face HIPAA-equivalent requirements alongside research-specific data governance requirements. Compliance scope for these workloads is frequently easier to manage in dedicated infrastructure than in cloud configurations.

The compliance dimension matters because compliance-driven repatriation produces value beyond pure cost optimization. Organizations that repatriate compliance-bound workloads to dedicated infrastructure frequently improve audit posture, simplify compliance documentation, and reduce ongoing compliance overhead. The financial savings on the cost dimension are reinforced by compliance benefits that have value independent of infrastructure pricing.

The Hybrid Cloud Architecture Outcome

The strategic outcome for most enterprises evaluating cloud repatriation is not complete cloud exit but hybrid architecture spanning dedicated infrastructure and cloud environments. The hybrid model captures the financial benefits of repatriating workloads that fit dedicated infrastructure while maintaining cloud presence for workloads that genuinely benefit from cloud elasticity or cloud-native services.

Successful hybrid architectures share several specific characteristics:

• Workload placement decisions follow utilization economics rather than ideology. Stable workloads run in dedicated colocation. Variable workloads run in cloud. The placement decision reflects workload requirements rather than universal preferences.
• Connectivity between dedicated and cloud environments enables seamless data movement. ServiceFabric, Open Cloud Exchange, and similar interconnection platforms provide low-latency private connectivity between colocation facilities and cloud providers. The connectivity eliminates the cloud egress cost dimension that often dominates cloud cost analysis for hybrid workloads.
• The interconnection ecosystem at Equinix data center facilities provides particularly strong hybrid cloud connectivity through their dense cloud onramp infrastructure. DataBank’s national network provides similar hybrid capability with stronger compliance positioning for regulated workloads — see our DataBank NYC guide for the complete capability analysis. CoreSite NY3 provides additional hybrid cloud positioning with ServiceFabric integration for workloads with significant cloud connectivity requirements. Digital Realty 60 Hudson supports Manhattan-required hybrid deployments with carrier hotel connectivity. Cologix Parsippany NJ offers cost-optimized hybrid infrastructure for disaster recovery and secondary workloads. Whichever provider you select, the carrier neutral data center architecture of major facilities provides the underlying connectivity foundation that hybrid architectures require.
• Operational tooling spans both environments without significant duplication. Monitoring, security, identity management, and deployment automation work across both dedicated and cloud environments rather than requiring separate operational tooling for each. The operational consistency reduces hybrid complexity to manageable levels.
• Compliance posture is maintained consistently across the hybrid architecture. Compliance certifications, security controls, and audit documentation apply to both environments with appropriate scope. Hybrid architecture introduces compliance complexity that requires structured management.
• Migration patterns work in both directions. Hybrid architectures support moving workloads from cloud to dedicated infrastructure when economics or compliance favor that direction and from dedicated infrastructure back to cloud when business requirements change. The flexibility to move workloads in both directions matters as business requirements evolve.

For most enterprises, the cloud repatriation conversation produces hybrid architecture rather than complete cloud exit. The specific workloads that repatriate vary by organization but the hybrid outcome is consistent across most enterprise repatriation initiatives.

Key Questions Enterprise CIOs Are Asking About Cloud Repatriation

At what workload size does cloud repatriation start to save money?

The crossover point depends on workload type and utilization pattern more than absolute size. For stable production workloads running 60-80 percent sustained utilization, repatriation typically saves money at any meaningful scale starting at approximately $200,000 annual cloud spend. For storage-heavy workloads with significant data egress, the crossover point can occur at $100,000 annual cloud spend. For variable workloads with high traffic variance, repatriation rarely saves money regardless of scale. The honest answer for most enterprises is that some workloads cross the threshold and others do not — portfolio analysis is more useful than universal scale thresholds. Metro Colo Advisory evaluates the specific workload portfolio for your situation at no cost.

Should I repatriate all workloads or only specific ones?

For most enterprises, partial repatriation produces better outcomes than complete cloud exit. The honest answer requires evaluating each workload category against the repatriation criteria. Stable production workloads with predictable utilization typically repatriate cost-effectively. Variable workloads with high demand variance typically remain in cloud. Workloads dependent on cloud-native services without dedicated equivalents face re-architecture costs that may not justify repatriation. The most common enterprise outcome is hybrid architecture where 30-60 percent of cloud workloads move to dedicated infrastructure while the remainder continues in cloud. Metro Colo Advisory provides workload-by-workload repatriation analysis at no cost.

How long does cloud repatriation typically take?

Partial repatriation timelines typically run 6-12 months from initial portfolio analysis through final workload migration. Complete cloud exit initiatives run 12-18 months for mid-market organizations and longer for enterprise organizations with significant cloud-native dependencies. The timeline is meaningfully longer than initial cloud migrations because workloads have developed operational dependencies during cloud deployment that require careful unwinding. The longest phases are typically workload migration in waves (12-24 weeks) and hybrid architecture optimization for organizations building hybrid models (8-16 weeks). Metro Colo Advisory manages repatriation timeline optimization for your specific workload portfolio at no cost.

What hidden costs do cloud repatriation projections often miss?

Three categories of hidden costs frequently affect repatriation economics. First, re-architecture costs for workloads dependent on cloud-native services without dedicated equivalents — these can be substantial for workloads built around serverless compute, managed AI services, or specific cloud integration platforms. Second, operational tooling investment for organizations that have built infrastructure tooling around cloud-native operational patterns — replacing or adapting tooling for dedicated infrastructure carries real cost. Third, parallel running costs during the migration period when workloads operate on both cloud and dedicated infrastructure simultaneously — these costs are temporary but meaningful during the migration timeline. Honest repatriation projections include these dimensions rather than focusing exclusively on steady-state savings. Metro Colo Advisory builds complete repatriation TCO models including transition costs at no cost.

What is cloud repatriation and why is it happening now?

Cloud repatriation is the deliberate movement of workloads from public cloud providers (AWS, Azure, GCP) back to dedicated infrastructure, typically professional colocation environments. The trend accelerated significantly between 2023 and 2025 as enterprise cloud bills exceeded original projections by 30 to 60 percent and the elasticity premium that justified cloud pricing during variable workload phases became expensive overhead during stable production phases. Major public cases including 37signals saving over $10 million across 5 years drove broader enterprise willingness to reevaluate cloud economics honestly. Industry analysts including IDC research now document repatriation as a mainstream strategic option rather than a contrarian position. The underlying economics have always favored dedicated infrastructure at high utilization — what changed is enterprise willingness to act on the math. Metro Colo Advisory evaluates cloud repatriation economics for your specific workload portfolio at no cost.

Which workloads benefit most from cloud repatriation?

Stable production workloads with sustained 50 to 85 percent utilization represent the clearest economic case for cloud repatriation, typically delivering 50 to 65 percent TCO reduction over a 5-year horizon. Database and storage-heavy workloads with significant persistent storage and high I/O throughput represent the second strongest case, with 55 to 70 percent TCO reduction common. Predictable batch processing workloads benefit at 30 to 50 percent TCO reduction. AI inference workloads at production scale benefit at 45 to 60 percent TCO reduction. Compliance-bound regulated workloads typically benefit on both cost and compliance dimensions. Variable customer-facing applications with high traffic variance, experimental research workloads, and short-duration training jobs generally stay better in cloud where elasticity is genuinely needed. The portfolio analysis approach evaluating each workload category against its utilization pattern produces materially better outcomes than treating the decision as universal. Metro Colo Advisory provides workload-by-workload repatriation analysis at no cost.

Can I run a hybrid cloud and colocation architecture?

Yes, and hybrid cloud colocation architecture is the most common strategic outcome for enterprise cloud repatriation initiatives. Hybrid architecture combines dedicated colocation infrastructure for workloads that fit dedicated infrastructure economics with cloud presence for workloads that genuinely benefit from cloud elasticity or cloud-native services. Successful hybrid architectures share several characteristics including workload placement decisions following utilization economics rather than ideology, low-latency private connectivity between colocation facilities and cloud providers (eliminating cloud egress costs), operational tooling that spans both environments without significant duplication, and consistent compliance posture across both environments. The interconnection ecosystem at Equinix data center facilities provides particularly strong hybrid cloud connectivity through their dense cloud onramp infrastructure. CoreSite NY3 provides similar hybrid cloud positioning with ServiceFabric integration. For most enterprises, hybrid cloud colocation produces better outcomes than complete cloud exit by capturing repatriation savings while maintaining cloud capability where genuinely needed. Metro Colo Advisory designs hybrid architectures matching your specific workload portfolio at no cost.

How much can I save by moving from AWS or Azure to colocation?

Repatriation savings vary significantly by workload type and utilization profile, but realistic ranges exist. For stable production workloads at 60 to 85 percent utilization, dedicated colocation typically delivers 40 to 65 percent total cost of ownership reduction over a 5-year horizon. For database and storage-heavy workloads with significant data egress patterns, savings frequently reach 55 to 70 percent TCO reduction because eliminating egress fees compounds with infrastructure cost savings. For typical mid-market workload profiles (100 vCPU equivalent with 50TB storage at stable utilization), 5-year savings typically run $575,000 to $810,000 versus continued cloud spend. Larger deployments produce larger absolute savings. Smaller deployments or workloads with high variance produce smaller savings or favor remaining in cloud. The honest analysis requires workload-specific TCO modeling rather than universal claims. Metro Colo Advisory builds complete repatriation TCO models comparing cloud spend against dedicated colocation across realistic 5-year scenarios at no cost.

What about cloud-native services that don’t exist in dedicated infrastructure?

Cloud-native services without dedicated infrastructure equivalents represent the most legitimate barrier to cloud repatriation and require careful evaluation. Serverless compute platforms, managed AI services, specialized analytics platforms, and certain integration services exist primarily as cloud-native offerings. Workloads heavily dependent on these services face significant re-architecture costs under repatriation that may not justify the cost savings. The honest approach is workload-by-workload evaluation rather than ideological commitment to complete cloud exit. Workloads with limited cloud-native service dependency repatriate cleanly. Workloads with heavy cloud-native dependency typically stay in cloud or get re-architected only when business requirements independently justify the engineering effort. The hybrid cloud colocation outcome captures the value of repatriating workloads that fit dedicated infrastructure while keeping cloud-native dependent workloads in cloud. Metro Colo Advisory identifies cloud-native service dependencies in workload portfolio analysis to determine realistic repatriation candidates at no cost.

How Independent Advisory Changes Cloud Repatriation Outcomes

Cloud repatriation analysis involves significant complexity that benefits substantially from independent expertise. The combination of workload-specific utilization analysis, total cost of ownership modeling, facility capability evaluation, compliance posture analysis, and migration planning creates a decision matrix that most internal IT teams have not built before.

Cloud provider sales teams will present cloud economics favorably regardless of whether specific workloads genuinely belong in cloud. Their commercial incentive is preserving and growing cloud spend rather than identifying workloads that would benefit from repatriation. Internal IT teams building repatriation analysis face the dual challenge of learning the analytical framework while maintaining current operational responsibilities.

Metro Colo Advisory is an independent colocation broker. We work for enterprises, not for cloud providers or colocation providers. Think of us the way you would think of a buyer’s agent in real estate. Our commission comes from the colocation provider you choose if and when you repatriate workloads, paid only when a deal closes. There is no cost to you. We have no financial stake in whether you repatriate workloads or stay in cloud — our only interest is the honest analysis of which workloads belong where.

Metro Colo Advisory serves enterprise clients nationally across all major US colocation markets including NYC metro, Chicago, Dallas, Atlanta, Phoenix, Northern Virginia, Silicon Valley, and other regional markets. Our independent cloud repatriation analysis applies equally across geographies — workload portfolio evaluation, TCO modeling, facility selection guidance, and migration planning expertise transfer cleanly between markets regardless of where your repatriated workloads will be deployed.

For cloud repatriation evaluations specifically we provide:

• Workload portfolio analysis identifying which specific workloads are repatriation candidates versus which should remain in cloud, with detailed utilization economics for each workload category.
• Total cost of ownership modeling comparing current cloud spend against dedicated colocation across realistic 5-year scenarios with hardware refresh cycles and capacity expansion provisions.
• Compliance posture evaluation for regulated industry workloads where dedicated colocation provides compliance benefits beyond cost savings.
• Facility selection guidance evaluating which providers and specific facilities best match your repatriated workload requirements including hybrid cloud connectivity for organizations building hybrid architectures.
• Migration planning support including workload-by-workload migration sequencing, parallel running coordination, and cloud termination procedures.
• Hybrid architecture design for enterprises building hybrid cloud colocation models rather than complete cloud exit.

For complete depth on related infrastructure decisions, see our cloud repatriation guide for the broader strategic framework, our high density colocation guide for AI workload repatriation specifics, our colocation pricing guide for current market pricing benchmarks, our colocation contract guide for contract term frameworks that apply to repatriation infrastructure, our HIPAA colocation guide for healthcare workload repatriation requirements, and our compliance colocation guide for the complete regulatory framework. For complete analysis of the NYC market context for repatriation deployments, see our NYC metro colocation market guide.

Get My Free Cloud Repatriation Analysis →

The cloud repatriation decision will define your infrastructure economics for the next 5-7 years. Cloud spend that feels manageable today compounds significantly as deployments scale and data movement patterns mature. Dedicated colocation properly evaluated for the right workloads delivers materially better economics. Get independent guidance before committing to either path.

Want to understand how Metro Colo Advisory works before filling out the assessment? See how Metro Colo Advisory works →