IT Infrastructure Consulting
Where Each Workload Should Run, the Cost Model Behind the Decision, and What a Fixed-Fee Engagement Delivers
Steady workloads typically cost 30 to 50 percent less on dedicated capacity, and bursty ones cost less in the cloud. Our engagements draw that line workload by workload, independent of anyone who would run the result.
Compare My Infrastructure CostsThe usual way to buy infrastructure advice is to ask the people who will run the result, and the answer tends to match what they sell. This page describes a different kind of engagement, written by an advisory practice that does not resell cloud capacity, hardware or managed services, prices the options on every side of the decision, and is paid the same engagement fee whatever the answer turns out to be. If the answer is to lease capacity, the provider chosen pays the placement fee, and that is disclosed before work starts.
Sources: Flexera 2026 State of the Cloud (waste); CBRE H1 2026 (vacancy); the two savings ranges are the Metro Colo Advisory market read from our cloud repatriation work, relative positions rather than quotes, and the repatriation range is consistent with a16z, which put repatriated workloads at one-third to one-half the cost of equivalent cloud.
What a Placement Engagement Does, Week by Week
An engagement runs from your billing exports and a short questionnaire to a readout in three to six weeks. The sort comes before the pricing, because pricing a workload that should not move wastes everyone’s time.
Metro Colo Advisory engagement timeline for a mid-size enterprise footprint; multi-region or multi-cloud footprints run toward the longer end. Figures as of October 2026. Free to cite with a link to this page.
Annual cloud and infrastructure spend, the main workloads and roughly how steady each one is, and the decision in front of you.
A reply from the person who would run the engagement, with a first read on whether a placement change is worth modeling.
If your infrastructure is already where it should be, we will tell you so, and it will have cost you one email.
The first read is free. A full engagement is a fixed fee agreed before work starts. If you later lease capacity, the provider you choose pays the placement fee.
Which Workloads Belong Where
Placement is decided by how a workload behaves, not by what it is called. Steady, egress-heavy, latency-sensitive work with few managed-service dependencies moves well; bursty work built on a cloud provider’s own services usually stays. This is the first pass we run on every engagement, before any pricing.
| Workload profile | Usual best home | Why |
|---|---|---|
| Steady compute, high utilizationCore application tiers, databases, serving | Colocation on owned hardware, or bare metal | Paying on-demand or even committed cloud rates for capacity that runs flat out all year is where the 30 to 50 percent comes from. |
| Egress-heavyContent delivery, data serving, ad tech, APIs | Dedicated, with flat-rate transit | Cloud egress is priced per gigabyte and is not covered by compute commitments; a dedicated port is a flat monthly fee. See the bare metal guide. |
| Latency-sensitiveTrading, real-time bidding, gaming | Colocation near the counterparty or exchange | Distance sets latency, and the right building beats any instance type. |
| AI trainingLarge clusters, weeks-long jobs | Owned GPUs in high-density colocation, or reserved GPU rental | Utilization decides it: at high, sustained utilization owning wins over three years; at low utilization renting does. See AI and GPU colocation and GPU rental. |
| AI inferenceServing models in production | Dedicated, close to the application | Steady, latency-sensitive and egress-heavy at once, which is the profile that moves best. |
| Bursty or seasonalBatch jobs, campaigns, peak events | Public cloud, on spot or on-demand | Paying only for the hours used beats owning capacity that sits idle most of the year. |
| Built on managed servicesManaged databases, queues, data warehouses, serverless | Public cloud, usually | The service is the lock-in, not the compute. Re-platforming it is a project in its own right and is costed separately when it is worth doing at all. |
| Dev, test and experimentationShort-lived environments | Public cloud | Speed of provisioning matters more than unit cost. |
| Regulated or sovereign dataHealth, financial, government, in-country rules | Colocation or private cloud in the required jurisdiction | Control of the hardware and the location is the requirement. See compliance colocation and managed private cloud. |
What Our IT Infrastructure Consulting Services Include
Every engagement is scoped to the decision in front of you, but the work draws on the same six pieces. Each one is there to make the final number defensible to a board, a lender or the other side of a transaction.
Workload placement analysis
Every workload sorted by what it does, what it depends on and how steadily it runs, into stay, move and move-later. This step is often skipped, and it is where the real answer lives.
The three-to-five-year cost model
Cloud as optimized, bare metal, colocation on owned hardware and hybrid, built from your billing exports and current market rates rather than list prices, with the inputs shown so your finance team can change them.
Migration cost and sequencing
What it takes to move each workload, in people, time and one-time cost, and the order that keeps production safe. A saving that costs two years of engineering to collect is not a saving.
Network and connectivity design
How the halves of a hybrid footprint connect, what private interconnect to the cloud costs, where the on-ramps are, and what the egress bill becomes. See data center connectivity.
Location and capacity
Which markets and facilities can take the dedicated footprint on your date, at your density, with the latency your users need. See data center site selection.
Hardware and platform mapping
What the dedicated side runs on, server class by server class, including the x86 and Arm question for workloads already built to run on both, and what it costs at list and at volume.
What We Do Not Do, and Why It Matters
Three things are deliberately absent from our engagements, and each one removes a reason to prefer one answer over another.
| Not included | Why it is left out | What you get instead |
|---|---|---|
| Managed services | An advisor who will also run your infrastructure has a reason to recommend infrastructure that needs running. | A decision you can hand to any operator, your own team or a managed provider you choose. |
| Hardware and cloud resale | A margin on servers or cloud credits is a reason to recommend servers or cloud credits. | Hardware priced at list and at the volume rates we see, with no markup and no preferred vendor. |
| A success fee on the saving | A fee tied to the size of the saving is a reason to overstate it. | An engagement fee fixed before work starts, the same whether the answer is move, stay or wait. |
One thing is paid by someone other than you: if the engagement ends in leasing colocation or bare metal capacity, the provider you choose pays our placement fee, as it would pay any broker. That is disclosed before work starts, and the engagement fee does not change with it.
What the Cost Model Starts From
The comparison is always between optimized cloud and the dedicated options, never between list-price cloud and anything, because comparing against list is how repatriation cases get overstated. The cloud repatriation guide compares the four paths, bare metal, colocation on owned hardware, optimized cloud and hybrid, with the egress arithmetic and the exit waivers; the colocation pricing guide and data center cost carry the market figures by metro. The model takes those as starting points and replaces them with your billing export and the rates we see in the market that month. The colocation side of the arithmetic looks like this.
When to Bring in an IT Infrastructure Consultant
Optimization has been done, the easy savings are taken, and the next round finds less.
The answer has to be written, defensible and independent of the people who run the systems.
A cloud enterprise agreement or a colocation lease, where the choice is to renew or to move and the model prices both.
GPUs at scale change the economics of owning versus renting and the density any facility has to support.
Everything in it has to go somewhere, and the choice is between cloud, colocation and bare metal workload by workload.
Two of everything, in two places, on two contracts.
How an IT Infrastructure Consulting Engagement Runs
The free workload review
You send your spend and workloads. Within 24 hours you hear back from the person who would run the engagement.
Scope and engagement letter
The question, the deliverables, the timeline and the fixed fee, in writing. Work starts on receipt of your billing exports and the questionnaire.
Baseline and placement
Your current spend by workload and cost type, then every workload sorted into stay, move and move-later with the reasoning for each.
The model and the report
Each scenario costed over the term with migration cost included, delivered as a report, a working spreadsheet and an executive summary, then a period of written questions and answers.
Placement, if you move
If the answer is to lease capacity, we run the data center RFP and negotiate the contract. The provider you choose pays for that step.
If the decision is made and the question is where to put the capacity, skip the engagement: send the requirement and we run the search and the negotiation. The provider you choose pays us, so it costs you nothing.
Six Infrastructure Decisions That Go Wrong
Comparing against list-price cloud
The case has to beat the bill after commitments, rightsizing and a negotiated agreement. Compared against list, almost anything looks cheaper. See cloud repatriation.
Moving everything, or nothing
The answer is almost always a line through the footprint. Re-platforming the managed-service layer takes years; leaving steady capacity on cloud rates costs every month.
Forgetting egress and the people
Data leaving the cloud is priced per gigabyte, and dedicated capacity needs engineers to run it. Both belong in the model at a loaded cost.
Asking the vendor
A cloud partner, a reseller and a managed provider each have a view, and it tends to match what they sell. Independent advice costs a fixed fee.
Deciding the where before the what
Picking a market or a facility before the workloads are sorted produces a deployment sized for the wrong footprint. Placement first, then capacity.
Treating the decision as final
Workloads that were bursty become steady as a product matures. The line through the footprint should be revisited as spend grows, which is why the model is handed over as a working spreadsheet.
We advise companies across the United States and Canada, with market guides for New York, Virginia, Texas, Chicago, Phoenix and Canada where the dedicated half of a footprint most often lands.
Related guides for the decision itself: cloud repatriation for the cloud-side arithmetic and the exit waivers, FinOps consulting for what to do when the bill is already optimized, VMware alternatives for the licensing question that often triggers the review, data center migration for the move itself, and wholesale colocation for footprints above 1 MW. Owners, developers and investors evaluating a site or a transaction are served by data center consulting and data center due diligence.
IT Infrastructure Consulting: Frequently Asked Questions
What is IT infrastructure consulting?
Outside advice on the systems a company runs on and where they should live: public cloud, private cloud, colocation on owned hardware, rented bare metal or a hybrid of them. A placement-focused engagement answers which workloads belong where, what each option costs over three to five years, and what it takes to move, delivered as a written report and a cost model.
What do IT infrastructure consulting services include?
Workload placement analysis, a three to five year cost model across cloud, bare metal, colocation and hybrid, migration cost and sequencing, network and connectivity design, location and capacity options, and hardware and platform mapping. Managed services, hardware resale and cloud resale are not included, which is what keeps the advice independent.
How much does IT infrastructure consulting cost?
Our engagements are a fixed fee, typically $50,000 to $195,000 depending on scope, agreed before work starts and the same whatever the answer. The first read, a free workload review, costs nothing. If the engagement ends in leasing capacity, the provider you choose pays the placement fee, which does not change the engagement fee.
How long does an infrastructure consulting engagement take?
Three to six weeks from receipt of billing exports and a short questionnaire, for a mid-size enterprise footprint. Larger or multi-region footprints take longer, and the free workload review that precedes it takes one email.
What is the difference between an IT infrastructure consultant and a managed services provider?
A consultant is paid for the decision; a managed services provider is paid to run the result. The two can be the same firm, but then the advice tends to point toward infrastructure that needs running. A consultant with no managed services, resale or success fee has no reason to prefer one answer.
Should we move off the cloud?
Partly, usually. For companies above about $50,000 a month, the steady third to half of the footprint typically saves 30 to 50 percent on dedicated capacity once every cost is counted, while bursty work and anything built on managed services stays. Those are market ranges, not quotes; the engagement draws the line workload by workload with your own numbers.
Who reads the report, and what happens after delivery?
It is written for the infrastructure leader who commissioned it and for finance and the board who will act on it: a report, a working model with its assumptions exposed, and an executive summary. A period of written questions follows delivery, and if the answer is to lease capacity, we run the search and negotiate the contract as a separate step paid by the provider.
What do you need from us to start an engagement?
Billing exports from each cloud and any colocation or hosting invoices, a short questionnaire on workloads, regions, managed-service dependencies, traffic and egress, and a blended engineering cost. The clock starts when those arrive. A mutual NDA covers everything before any data changes hands.
How does a hybrid cloud infrastructure work?
Steady workloads run on dedicated capacity, variable ones stay in the public cloud, and a private interconnect joins the two so the application behaves as one system. Most placement engagements land here. The hybrid cloud colocation guide covers the connection options and their costs.
Do you work with AI and GPU infrastructure?
Yes. The owning versus renting decision for GPUs, the density a facility has to support, and where rack-scale capacity exists are part of the placement analysis when AI workloads are in scope. See the AI and GPU colocation and GPU rental guides.
Can you prepare the analysis for a board or a transaction?
Yes. The report, the model with its assumptions exposed, and the executive summary are written to be read by finance and a board, and to stand up to questions from a lender or the other side of a transaction. Confidentiality is covered by a mutual NDA before any data changes hands.
How do we start?
Send your infrastructure spend and the main workloads. You will hear back within 24 hours from the person who would run the engagement. If a placement change is not worth modeling, you will be told so.
Start With the Free Workload Review
Send your annual infrastructure spend, the main workloads and how steady each one is, and the decision in front of you. You'll hear back within 24 hours from the person who would run the engagement. If your infrastructure is already where it belongs, we will say so.
Independent of every cloud, operator and vendor. A full engagement is a fixed fee agreed before work starts, and placement of any capacity you lease is paid by the provider.
The decision: cloud repatriation, FinOps consulting, hybrid cloud colocation, bare metal and managed private cloud.
The numbers: data center cost, the colocation pricing guide, the cloud vs colocation calculator and the data center lease guide.
AI and GPU: AI and GPU colocation, GPU rental and GPU cluster rental.
Placing the capacity: data center site selection, the data center RFP, how a data center broker works and disaster recovery colocation.
