Always-on compute
Servers that run at the same level every hour of the month pay for flexibility they never use.
Most companies don’t need to leave AWS, Azure or Google Cloud. But steady, always-on workloads, heavy data transfer and GPU inference can cost less in a colocation building with a direct connection to your cloud. In four weeks we show you which workloads, what each path costs, and when staying put is the better answer.
Servers that run at the same level every hour of the month pay for flexibility they never use.
Moving data out of the cloud is often the line on the bill that grows fastest.
Inference that runs most of the day behaves like a steady workload, not a burst.
If your VMware terms changed, the question of where those hosts should live is open again.
A cloud-versus-colo comparison is easy to rig either way. Ours includes the costs on both sides.
We compare against the cloud with the discounts you already have or could get, such as savings plans, reservations and committed use discounts, not against list price.
Colocation is priced from current provider quotes for your footprint, not from averages.
Hardware, migration and exit costs come from your team’s or partners’ quotes and are shown separately.
We only consider buildings with a direct connection to your cloud. See the New York market guide for on-ramps in our home market.
Who runs the servers after the move, and what that costs, is part of the answer.
One lead runs the review from the first bill to the recommendation.
Three to six months of billing and usage exports, the discounts in place, and the VMware inventory. Read-only.
Steady, variable, data-heavy, GPU, regulated or latency-sensitive.
Three-year cost of staying in the cloud and of a hybrid setup with the steady part in colocation next to it.
What stays, what moves and in what order, a colocation requirement ready to go to market, and partners for the move.
No. Most companies that move workloads move only some of them. IDC said in 2024 that only 8 to 9% of companies plan to move everything off public cloud. The review shows which workloads, if any, cost less in colocation with a direct connection to your cloud; the rest stays where it is.
Compute that runs at a steady level around the clock, workloads that move a lot of data out of the cloud, GPU inference that runs most of the day, and VMware estates facing new license terms. Spiky, seasonal or fast-changing workloads usually stay in the cloud.
Read-only access to three to six months of billing and usage exports (for example the AWS Cost and Usage Report, an Azure cost export or a Google Cloud billing export) and, for VMware, the host and license inventory. We never change your cloud accounts.
No. Bargun reviews and recommends, then finds and contracts the colocation space and writes the build spec if you go ahead. Servers, migration and day-to-day operations are done by your team or by partners you choose; we can introduce some.
A fixed fee, agreed before we start and based on the size of your estate. If you later contract colocation through Bargun, any commission we earn from a provider is disclosed to you in writing before you sign.
Tell us which clouds you use, roughly what runs all day, and where your team is. We’ll tell you whether a review is worth it, and if it isn’t, we’ll say so.
Figures as of October 2026, each with its source and date.
AWS, Azure, Google Cloud and VMware are trademarks of their owners. Bargun is independent of them.