Cloud bills grow quietly: forgotten test environments, oversized machines, idle databases, data transfer nobody planned for. We find these, fix them and put routines in place so they stay fixed. The other half of a lower bill is the price you pay per unit, and that is where commitments and discounts come in.
Understanding the bill
We start with the billing data. We go through the invoices and usage reports line by line and sort the cost by service, environment and team, using tagging, budgets and alerts so that each team can see what it spends. This usually shows what actually drives the bill.
Paying less per unit
Cloud vendors price the same capacity in several ways, and the right mix depends on how steady your usage is.
Reservations and commitments
For the part of your usage that does not move, a commitment is much cheaper than on-demand prices: reserved instances and Savings Plans on AWS, reservations and savings plans on Azure, and the equivalent on Huawei Cloud. We size them to your stable base, not to the peak, and check regularly that they are actually used, so that you do not pay for capacity you do not need.
Spot capacity
Workloads that can be interrupted, such as batch jobs, build agents and development environments, can run on spot capacity at a much lower price.
Volume discounts
Vendors give better prices as spend grows, but the big agreements are out of reach for many companies. If you buy through us as a reseller, you get discounts set for your volume.
Cloud billing
When you buy through us, the vendor invoices us and we invoice you. The reseller discount comes with that, and so do payment terms that give you more time to pay and some conveniences in how you pay. More on this under Cloud Reselling.
Using less
Before you commit to anything, you should only be paying for what you use. We remove unused resources, match machines and databases to the load they actually carry, and set up schedules that stop non-production environments at night and at weekends. Autoscaling follows demand instead of a worst-case guess, and lifecycle rules move old storage to cheaper tiers.
When the cloud is the wrong place
Sometimes the answer is that a workload does not belong in the cloud. We run a platform that serves more than 80 billion requests a day on our own bare-metal servers at OVHcloud, so we know what each option costs, and we will tell you when a move makes sense, in either direction.
On the platform we built in the Project Synergy case study, compute costs fell by up to 30%.