Share this:
IT capacity planning is the practice of forecasting how much compute, storage, and network capacity your workloads will need, and when, so you can buy or reclaim it before it runs out. For 2027, the job has changed: memory and flash prices are still rising and lead times are long, so a late forecast now costs real money, not just a rushed PO.
If you are building next year’s infrastructure budget right now, your capacity plan is the part of it that is most exposed to the hardware market. Getting it right is how you avoid both an outage and an emergency purchase at peak pricing.
Key takeaways
- Component prices are moving against buyers: TrendForce forecast conventional DRAM contract prices up 13 to 18 percent and NAND flash up 10 to 15 percent quarter over quarter in 3Q26, with enterprise SSD prices still climbing into 4Q26.
- Lead time is now a capacity metric. When large DRAM orders can take 40+ weeks, your forecast horizon has to be longer than your procurement cycle.
- The cheapest capacity is the capacity you already own. Reclaiming orphaned, idle, and overprovisioned resources should come before any new buy.
- Hybrid estates need one capacity model across on-prem and cloud, or you will double-count headroom in one place and miss exhaustion in another.
- A defensible plan ties every capacity request to a date, a workload, and a cost your CFO can check.
What is IT capacity planning, and why is it harder for 2027?
IT capacity planning answers three questions: what do we have, what will we need, and when will we cross the line. It is harder for 2027 because the cost and timing of the answer to “buy more” have both gotten worse.
For most of the last decade, capacity planning ran on a quiet assumption: hardware gets cheaper every year, so waiting is rarely expensive. That assumption no longer holds. Gartner’s July 2026 forecast projects data center systems spending to grow 62.5 percent in 2026, to $822 billion, and notes that technology budgets are strained by rising hardware and memory costs and supply constraints. AI buildouts at hyperscalers are absorbing the same memory and flash that your next storage array and server refresh depend on.
If you want the fundamentals first, our primer on what capacity planning is covers the basics. This post focuses on how an IT Director should run the process when the market is working against you.
Why are rising memory and flash prices a capacity planning problem?
Because they turn a timing mistake into a budget overrun. If you run out of headroom and have to buy in a hurry, you pay whatever the market is charging that quarter, and you may wait months to receive it.
TrendForce’s 3Q26 forecast called for conventional DRAM contract prices to rise 13 to 18 percent and NAND flash to rise 10 to 15 percent quarter over quarter, with server and AI demand holding prices up. In September, TrendForce said enterprise SSD prices will stay on an upward trend in 4Q26 as North American cloud providers raised their demand forecasts. On the supply side, server maker Inventec warned in July that lead times for large DRAM orders had passed 40 weeks.
Put those together and the practical effect is simple: a capacity plan that only looks one or two quarters out is now shorter than the time it can take to get hardware on the floor.
| Old planning assumption | 2027 reality | What to change |
|---|---|---|
| Hardware gets cheaper if you wait | Memory and flash prices are still rising quarter over quarter | Price your forecast at current quotes, not last year’s |
| Lead times are a few weeks | Large memory orders have been reported at 40+ weeks | Extend the forecast horizon past your longest lead time |
| Buy headroom, sort it out later | Every unused terabyte is paid for at peak pricing | Reclaim idle and orphaned capacity before buying |
| Cloud is the overflow valve | Cloud providers are the ones driving the demand | Model the cloud cost of burst, not just its availability |
| Vendor tools cover their own arrays | Most estates run several storage and compute vendors | Forecast from one normalized, multi-vendor view |
How do enterprises do IT capacity planning across hybrid infrastructure?
They put on-prem and cloud capacity into one model with one set of units, then forecast each resource pool against its own exhaustion date. The failure mode is planning each platform in its own console and adding up the answers in a spreadsheet.
In a hybrid estate, headroom in one place does not cancel exhaustion in another. Free space on a VMware cluster does not help a full IBM Power frame, and spare cloud quota does not help a latency-sensitive database that has to stay on-prem. A useful hybrid model does four things:
- Normalizes units across vendors (usable capacity, not raw; vCPU and memory per host, not just host counts).
- Tracks growth per pool, not as one company-wide average that hides the fast-growing workloads.
- Prices the capacity on-prem and in cloud, so you can compare the cost of adding a shelf with the cost of bursting a workload.
- Maps capacity to owners, so growth has a name attached. This is where showback or chargeback earns its keep.
What is the IT capacity planning process? A 6-step framework
The process is: inventory, measure, forecast, reclaim, price, and commit. Each step has a clear output you can show to someone else.
- Inventory everything you own. Every array, host, cluster, and cloud account, including the ones nobody has logged into lately. Output: one list, one set of units.
- Measure real utilization and growth. Use at least 90 days of history per pool, and flag seasonality (quarter-end, open enrollment, peak retail). Output: growth rate per pool.
- Forecast exhaustion dates, not percentages. “Cluster B hits 80 percent in March” is actionable. “We are at 64 percent” is not. Set your trigger threshold so the exhaustion date lands after your realistic lead time. Output: a dated exhaustion calendar.
- Reclaim before you buy. Orphaned volumes, powered-off VMs, stale snapshots, and oversized allocations are hidden costs that become free capacity once you find them. Output: reclaimable capacity, by pool.
- Price the remaining gap. Use current quotes and assume prices keep moving. Compare buy now, buy later, and burst to cloud for each gap. Output: a cost per option.
- Commit and review monthly. Lock the purchases that fall inside your lead time, and revisit the rest every month as actuals come in. Output: a rolling plan, not an annual guess.
This maps closely to the Forecasting capability in the FinOps Framework, which the FinOps Foundation describes as modeling the future cost and value of IT systems from historical patterns and planned changes. The difference for an IT Director is that the model has to include the hardware on your own floor, not only the cloud bill.
How do you defend an IT capacity plan to your CFO?
Show the date, the cost of waiting, and what you already reclaimed. CFOs rarely push back on a purchase that prevents a dated, specific outage. They push back on round numbers and “we need more headroom.”
A one-page capacity brief that holds up in a budget review includes:
- The exhaustion date for each pool that needs investment, with the growth trend behind it.
- The capacity you reclaimed first, and what it saved.
- The price and lead-time exposure of waiting a quarter, using current market data.
- Who owns the growth, so the cost lands with the business unit driving it.
If you are also ranking next year’s projects, our guide to prioritizing IT infrastructure projects for your 2027 budget pairs well with this: capacity work tends to score high on risk reduction and cost of delay.
Where Visual One Intelligence® fits
Visual One Intelligence® is a Hybrid FinOps™ and IT operations management platform that brings on-prem storage and compute into the same cost and capacity model as cloud. For capacity planning, that means multi-vendor storage and compute reporting in one normalized view, capacity forecasting by pool, and cost attribution through showback and chargeback. It is also recognized by the FinOps Foundation as a FinOps Certified Platform for hybrid infrastructure. The practical outcome: you can see which pools run out first, what you can reclaim, and what the rest will cost, before the budget is locked.
Frequently asked questions
What is IT capacity planning?
IT capacity planning is the process of forecasting the compute, storage, and network resources your workloads will need and when you will need them. It compares current utilization and growth trends against available capacity to produce exhaustion dates, so you can reclaim, buy, or shift workloads before performance suffers or an outage occurs. In hybrid estates it covers on-prem and cloud resources in one model.
How far ahead should an IT capacity plan look?
At minimum, longer than your longest procurement lead time plus installation time. With large memory orders reported at 40 weeks or more in 2026, many IT teams now plan 12 to 18 months out for on-prem hardware, while reviewing the plan monthly. Cloud capacity can be planned on shorter cycles, but its cost still needs a forecast.
What utilization threshold should trigger a purchase?
There is no single right number. Many teams use 70 to 80 percent of usable capacity as a trigger, but the better approach is to work backward from lead time: set the trigger so that, at your current growth rate, new capacity arrives before you hit the level where performance or resiliency degrades. Faster-growing pools need lower triggers.
What is the difference between capacity planning and capacity management?
Capacity management is the ongoing operational work of monitoring and tuning resources you already have. Capacity planning looks forward: it forecasts future demand and decides what to buy, reclaim, or move, and when. In practice they share the same data, and a good plan depends on accurate, current capacity management data across every vendor.
How do rising memory and SSD prices affect capacity planning?
They raise the cost of buying late. When component prices climb quarter over quarter, an emergency purchase costs more than a planned one, and long lead times can leave you short while you wait. The response is to forecast further out, reclaim unused capacity first, and lock purchases that fall inside your lead-time window.
See your hybrid capacity in one view
Want to know which pools run out first, and what you can reclaim before you buy? See a live demo of Visual One Intelligence® and get a capacity picture across your storage, compute, and cloud that you can take into your 2027 budget review.
