A budget line that grows because the price went up has not grown. It has only become more expensive. That distinction is about to matter more than usual, because the headline number for 2027 IT spending is almost entirely a price story, and most budget templates have no column that shows it.
On October 21, Gartner forecast worldwide IT spending of $7.35 trillion in 2027, up 15% from roughly $6.39 trillion this year. It is the first time the total has crossed $7 trillion. Read the release closely, though, and the growth is not coming from companies buying more. It is coming from companies paying more for the same things, driven by what Gartner analyst John-David Lovelock calls “memflation.”
For a finance or IT leader building next year’s budget right now, that changes the planning question. The question is not “how much more do we need?” It is “how much of our increase is already spoken for before we fund a single new project?”
Gartner’s 15% is mostly a price story
Here is how Gartner’s forecast breaks down by segment, in billions of US dollars:
| Segment | 2026 spend | 2026 growth | 2027 spend | 2027 growth |
|---|---|---|---|---|
| Data center systems | 879 | 73.8% | 1,366 | 55.5% |
| Software | 1,464 | 15.2% | 1,679 | 14.6% |
| IT services | 1,557 | 4.4% | 1,633 | 4.9% |
| Communications services | 1,352 | 4.1% | 1,391 | 2.9% |
| Devices | 847 | 7.2% | 916 | 8.1% |
| IaaS (subset of the above) | 292 | 31.3% | 367 | 25.7% |
| Total IT | 6,391 | 14.5% | 7,352 | 15.0% |
Source: Gartner, October 2026.
Data center systems is the line doing the work: up 73.8% this year and another 55.5% next year. Part of that is genuine hyperscaler and AI buildout. But Gartner is explicit about the other part. Memory prices more than doubled over the past twelve months, and the increase has spread through storage, servers, networking, PCs and devices. Lovelock expects the “memflation” effect “to persist through 2027 before hardware pricing begins to stabilize.”
The sentence that should get a CFO’s attention is a quieter one. Higher average selling prices “continue to support spending growth,” Gartner says, but “the resulting pressure on affordability is expected to suppress end-user demand.” Translated: the forecaster expects buyers to spend more dollars on fewer units. That is the definition of a price-driven increase.
The memory data backs it up. TrendForce reported on October 20 that DRAM contract prices are expected to rise about 20% quarter on quarter in Q4 2026, after roughly 80% in Q2 and 40% in Q3. The pace is slowing, but slowing inflation is still inflation, and it compounds on a base that has already more than doubled. TrendForce’s September 30 outlook put conventional DRAM at 10% to 15% for the quarter and NAND flash at 15% to 20%, and said server DRAM will “remain undersupplied in 4Q26.”
The two numbers that don’t reconcile
Set Gartner’s market forecast beside Gartner’s own CIO survey and the gap is hard to miss. The 2027 CIO and Technology Executive Survey projects IT budgets will grow by an average of just 3.7% next year. Gartner adds that after inflation, many organizations will see only marginal gains in available funding.
So the market grows 15% and the typical enterprise budget grows 3.7%. Both can be true at once, because the market number is concentrated in hyperscalers and AI infrastructure buyers, while the budget number describes everyone else. But for the “everyone else” group, the implication is uncomfortable: suppliers are raising prices at multiples of the rate your budget is growing.
The same survey adds a third number that makes the squeeze worse. Funding for agentic AI is expected to rise by an average of 31.8%, the fastest-growing investment area in the survey. A budget growing 3.7% overall that grows one line by 31.8% has to shrink something else. With hardware and software prices rising, the “something else” is usually whatever was planned at last year’s unit prices.
Where memflation lands in an IT budget
Memory inflation does not show up evenly. It hits some lines directly, some indirectly and some barely at all, and treating it as a blanket percentage is how budgets end up wrong in both directions.
End-user devices
This is the most direct hit. In February, Gartner estimated that DRAM and SSD costs would rise 130% by the end of 2026, enough to add 17% to PC sticker prices. Dell raised server prices on December 10 and, according to COO Jeff Clarke, “had tens of thousands of open quotes in the PC business and changed them all on Jan. 6.” If your 2027 plan includes a laptop refresh priced off a 2025 quote, that line is understated before anyone touches it.
On-premises servers and storage
The second direct hit, and often the larger one in dollar terms. Server configurations are memory-heavy, and enterprise SSD pricing is accelerating in Q4 according to TrendForce. Clarke’s other remarks in the same CIO Dive interview are the ones procurement teams should pin to the wall: “We’re compressing discounting” and “Our quotes are valid for the shortest period of time they’ve ever been.” A server refresh budgeted in September may not survive to a purchase order in February at the same price.
Cloud infrastructure
Less direct, and this is where teams tend to overcorrect. Hyperscalers have so far absorbed most memory inflation in their on-demand compute pricing, but not all of it. AWS has raised EC2 Capacity Block prices for GPU instances repeatedly this year, and Microsoft is raising Azure VM prices by up to 17% in seven regions on February 1. The broader pattern, covered in our look at how the memory shortage reaches regular cloud bills, is that increases arrive through specific regions, legacy instance series and capacity products rather than as a list-price headline. Budget cloud for selective increases tied to where your workloads actually run, not a flat uplift.
Software and SaaS
Software growth of 14.6% is not memory-driven, but it behaves the same way in a budget: the price rises while the footprint stays flat. Our analysis of the latest SaaS benchmark data found that portfolios stopped growing while spend kept climbing, driven by renewal increases and metered AI charges. For budgeting purposes, software belongs in the same “price, not volume” bucket as hardware.
Split every 2027 line into price and volume
The fix is not a new tool. It is a column. Every material budget line for 2027 should show the change from 2026 as two separate numbers: what changed because unit prices moved, and what changed because you are buying more or less. Operational finance teams in manufacturing and telecom have done this for decades as price and volume variance analysis; IT budgets mostly skipped it because, for most of the last two decades, unit prices in technology fell on their own.
Here is an illustrative $10 million IT budget built that way. The percentages are planning assumptions drawn from the sources above, not benchmarks; substitute your own quotes and contracts.
| Line | 2026 base ($K) | Price assumption | Price effect ($K) | Volume assumption | Volume effect ($K) | 2027 ($K) |
|---|---|---|---|---|---|---|
| End-user devices | 1,200 | +17% | +204 | flat | 0 | 1,404 |
| On-prem servers and storage | 1,000 | +20% | +200 | flat | 0 | 1,200 |
| Cloud infrastructure | 3,000 | +2% | +60 | +12% | +360 | 3,420 |
| SaaS and software | 2,800 | +7% | +196 | flat | 0 | 2,996 |
| AI services and APIs | 500 | −25% | −125 | +60% | +300 | 675 |
| IT services and contractors | 1,500 | +3% | +45 | flat | 0 | 1,545 |
| Total | 10,000 | +580 | +660 | 11,240 |
(The cross term, price change multiplied by volume change, is left out for readability. It is small at these rates.)
Three things fall out of this table that a single “2027 budget” column would hide.
Price alone consumes the raise. Price effects add $580K, or 5.8% of the base. A budget that grows at Gartner’s 3.7% average gives you $370K. You are $210K short before funding any new demand, including the AI growth leadership has probably already promised.
The AI line moves opposite to everything else. Token prices are falling while usage climbs, so the AI line has a negative price effect and a large volume effect. That is the one place where cost control is a volume conversation (routing, caching, which workloads run at all), and it is also the line most likely to be forecast badly, for reasons we covered in why better AI optimization makes forecasts less reliable.
The requested total is $11.24 million, a 12.4% increase. Presented as one number, that looks like an IT department asking for three times the average budget growth. Presented as $580K of supplier pricing and $660K of chosen growth, it becomes a conversation about which growth to fund, which is the conversation finance actually wants to have.
Buying hardware while quotes expire in days
Once price is visible as its own line, the procurement decisions get clearer.
Decide refresh timing on purpose. If Gartner is right that memory pricing stabilizes after 2027, a one-year extension on a four-year laptop refresh cycle defers a purchase into a cheaper market. That only works if the fleet can tolerate it, and it trades a capital saving for higher support and failure costs. Model both before deciding.
Pull forward only what is already approved and specified. With DRAM still rising quarter on quarter, buying approved server capacity in Q4 rather than Q1 can beat the next round of increases. Buying unapproved capacity early to “lock in” pricing is how a hardware budget turns into a stranded-asset write-down.
Change how you evaluate quotes. Short validity windows and compressed discounts mean the old habit of collecting three quotes over six weeks no longer produces comparable numbers. Get quotes in the same week, ask for price protection terms explicitly, and treat any multi-month validity a vendor offers as a concession worth trading for.
Reconsider the on-prem versus cloud line where it is close. Hyperscalers buy memory at a scale and contract structure that most enterprises cannot match, which is part of why on-demand compute has not repriced the way servers have. For workloads that were marginal candidates either way, the 2027 hardware premium is a legitimate input to that decision. It is not a reason to reverse a repatriation case that was built on sound math.
What to bring to the CFO in November
Budget season for calendar-year companies is now. A few practical moves:
Lead with Gartner’s two numbers side by side, 15% market growth and 3.7% budget growth, because they frame the problem in someone else’s words. Then show the price and volume split for your own lines. A CFO who sees that supplier pricing alone consumes the planned increase will either fund the gap or tell you which growth to cut. Both are better outcomes than a budget that quietly misses by 6% next June.
Tie each price assumption to evidence: the quote date, the contract renewal clause or the published price change. Price assumptions without evidence get negotiated down in budget review; price assumptions with a TrendForce reference and a dated vendor quote do not.
Finally, set a checkpoint. Memory pricing is moving every quarter, and Gartner itself expects the pressure to ease only after 2027. Revisit the price column in Q2, when actual hardware quotes and the first round of 2027 renewals will show whether your assumptions were too high or too low. Separating price from volume is useful at budget time, but it pays off most in the mid-year reforecast, when the variance comes with an explanation already attached.