Free tools Windows power users keep installed
One-click scans. No signup required.
Some links on this page are affiliate links: if you buy through them we may earn a commission, at no extra cost to you.
Gartner’s 7.9% figure was a July 15, 2025 forecast for worldwide IT spending in calendar year 2025: $5.43 trillion, up from 2024. It was not a measure of infrastructure spending alone, and it is no longer Gartner’s latest outlook. On July 27, 2026, Gartner forecast 14.2% growth in worldwide IT spending for 2026, to $6.37 trillion. The through-line is rising investment in AI infrastructure, data-center systems and cloud capacity—but neither forecast means every company should raise its IT budget by the same amount.
What Gartner’s 7.9% forecast measured
In a July 15, 2025 release, Gartner forecast that worldwide end-user IT spending would total $5.43 trillion in 2025, a 7.9% increase over 2024. The estimate was denominated in U.S. dollars and covered data-center systems, devices, software, IT services and communications services.
That scope matters. The figure was not a prediction that every category—or every organization’s budget—would grow 7.9%. Nor was it an AI-spending or infrastructure-only total. It was a global forecast across a broad set of technology markets. Gartner said AI-related infrastructure, particularly data-center systems, was a major growth driver even as uncertainty weighed on some software and services decisions.
Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Repair Windows errors before they cause bigger problems3Fix the driver behind crashes, sound loss and screen glitchesThe 7.9% figure in context
Gartner’s forecasts are dated snapshots, not fixed outcomes. Its outlook for 2025 changed over time, and its estimates for 2026 rose through successive releases:
#1 Best Overall
| Forecast published | Year being forecast | Worldwide IT spending forecast | Growth forecast |
|---|---|---|---|
| Oct. 23, 2024 | 2025 | — | 9.3% |
| Jan. 21, 2025 | 2025 | — | 9.8% |
| July 15, 2025 | 2025 | $5.43 trillion | 7.9% |
| Feb. 3, 2026 | 2026 | $6.15 trillion | 10.8% |
| Apr. 22, 2026 | 2026 | $6.31 trillion | 13.5% |
| July 27, 2026 | 2026 | $6.37 trillion | 14.2% |
The 2025 estimates appeared in Gartner releases dated Oct. 23, 2024, Jan. 21, 2025 and July 15, 2025. The 2026 outlook was revised in releases dated Feb. 3, Apr. 22 and July 27, 2026. Those revisions reflect changing assumptions about demand, investment, supply and economic conditions; they do not turn earlier forecasts into reported results.
For a reader encountering the 7.9% headline now, the useful distinction is simple: it describes Gartner’s 2025 outlook at a particular point in time. The July 2026 forecast is the later outlook for a different calendar year—not a direct revision of the 2025 number.
Why AI spending shows up as infrastructure spending
AI systems need more than applications and subscriptions. They need compute, memory, fast connections between components, storage for data and models, and facilities capable of supplying power and removing heat. For large-scale AI, that often means dense accelerator-equipped servers, high-bandwidth memory and high-speed network fabrics, along with the software and operations needed to run them.
The Tool Desk
Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →Outbyte Driver Updater FREEScan for outdated or missing drivers - takes under a minuteDriver Scan →Investment also occurs at several points in the supply chain. Cloud providers and technology companies build capacity; businesses may then purchase access to that capacity as GPU instances, managed AI platforms or inference services. As a result, a company can use substantial AI infrastructure without buying or operating a server itself. Provider investment may ultimately be reflected in cloud charges or bundled software services.
Gartner’s July 2025 release forecast a major change in server spending. It said AI-optimized server spending, negligible in 2021, was expected to reach a scale that would make it roughly three times traditional-server spending by 2027. That was a forecast about spending mix, not a claim that conventional servers would disappear or that the projection had already been achieved.
The equipment stack is broader than GPUs. It includes CPUs, accelerators or custom AI chips, memory, storage, networking, data-center facilities, power systems and cooling. Software for orchestration, security, monitoring and data management helps make the hardware usable. A shortage or bottleneck in any of these layers can limit useful capacity even when a buyer has secured accelerator hardware.
Which areas are growing—and why the numbers differ
Data-center systems are central to the infrastructure story. Gartner’s February 2026 forecast projected data-center spending to rise 31.7% that year, exceeding $650 billion. Its April release projected data-center systems spending above $788 billion. These figures come from different forecast dates and should be read as successive estimates, not added together. In July, Gartner identified data-center systems and infrastructure as a service (IaaS) as leading growth segments.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Cloud infrastructure and IaaS let customers access compute, storage and networking without building a facility. AI capacity may be sold as accelerator instances, dedicated or reserved capacity, managed Kubernetes, hosted model platforms or inference endpoints. This shifts some infrastructure purchasing from enterprise capital budgets to provider investment and customer operating expenses; it does not remove the underlying cost.
Software is also part of the expansion, but software growth is not the same thing as infrastructure growth. Gartner’s April 2026 outlook pointed to momentum across AI infrastructure, software and IaaS. These categories support one another, but they have different buyers, cost structures and adoption risks.
Gartner also published separate AI-spending forecasts during 2026. A Jan. 15 forecast put worldwide AI spending at $2.52 trillion for 2026 and estimated that AI infrastructure would add about $401 billion in spending that year. A later May 19 forecast put AI spending at $2.59 trillion, up 47%, and said AI infrastructure would account for more than 45% of the total. Use the date attached to each estimate: these were different forecast releases, not two measures to combine.
Likewise, do not add AI-spending and worldwide-IT-spending totals as if they were separate markets. AI-related purchases can sit inside broader IT categories, and the market definitions differ. Spending forecasts say how much Gartner expects the market to spend; they do not establish that the investments will be profitable or deliver business value.
Who is actually paying?
Early investment in generative-AI infrastructure has been driven in significant part by technology suppliers building capacity: hyperscale cloud companies, AI providers and server and infrastructure vendors. Enterprises also create demand when they buy cloud capacity, AI software, consulting or managed services. Those are related parts of the market, but the organization paying for a service is not always the organization that bought the physical servers.
Rank #4
- Funny Roofer Design for roofers, roofing contractors, construction workers and tradesmen who spend their days climbing ladders, fixing shingles and keeping every roof overhead where it belongs. Tradesman, roofing crew, roofer dad and funny roofer.
- Built around popular roofing keywords such as roofer, roofing, roofing contractor, construction worker, roof repair and tradesman with a light jobsite joke for proud professionals with a head for heights.
- Hardcover journal with 240 line-ruled pages (120 sheets)
- Built-in elastic closure and ribbon bookmark
- Includes an expandable inner storage pocket and a pen holder
This distinction explains why rising global IT spending does not imply that every business is expanding its own data center. One company may buy and operate a cluster; another may rent cloud accelerators for a project; a third may use a managed AI service whose infrastructure costs are embedded in a subscription or usage charge. Each approach has different cost, control and staffing implications.
The physical and operational constraints
More demand does not automatically translate into immediately available capacity. Buyers and providers may face accelerator and memory supply limits, long equipment and construction lead times, power-grid and facility constraints, cooling challenges, and network bottlenecks. Data residency, security rules and regional availability can further narrow the set of suitable locations. Specialized infrastructure also requires people who can operate, secure and maintain it.
Capacity that exists on paper may not be usable for a particular workload. A customer can encounter regional shortages, quotas, provisioning delays, incompatible software, data-transfer charges or network limitations. And a large cluster can still be uneconomic if it is poorly utilized, starved of data, constrained by memory or left idle between jobs.
Windows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallCrashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteGartner’s July 2026 release connected growth in data-center systems and IaaS to what it described as the largest infrastructure project ever attempted by humanity. That is Gartner’s characterization, not a standardized measurement. The practical point is that expanding AI capacity depends on facilities, power and supply chains as well as technology budgets.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What the forecast means for an enterprise budget
A global market forecast is context, not a budget target. Before committing to infrastructure, buyers should identify the workload, estimate demand and compare the complete cost of viable deployment models.
- Separate workload types. Training, fine-tuning, batch inference, real-time inference, embeddings, data preparation and evaluation have different compute, latency and memory needs. A system suited to training may not be the economical choice for high-volume inference.
- Estimate demand and utilization. Model expected throughput, peak periods, idle time and growth. Assess the cost per useful result, not only the headline price per accelerator-hour. For inference, test whether caching, quantization, model choice or scheduling can change the economics.
- Compare ownership, cloud and colocation. Owning infrastructure may suit sustained, predictable workloads when an organization can keep it busy and has power, facilities and experienced operators. Cloud or managed services can fit uncertain, intermittent or fast-changing demand, but capacity, regional availability, data transfer, proprietary tooling and contract commitments matter. Colocation can offer site infrastructure and connectivity while leaving hardware control with the customer.
- Count the whole stack. Include hosts, memory, storage, network fabric, power and cooling, facility costs, software, cluster management, monitoring, security, backups, data movement and engineering labor. Account for idle capacity and hardware depreciation as well as the purchase or rental price.
- Protect flexibility. Validate accelerator availability and lead times; check quotas, service regions and support terms. Understand minimum commitments, egress charges and exit paths before relying on a provider-specific platform. Where practical, preserve portability without taking on more operational complexity than the team can support.
- Set a measurable investment gate. Expand beyond pilots when the workload has a defined owner, production demand, acceptable utilization and a credible value case. Track cost and performance against a baseline; avoid reserving large amounts of capacity solely because aggregate market spending is rising.
Build-versus-buy is not a universal rule. Predictable, sustained use and high expected utilization can strengthen the case for owned systems, provided power, cooling and operational expertise are available. Variable or experimental demand often favors rented or managed capacity. Colocation sits between those choices for organizations that want hardware control without constructing a data center. Actual economics depend on utilization, workload, region, energy, contract terms and staffing.
Risks behind the growth story
Gartner’s February 2026 outlook acknowledged concerns about an AI bubble while still forecasting strong infrastructure growth. High investment is evidence of expected demand, not proof that every AI application will succeed. Providers may continue building in anticipation of future usage even if some projects are delayed, abandoned or less profitable than expected.
Other risks include underused accelerators, rapid hardware obsolescence, volatile prices and supply, power delays and cloud dependence. In production, inference may become a recurring cost that exceeds the initial training expense; Gartner’s October 2025 forecast for AI-optimized IaaS projected $37.5 billion of spending in 2026, with 55% supporting inference. Those were Gartner forecasts, not audited market results. The figures underline why buyers should evaluate inference separately, rather than assuming the training bill captures the long-term cost.
The takeaway for technology leaders
The 7.9% headline captured Gartner’s July 2025 expectation for broad global IT spending in 2025. The later 14.2% forecast for 2026 shows how much the outlook changed as AI infrastructure and cloud investment accelerated. Together, the figures describe a powerful market trend—not a guaranteed return, a uniform increase across categories or an instruction for every organization to buy more hardware.
For buyers, the sound response is to match infrastructure to measured workload demand. Plan for inference as well as training, test utilization, account for power and the surrounding systems, and choose owned, cloud or colocated capacity according to cost, control and flexibility. The investment case rests on useful output and sustainable economics, not on the size of the global forecast.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
