IDC forecasts that Global 1,000 companies will underestimate AI infrastructure costs by 30% through 2027, according to CIO. That is a forecast for a defined group and time horizon—not a measured result for every company or a guarantee that any one budget will be short by exactly 30%. The practical warning is that budgets built around a pilot or model-compute bill can miss the wider costs of running AI as usage grows.
What does the 30% forecast mean?
The figure comes from an IDC forecast reported by CIO. The underlying forecast methodology is not available in the reviewed material, so the percentage should be treated as an attributed projection, not an independently verified average or a universal rule. It does not specify a fixed surcharge that companies should add to every AI project budget.
| # | Preview | Product | Price | |
|---|---|---|---|---|
| 1 |
|
MINISFORUM MS-02 Ultra Workstation Mini PC, Intel Core Ultra 9 285HX (24C/24T, up to 5.5GHz), PCIe... | $1,659.00 | Buy on Amazon |
| 2 |
|
GMKtec EVO-X2 AI Mini PC Ryzen Al Max+ 395 Superchip 128GB LPDDR5X 2TB SSD | $3,649.99 | Buy on Amazon |
It does point to a budgeting problem: AI costs can change with usage, workload design, and the systems needed to operate models safely. IDC vice president of infrastructure and operations research Jevin Jensen described the shift this way: “AI has moved technology spending from predictable consumption to probabilistic behavior.” For finance and technology leaders, that makes ongoing measurement more useful than relying only on the initial business case.
Which costs can a narrow AI budget miss?
Model training or inference is only part of the bill. The mix depends on the workload, where it runs, how many people use it, and what controls the organization requires. IDC’s Jevin Jensen identifies GPUs, inference, networking, and tokens as complex budget factors, alongside security, governance, and employee training.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
#1 Best Overall
- High-Performance AI Processor：The MS-02 Ultra features an Intel Core Ultra 9 285HX (24C/24T, up to 5.5 GHz, 13 TOPS NPU), delivering fast and efficient performance for AI inference, algorithm development, and media workloads. A PCIe x16 expansion slot supports desktop-class GPU upgrades for advanced model training and accelerated computing tasks. It's ideal for creators, engineers, and teams handling intensive parallel workloads.
- 4 × M.2 PCIe 4.0 + 4 × DDR5 SODIMM slots:Four DDR5 SODIMM slots support up to 256 GB of memory, while ECC helps maintain data integrity in mission-critical environments. Four PCIe 4.0 M.2 slots support up to 24 TB of storage, supporting RAID 0/1/5/10, combining high-speed performance with data protection. It allows for the creation of independent scratch disks, media libraries, and project drives, providing high-throughput for production workflows.
- PCIe & USB 4.0 v2: Up to three PCIe slots can be equipped, including a dual-slot x16 GPU. The main slot supports PCIe 5.0, meeting the needs of high-bandwidth creative and computing workloads. USB 4.0 v2 (80Gbps) supports high-bandwidth external storage and displays.
- Ultra-fast Networking: Wi-Fi 7 further enhances wireless performance with next-generation speeds and low-latency stability. Intelligent bandwidth switching optimizes throughput in different network environments, ensuring optimal performance for enterprise or local networks. Dual 25GbE ports (providing up to approximately 3.125 GB/s bandwidth, about 25 times faster than traditional 1GbE), enabling seamless large-scale file transfers and parallel computing. 10GbE and 2.5GbE ports, with support for Intel vPro technology, ensure enterprise-grade remote management and deployment flexibility.
- Server-grade thermal architecture: Utilizing a dedicated CPU/GPU airflow design, equipped with a 6-pipe dual-fan cooler, it maintains stable performance even under sustained loads, delivering up to 140W Turbo power while maintaining a 100W TDP, and operating with noise levels as low as 36 dB. An integrated 350W power supply ensures stable and reliable output for demanding computing tasks and fully loaded extended configurations.
- Compute and model use: GPUs and other compute resources may support training and inference; token-based usage can also vary as requests and adoption change.
- Data movement and connectivity: Networking costs can rise when data must move between systems, users, models, or cloud environments.
- Operations and control: Monitoring, drift detection, logging, and validation consume resources. Cisco’s Nik Kale says supporting systems can cost as much as, or more than, model inference in some environments; that is an observation about some environments, not a universal cost ratio.
- Security and governance: Access controls, data handling, and oversight need to be planned as part of deployment rather than treated as unrelated overhead.
- People and adoption: Training employees and supporting a wider user base add organizational costs, while broader adoption can also increase infrastructure consumption.
Kale also cautions that model use may spread beyond the team originally included in the plan. A budget based on a small pilot can therefore miss both higher consumption and the operational systems needed when access expands.
How do cloud and on-premises costs differ?
Neither deployment model is automatically the cheapest. Cloud spending is generally an operating expense tied to the workloads and services used; on-premises deployments involve hardware investment as well as ongoing operating costs. Both require workload-specific cost management. Existing infrastructure may be sufficient for some AI projects, while other workloads may require added capacity.
| Deployment choice | Cost questions to model | Planning caution |
|---|---|---|
| Cloud | How workload volume, model use, networking, and supporting services affect recurring spend. | Track consumption and allocate costs to the teams and workloads creating them; do not assume cloud is always more expensive. |
| On-premises | Hardware investment alongside staffing, power, cooling, operations, and the workload’s actual utilization. | A hardware purchase alone does not remove cost uncertainty or guarantee lower total cost than cloud. |
| Hybrid | Which workloads run in each environment, how data moves between them, and what governance and operations each requires. | Hybrid is a design option, not an automatic cost saving; compare the full workload-specific costs. |
Before choosing, compare utilization patterns, data movement and governance needs, staffing and skills, expected business value, and power, cooling, and capacity requirements. Deloitte also flags memory-component costs, longer procurement times, potential wafer-cost increases, grid interconnections, and air versus liquid cooling as factors that may affect some deployments. These are planning variables, not fixed charges that apply to every project.
Why does the wider infrastructure environment matter?
AI budgets sit within a physical infrastructure market where electricity and capacity can be constraints. The International Energy Agency reported that data-center electricity demand rose 17% in 2025 and described bottlenecks involving supply chains and grid connections. That sector-level statistic does not measure the effect on an individual enterprise’s bill, but it is a reason to include power availability and cooling in capacity planning rather than treating them as afterthoughts.
Rank #2
- EVOLUTION RYZEN AI MAX+ 395 MINI PC - GMKtec EVO-X2 is the next evolution in AI mini PC Ryzen Strix Halo series. Thanks to AMD Simultaneous Multithreading (SMT) the core-count is effectively doubled, to 32 threads. Ryzen AI Max+ 395 has 64 MB of L3 cache and can boost up to 5.1 GHz, depending on the workload. The Ryzen AI Max+ 395 is currently rated as the "most powerful x86 APU" on the market for AI computing.
- AI NPU with XDNA 2 ARCHITECTURE - Powered by 16 “Zen 5” CPU cores, 50+ peak AI TOPS XDNA 2 NPU and a truly massive integrated GPU driven by 40 AMD RDNA 3.5 CUs, the Ryzen AI MAX+ 395 is a transformative upgrade and delivers a significant performance boost over the competition. The Ryzen AI Max+ 395 excels in consumer AI workloads like the llama.cpp-powered application: LM Studio. Shaping up to be the must-have app for client LLM workloads, LM Studio allows users to locally run the latest language model without any technical knowledge required and unleash their creativity and productivity.
- AMD RADEON 8090S iGPU GAMING PC - The AMD Radeon RX 8060S offers all 40 CUs with up to 2.9 GHz graphics clock and uses the new RDNA 3.5 architecture. The powerful iGPU is positioned between an RTX 4060 and 4070 laptop GPU and therefore enables gaming in FHD at maximum details in most demanding games. The 8060S can also utilize the full 128GB pool, which is perfect for running LLMs such as Deepseek 70B Q8, which runs comfortably on this machine.
- EIGHT CHANNEL LPDDR5X - LPDDR5X is a new ground breaking memory small form factor installed on-board. With blazing speeds up to to 8000MT/s, it runs 1.5x faster than the DDR5 SODIMMs; 90% better performance over DDR5 SODIMMs in video conferencing and photo editing; 30% better performance in productivity apps; 12% better performance in digital content workloads.
- QUAD SCREEN 8K DISPLAY SUPPORT - EVO-X2 AI Mini PC support 4-screen 4K/8K output via HDMI 2.1 (8K@60Hz), DisplayPort 1.4 (4K@60Hz), and dual USB 4 40Gbps Transfer speed (supporting PD3.0/DP1.4/DATA). Ideal for gaming, video editing, and multitasking, it provides expansive and crisp multi-display support.
The IEA also notes that efficiency per AI task is improving while usage is rising. Efficiency gains alone therefore do not establish that total infrastructure demand—or an organization’s costs—will fall. The agency reported that five large technology companies’ capital expenditure exceeded $400 billion in 2025 and was set to rise another 75% in 2026; those figures describe sector investment, not a cost estimate for an enterprise AI deployment.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What do companies expect their budgets to do?
Deloitte’s 2025 survey, published in March 2026, found that 86% of respondents expected AI infrastructure budgets to increase over the following three years. Respondents expected budgets to more than triple on average, while large enterprises projected almost four times the current level. Those are expectations, not realized spending or a forecast that applies to every company.
The survey was fielded in November and December 2025 among 515 US business and technology decision-makers at director level or above, across five industries. Their organizations had at least US$500 million in revenue. The sample and geography matter: the results describe expectations among those respondents, not all businesses or a specific organization’s likely costs.
Quick Recap
How should finance and technology leaders plan?
- Build a workload-level baseline. Estimate compute, inference, tokens, networking, data movement, and supporting services for each intended use—not just the pilot.
- Include operating requirements. Account for security, governance, training, monitoring, logging, drift detection, validation, staffing, power, and cooling where they apply.
- Assign ownership and allocation. Make usage visible by workload or team so finance and technology leaders can see what is driving consumption and who benefits from it.
- Track actual use continuously. Review consumption and costs regularly, update forecasts as adoption expands, and revisit assumptions when pricing or capacity conditions change. IDC’s account of its FinOps guidance and Deloitte’s advice both emphasize tracking and auditing AI consumption.
- Test spending against business value. Tie capacity decisions to expected ROI and adjust investment when the realized benefits, utilization, or operating costs differ from the plan.
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.
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Fix the driver behind crashes, sound loss and screen glitches3Clear out junk files and repair common Windows errors
