Quick wins for a faster PC:
Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Clear out junk files and repair common Windows errorsFree Scan →Scan for outdated or missing drivers - takes under a minuteDriver Scan →iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
The $690 billion headline is an upper-end estimate of 2026 capital spending by hyperscalers and adjacent AI-infrastructure providers—not a single measure of cloud revenue. The total is methodology-dependent: estimates differ on whether they include only the largest cloud companies or also enterprises, newer infrastructure providers, power projects and semiconductor-related investment. A midpoint estimate based on company guidance is about $610 billion, so the headline should be treated as a high-case projection rather than a settled total.
What the $690 billion estimate includes—and what it does not
Cloud spending can describe several different economic flows. A cloud provider’s capital expenditure pays for facilities and equipment. Public-cloud end-user spending measures what customers consume. AI-optimized infrastructure spending isolates systems designed for artificial-intelligence workloads. Total AI spending also includes software, services and other categories outside cloud infrastructure.
The headline figure is intended to aggregate capital expenditure by the major hyperscalers and related AI-infrastructure companies during calendar year 2026. It is not an official industry revenue statistic, and there is no single accounting standard behind the aggregate. Company guidance can change at each earnings cycle, while analysts may use different company lists, geographies and inclusion rules.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
The market figures that put the buildout in context
| Measure | Forecast or reported value | What it measures |
|---|---|---|
| Worldwide IT spending | $6.37 trillion in 2026, up 14.2% from 2025 | Gartner’s total global IT-spending forecast; data-center systems and infrastructure as a service are the fastest-growing segments. |
| AI-optimized IaaS | $42 billion in 2026 | Gartner’s forecast for infrastructure-as-a-service optimized for AI workloads, growing 96% during the year. |
| AI-optimized IaaS | $66 billion in 2027 | Gartner’s following-year forecast for the same category. |
| Inference share of AI-optimized IaaS | 55% in 2026; 59% in 2027 | Gartner’s estimate of the portion supporting model inference rather than training. |
| Worldwide public-cloud end-user spending | $723.4 billion in 2025, up from $595.7 billion in 2024 | Gartner’s forecast for customer spending on public-cloud services. |
| Worldwide AI spending | $2.59 trillion in 2026, up 47% year over year | Gartner’s broad AI-spending forecast; AI infrastructure accounts for more than 45%. |
| Hyperscaler capital expenditure | About $610 billion in 2026 | Axios’s midpoint estimate derived from company guidance. |
| Cloud-infrastructure services | $143 billion in 2026 Q2, up $43 billion year over year | Synergy Research Group figure reported by ITPro; GenAI-specific cloud services grew 165% year over year. |
These figures cannot be added together. For example, public-cloud consumption is a customer-spending measure, while hyperscaler capital expenditure is an investment measure. AI spending includes categories that are not necessarily purchased through a public cloud.
#1 Best Overall
Why artificial intelligence is driving the spending surge
Training large models requires dense clusters of accelerators, high-speed networking, large storage systems and software capable of coordinating thousands of chips. Production inference creates a different but continuing demand: models must answer user and business requests around the clock, often close to the data and with predictable latency. That turns AI capacity into an ongoing operating requirement rather than a one-time training project.
Gartner Distinguished Vice President Analyst John-David Lovelock described data-center systems and infrastructure as a service as the top growth segments, reflecting accelerating investment in AI infrastructure, cloud platforms and intelligent applications. Gartner analyst Hardeep Singh said the growth is driven by infrastructure for large-language-model training and the rapid operationalization of AI across enterprise applications and workflows.
Rank #2
The shift toward inference matters because it signals that cloud demand is moving from experimentation to production. Training can involve a concentrated burst of spending for a new model; inference requires capacity to remain available as applications gain users. Gartner expects inference to represent a majority of AI-optimized IaaS spending in 2026 and an even larger share the following year.
Free tools Windows power users keep installed
One-click scans. No signup required.
Where the capital goes
Hyperscaler budgets are not spent on processors alone. The buildout typically includes:
Rank #3
- Data-center construction, land preparation and internal electrical systems.
- Servers containing AI accelerators and conventional CPUs.
- High-bandwidth networking fabrics that connect accelerators and storage.
- Storage for training datasets, model checkpoints, logs and customer workloads.
- Liquid-cooling and other thermal-management systems for dense equipment.
- Power generation, substations and transmission capacity needed to operate new campuses.
- Cloud software platforms for provisioning, security, orchestration and workload management.
Because these components have different delivery schedules, a provider can announce a large capital plan before all of the resulting computing capacity is online. Spending may therefore appear in one year while usable capacity, customer revenue and electricity demand arrive over several years.
Who is spending, and what is known about each company
The principal spenders identified in the outlook are Amazon Web Services, Microsoft Azure, Google Cloud, Meta and Oracle, alongside other cloud and AI-infrastructure providers. Their investment is concentrated in data centers, accelerators, networking, storage, cooling and power systems.
The available forecasts do not establish a reliable 2026 total for each named company. Guidance is issued on different schedules and may cover broader capital programs than cloud alone. A company-by-company ranking should therefore be labeled with the date of the guidance and the exact definition of capital expenditure used. The aggregate midpoint is more defensible than presenting an unverified split among providers.
How to judge any “nearly $700 billion” cloud claim
- Identify the flow. Check whether the number is capital expenditure, customer cloud revenue, AI-infrastructure spending or total AI spending.
- Check what hardware and facilities are included. Some estimates include accelerators and servers but exclude power projects, buildings or semiconductor investment; others include them.
- Check the company and geography list. Ask whether the calculation covers only the largest hyperscalers or also enterprises, neoclouds and regional providers.
- Check the forecast vintage. A 2026 estimate made before a quarterly earnings update may differ from one made afterward.
- Check whether it is a midpoint, high case or realized result. A midpoint based on guidance is not equivalent to money already spent, and an upper-end case should not be reported as a confirmed total.
Who is positioned to benefit from the buildout
Demand is likely to reach several layers of the infrastructure supply chain: accelerator and server manufacturers, networking suppliers, storage providers, data-center builders, cooling specialists, utilities and power-equipment companies, and software vendors that operate AI clusters. The spending also creates work for firms that connect new campuses to local grids and manage the resulting infrastructure.
Best Value
Those opportunities do not remove execution constraints. Deliveries depend on chip availability, construction schedules, grid connections, cooling design and the ability to turn installed equipment into billable workloads. If customer demand or model economics change, providers can adjust later phases of their plans even after committing to early projects.
What readers should conclude
The cloud industry is entering an infrastructure cycle measured in hundreds of billions of dollars because AI requires far more compute, power, networking and cooling than earlier cloud applications. The $690 billion figure is best used as an upper-bound description of that planned 2026 investment. The more comparable midpoint is about $610 billion, while Gartner’s separate forecasts show how public-cloud consumption, AI-optimized IaaS and total AI spending describe different parts of the same expansion. Treating those measures as interchangeable would make the market look larger—and more precise—than the evidence supports.
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.
