The “$690 billion cloud spending spree” is an upper-end estimate of what hyperscalers and adjacent AI-infrastructure providers could spend on capital projects during calendar year 2026. It is not an official measure of cloud revenue, and no single published methodology establishes that exact total. An Axios midpoint based on company guidance is about $610 billion, illustrating how much the headline depends on which companies, facilities and AI-related investments are counted.
What the $690 billion figure includes
The estimate describes capital expenditure: money committed to long-lived infrastructure rather than sales collected from cloud customers. Its likely core is the buildout by Amazon Web Services, Microsoft Azure, Google Cloud, Meta and Oracle, alongside other AI-infrastructure providers. Depending on the calculation, the total may also include neocloud operators, enterprise projects, power facilities and semiconductor-related investment.
As an Amazon Associate I earn from qualifying purchases.
That scope is why the number should be treated as a high-case aggregate, not a precise industry total. Company guidance can change at earnings updates, and estimates differ over whether a data center, power project or hardware purchase is classified as cloud or AI infrastructure.
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREERepair Windows errors before they cause bigger problemsFix Now →How the major market measures compare
| Measure | Value | Period and qualification | What it measures |
|---|---|---|---|
| Hyperscaler and adjacent-provider capex | About $690 billion | 2026 upper-end estimate | Aggregate infrastructure investment; methodology varies |
| Hyperscaler capex | About $610 billion | 2026 midpoint from company guidance, reported by Axios | Capital spending by the companies included in the estimate |
| Worldwide IT spending | $6.37 trillion | 2026 Gartner forecast, up 14.2% from 2025 | Total IT spending; data-center systems and IaaS are the fastest-growing segments |
| Worldwide public-cloud end-user spending | $723.4 billion | 2025 Gartner forecast, versus $595.7 billion in 2024 | Customer spending on public-cloud services |
| AI-optimized IaaS | $42 billion | 2026 Gartner forecast, up 96%; $66 billion forecast for 2027 | Cloud infrastructure optimized for AI workloads |
| Inference share of AI-optimized IaaS | 55% | 2026 Gartner forecast; 59% forecast for 2027 | Share supporting deployed-model inference rather than primarily training |
| Worldwide AI spending | $2.59 trillion | 2026 Gartner forecast, up 47% year over year | All AI spending categories; infrastructure is more than 45% |
| Cloud-infrastructure services | $143 billion | Second quarter of 2026, according to Synergy Research Group figures reported by ITPro; $43 billion higher year over year | Quarterly services market; GenAI-specific cloud services grew 165% year over year |
These figures cannot be added together. Public-cloud end-user spending is revenue flowing from customers to providers, while hyperscaler capex is investment made by providers and partners. AI spending includes software, services and systems beyond cloud infrastructure. Each measure answers a different question.
#1 Best Overall
Why AI is driving the buildout
Training needs enormous concentrated capacity
Training large language models requires clusters of accelerators connected by high-speed networking, with storage systems able to feed data continuously. A provider must also supply cooling, electricity and the software layer that schedules and monitors the hardware. Expanding those clusters requires new data-center space and supporting power infrastructure, not just additional virtual machines.
Inference is becoming the larger operating workload
Gartner expects inference to represent 55% of AI-optimized IaaS spending in 2026, rising to 59% in 2027. Training remains capital-intensive, but production applications generate recurring inference demand whenever users submit prompts, documents, images or other inputs. That shift helps explain why providers are building capacity for everyday enterprise workloads as well as model development.
Rank #2
“This growth is driven by continued demand for infrastructure to support large language model (LLM) training and the rapid operationalization of AI across enterprise applications and workflows.”
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Fix the driver behind crashes, sound loss and screen glitches3Repair Windows errors before they cause bigger problemsSpecial offer. See more information about Outbyte and uninstall instructions. Please review EULA and Privacy policy.— Hardeep Singh, Gartner senior principal research analyst
Who is spending and where the money goes
The largest named spenders in the estimate are Amazon Web Services, Microsoft Azure, Google Cloud, Meta and Oracle, with other cloud and AI-infrastructure companies contributing. The available forecasts do not provide a definitive, apples-to-apples dollar split for each company, so claims that one provider is responsible for a specific share should be date-stamped and tied to that provider’s own guidance.
- Data centers: land, buildings, electrical systems and fit-outs for high-density computing.
- AI accelerators and servers: processors, memory and complete systems for training and inference.
- Networking: high-bandwidth fabrics linking accelerators, storage and regional facilities.
- Storage: systems for training datasets, model checkpoints and application data.
- Cooling: including liquid-cooling equipment needed for dense accelerator racks.
- Power: generation, substations, transmission and other grid-related capacity.
- Operating software: platforms that provision, orchestrate and monitor cloud and AI infrastructure.
Spending therefore reaches construction firms, equipment manufacturers, chip and server suppliers, networking vendors, utilities and software providers. The size of the aggregate does not by itself show which supplier will capture the most revenue or profit.
How to judge whether a cloud-spending estimate is credible
- Identify the economic flow. Check whether the number is provider capex, customer cloud revenue, infrastructure-services revenue or total AI spending.
- Check what physical assets are included. Some estimates count only servers and accelerators; others add data-center construction, power projects, cooling and networking.
- List the companies and geography. Confirm whether the total covers only the major hyperscalers or also enterprises, neoclouds and adjacent infrastructure providers, and whether it is worldwide or limited to a region.
- Match the calendar year and forecast vintage. A 2026 forecast issued before a major earnings update can differ from one revised later in the year.
- Separate a midpoint, high case and realized spending. The $690 billion headline is an upper-end estimate, while the approximately $610 billion figure is a midpoint based on company guidance; neither is audited realized spending.
What the estimate means for cloud customers
More capacity can improve access to accelerators and support larger AI deployments, but construction does not automatically make every workload cheaper. Providers must recover the cost of hardware, buildings, electricity, networking and operations through usage charges and long-term contracts. Customers should therefore treat the capex headline as evidence of supply expansion, not as a promise of lower prices or unlimited availability.
Recommended Free Tools
Bottom line
The $690 billion figure captures the scale of the 2026 AI infrastructure race, but it is best read as a high-end capex estimate. Gartner’s separate forecasts show why the buildout is happening: AI-optimized IaaS is growing rapidly, inference is taking a majority share, and data-center systems and IaaS are leading IT growth. To compare headlines accurately, always separate capital investment from cloud revenue, record the forecast date, and inspect exactly which companies and physical projects are counted.
Quick Recap
Best Value
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.
