Spending related to AI is projected to reach $2.59 trillion by 2026. According to Gartner's forecast, more than 55% of that—$1.43 trillion—will go toward AI infrastructure. The roughly $1 trillion figure that Lovelock explained to The Register refers to tech companies' own IT spending, not AI-specific expenditure. To gauge the true scale of this investment, we need to separate global IT purchasing from the AI market, and distinguish infrastructure from each company's capital expenditure.
Cost recovery doesn't come down to a single invoice, either. In enterprise software, AI features are being folded into base plans; in the cloud, charges scale with usage. For PCs and smartphones, rising memory prices are passing through to device costs. Meanwhile, buyers of IT services are demanding price cuts, citing productivity gains from AI. So while prices are rising for software and devices, buyers of outsourced services are pushing back with demands for lower prices.
What the $1.43 Trillion Buys—And How It Differs from the ~$1 Trillion Figure
Four distinct large numbers are circulating around 2026 AI investment estimates. The $2.59 trillion and $1.43 trillion figures come from Gartner's published AI spending forecast, while the $6.37 trillion and roughly $1 trillion figures are the latest numbers Lovelock shared in an interview with The Register. Comparing these without aligning their definitions risks conflating construction costs, software usage fees, and semiconductor purchases as if they were all the same category of capital expenditure.
| Figure | Scope | Nature |
|---|---|---|
| $6.37 trillion | Global IT spending | Latest forecast covering hardware, software, IT services, and telecom |
| $2.59 trillion | Global AI spending | Market forecast covering AI services, software, models, infrastructure, etc. |
| $1.43 trillion | AI infrastructure | AI-related IaaS, servers, networking, semiconductors, devices |
| ~$1 trillion | Tech companies' own IT spending | Estimate obtained by The Register through independent interviews with Gartner |
The $1.43 trillion in AI infrastructure represents 55.15% of the $2.59 trillion total AI spending figure—a 46.7% increase from $975.581 billion in 2025. This category includes not only AI servers purchased by cloud providers, but also AI-related IaaS, networking, semiconductors, and end-user devices. It is not synonymous with the construction costs of data center buildings.
The separate ~$1 trillion figure represents tech companies' own IT spending, which Gartner's John-David Lovelock told The Register is expected to grow 34.7% in 2026. This is not a figure specifically for AI capital expenditure. Moreover, according to Lovelock, Gartner's IT spending statistics exclude buildings and cooling systems. While undeniably a massive figure, it doesn't represent the total physical construction costs of AI infrastructure either.
Global IT Spending Revised Upward by $285.9 Billion in Nine Months
The 2026 global IT spending forecast rose from $6.084 trillion in October 2025 to $6.37 trillion by July 2026. That's an increase of $285.9 billion over nine months, equivalent to 4.7% of the original forecast. The year-over-year growth rate was also revised upward, from 9.8% to 14.2%.
Data centers are at the core of this upward revision. In October 2025, Gartner projected 2026 data center systems spending at $582.446 billion, a 19% year-over-year increase. By April 2026, this had been revised to $787.99 billion, a 55.8% increase—meaning the figure rose by $205.544 billion, or 35.3%, in just six months.
According to the most recent interview, IaaS alone is expected to reach $287 billion in 2026, a 29.3% increase—accelerating further from 25.3% growth in 2025. Spending on AI models has also been revised upward to 110% year-over-year growth, with Gartner adding $6 billion to its previous forecast. Investment in server deployment and spending on running models atop that infrastructure are both expanding simultaneously.
However, not all of this spending growth reflects increased sales volume. Gartner's April forecast noted that HBM prices are hitting record highs due to demand and supply constraints, and that rising general memory prices are also pushing up average selling prices for devices. While device spending is projected to grow 9.8% as of July, Lovelock explains that a substantial portion of that increase is attributable to price increases rather than volume. AI infrastructure demand combined with supply constraints is driving up HBM prices, and rising general memory costs are showing up in the devices purchased by both businesses and consumers.
Microsoft and Google Are Baking AI Into Base Pricing
On July 1, 2026, Microsoft raised prices across several commercial Microsoft 365 plans. Microsoft 365 E3, which includes Teams, rose 8% from $36 to $39 per user per month, while Business Standard rose 12% from $12.50 to $14. Frontline's F1 plan, which doesn't include Teams, saw a 43% increase from $1.75 to $2.50, though some plans remained unchanged.
Alongside the price changes, Microsoft added Copilot Chat, which can now recognize inboxes and calendars. Agents for Word, Excel, and PowerPoint are also now available. Security, IT management features, and mailbox storage have all increased across various plans. This makes it impossible to attribute the price increases solely to AI computing costs. Still, the pattern is clear: rather than offering AI as a separate add-on contract, Microsoft is folding it into the renewal pricing of widely used base plans. Standalone Copilot was not affected by this round of price increases.
Google Workspace adopted a similar bundling approach for AI in 2025. Business Standard shifted from $12 per month without AI to $14 per month including Gemini. Customers who had previously paid for the Gemini Business add-on were paying $32 per month, meaning those migrating from that plan saw their total costs drop by $18. Meanwhile, customers who hadn't purchased the add-on saw their base price rise by $2, a 16.7% increase.
Under this bundling approach, total costs decrease for customers who previously held add-on contracts, while customers who never purchased the add-on are now subject to the new, higher base price. This allows vendors to rapidly expand the customer base with access to AI features, but makes it harder for buyers to strip out unwanted features to control spending. The shift Gartner forecast in October 2025—"AI gets built into existing software, and both features and software prices rise"—is now showing up in 2026 renewal invoices.
Hardware and Cloud Investments Recover Costs on Different Timelines
Alphabet expects 2026 capital expenditure to fall between $195 billion and $205 billion—2.1 to 2.2 times its 2025 actual figure of $91.447 billion. However, this capital expenditure isn't dedicated solely to Google Cloud or AI; it supports company-wide computing needs including foundation model development, Search, and YouTube.
In Q2 2026 alone, Alphabet invested $44.924 billion in capital expenditure. Google Cloud revenue for that same quarter was $24.768 billion, with operating income of $8.814 billion. Cloud revenue grew 82% year-over-year, converting enterprise demand into revenue. Even so, Alphabet's free cash flow (as the company defines it) showed a deficit of $5.855 billion. Cloud profit margins and company-wide capital expenditure paid upfront operate on entirely different timelines.
Alphabet's management explained that decisions about whether input costs can be passed through to pricing are evaluated based on multi-year return on invested capital, factoring in long-term contracts and renewal demand. This doesn't mean that increasing capacity immediately translates into proportional price increases. Recovery comes through a combination of cloud usage volumes, product sales like TPUs, contract terms, and renewal pricing. To break down the 29.3% IaaS growth into volume versus unit price components, one would need to examine each company's disclosed usage figures and effective unit prices, as well as track how these translate into profit margins and cash flow.
Upward Price Pressure Meets Downward Price Demands in IT Services
Corporate CIOs are pushing back against the price increases vendors are proposing across software and hardware alike. According to what Lovelock told The Register, the one area where buyers are actually succeeding in negotiating lower prices is IT services. When service providers add AI capabilities, customers anticipate efficiency gains and demand lower prices in return. AI adoption doesn't necessarily strengthen suppliers' pricing power.
Some costs fall outside these statistics entirely. As Lovelock explained to The Register, Gartner's IT spending statistics exclude buildings and cooling equipment, and electricity demand is rising separately. According to the U.S. Department of Energy's summary of LBNL's updated 2025 estimates, U.S. data centers are projected to account for 11.8% of national electricity usage by 2030, with scenario ranges spanning 9.5% to 15.3%. This is a forecast of U.S. electricity demand—not an AI company's investment figures or current actuals. But it does illustrate that server purchase amounts alone cannot capture the full picture of infrastructure expansion.
While Gartner views 2026 as a turning point where mainstream enterprise AI spending accelerates, what companies are actually choosing to implement on the ground right now are incremental measures aimed at efficiency and productivity gains. Will usage grow even as software renewal prices rise? Will usage-based revenue from IaaS and AI models absorb infrastructure costs? Will productivity gains in IT services actually flow back to customers? Tracking these three pricing dynamics will reveal whether mainstream enterprise adoption can catch up to this wave of advance investment—and whether the $2.59 trillion in spending can be converted into sustainable demand.