OpEd: The AI Buildout Is Already on Your P&L, Even If You Never Own a GPU

This is an opinion piece. The facts and figures below are sourced and dated. The conclusions are Yayah Creative Co's own view, not investment, legal, or financial advice.

There is a version of the AI infrastructure story that reads like sports coverage. Who is building the biggest cluster, who signed the largest power deal, whose chip is fastest this quarter. It is genuinely interesting (we wrote about the chip war behind the AI boom for exactly that reason), and for most business owners it is also completely useless.

Here is the version that actually matters if you run a ten-person shop or a regional services company. The largest technology companies in the world are spending an amount of money on AI infrastructure that has no real precedent, and you are going to pay a small slice of it. Not through a purchase order. Through your software renewals, your utility bill, and the rate at which your competitors get faster than you.

The Number

Futurum Group's February 2026 analysis put combined 2026 AI infrastructure commitments from the five largest US cloud and technology companies at $660 to $690 billion, up from roughly $380 billion in 2025. CNBC's reporting on the same earnings cycle landed in the same neighborhood, describing AI spending approaching $700 billion for the year. Amazon alone guided to about $200 billion in capital expenditure, with Alphabet at $175 to $185 billion, Meta at $115 to $135 billion, and Microsoft at $120 billion or more.

A near doubling of spending in a single year is not a rounding error in the economy. It is a large, physical, electricity-hungry construction program, and construction programs get paid for.

Our view is simple: that money reaches your business through three channels, and each one deserves a different response.

Channel One: Your Software Renewal

The most direct path runs through the tools you already pay for.

Zylo's 2026 SaaS Management Index found that 79 percent of IT leaders encountered price increases at renewal in the prior 12 months, and 78 percent hit unexpected charges tied to consumption or AI features. The pattern is less about a new AI product line and more about AI being folded into the plan you already have, with the plan price moving up to match.

The examples are easy to find in the open. HubSpot charges $10 per 1,000 AI credits beyond the allotment in each subscription tier. Zendesk prices its AI resolution agent at $1.50 per resolved conversation. Neither of those is unreasonable on its face. Both change the shape of your bill from a number you can forecast to a number that moves with usage.

That is the real shift, and it is worth naming plainly. Software used to be a fixed cost you negotiated once a year. For any tool with AI in it, software is becoming a variable cost that behaves more like shipping or ad spend.

What to do: before your next renewal, pull twelve months of invoices for your top five tools and separate the subscription line from the usage line. If you cannot tell which is which, that is the finding. Ask the vendor for the AI credit or consumption terms in writing, and ask what happens when you exceed them. Then budget the variable portion as a range, not a number, the same way we suggest handling the rest of your Q4 and 2027 marketing budget.

Channel Two: Your Utility Bill

The second channel is slower and less obvious, and in some regions it is already the larger one.

The International Energy Agency's Energy and AI report, published in April 2025, estimated that data centres consumed about 415 terawatt-hours of electricity in 2024, roughly 1.5 percent of global consumption, and projected that figure to more than double to around 945 terawatt-hours by 2030. The IEA also projected that in the United States, data centres account for nearly half of electricity demand growth through the end of the decade.

You can see that pressure showing up in wholesale markets right now. PJM Interconnection, the grid operator serving more than 65 million people across 13 states and Washington DC, cleared its most recent capacity auction at a record $16.4 billion for the 2027 to 2028 delivery year. PJM's independent market monitor attributed about $6.5 billion of that, roughly 40 percent, to data centers, and noted that roughly $6.2 billion of the total related to forecasted data centers that have not been built yet. Across the last three auctions, the monitor put the data center share at $21.3 billion of $47.2 billion in cleared capacity costs. Capacity costs are a component of retail rates, and PJM customers were already looking at bill increases in the range of 1.5 to 5 percent for the twelve months beginning June 2026.

None of that is a claim that any particular company did anything wrong. Demand is demand, and grids price scarcity. But if your business has a shop, a bay, a walk-in cooler, a paint booth, or a server closet, electricity is a real line on your P&L and the trend is not your friend.

What to do: treat energy like the variable input it has become. Get a copy of your rate schedule and find out whether you are on a demand charge. Ask your utility what efficiency rebates are open right now, because many of them are use-it-or-lose-it. If you are renewing a lease, ask who controls the meter.

Channel Three: The Adoption Gap

The third channel is the one nobody invoices you for.

The US Census Bureau's Business Trends and Outlook Survey put overall US business AI use at 19.8 percent as of May 3, 2026, with usage hovering between 17 and 20 percent from December 2025 through May 2026. The split by size is the part worth reading twice. Thirty-seven percent of firms with at least 250 employees reported using AI. Thirty-two percent of firms with 100 to 249 employees did. Fewer than 20 percent of firms with four or fewer employees did.

Here is the honest read. Most of the money being spent on AI infrastructure is being spent by and for large organizations, and large organizations are adopting faster. If you are a small business waiting for the hype to settle, the hype settling is not the same as the gap closing.

Here is the more useful read. The capability those billions are building is available to you at consumer pricing. A five-person contractor and a five-thousand-person enterprise can buy access to broadly the same models, as long as you are not running on a free tier that is two generations behind. The enterprise has more data, more process, and more people to deploy it. The small business has something the enterprise does not, which is the ability to change a workflow on a Tuesday because the owner decided to.

What to do: pick one task your team does every week that is mostly writing, sorting, or summarizing, and run it through an AI tool for a month with a named person responsible for checking every output. One task, one owner, one month. That is the whole program. If you want a starting point, our guide to using ChatGPT and AI tools for marketing covers the tasks that are safe to hand off first, and building your first AI agent is the natural next step once the pilot works. It costs less than the software price increase you are already absorbing.

Our Take

The AI buildout is not a story about chips. It is a story about where costs land.

Somebody is paying for $660 billion in infrastructure this year, and the payment is being distributed across software renewals, electric bills, and the widening distance between companies that automate a workflow and companies that do not. You did not sign up for any of that, and you are in it anyway.

The businesses that come out of this well will not be the ones that read the most AI news. They will be the ones that got specific: who audited their variable software costs, who knows their demand charge, and who has one workflow running better than it did last quarter. That is unglamorous work. It is also the only part of this you control.

Want help figuring out which part of your marketing operation is worth automating first, and which part is not? See how we work.

Sources

• →Futurum Group, "AI Capex 2026: The $690B Infrastructure Sprint" (February 12, 2026)
• →CNBC, "Tech AI spending approaches $700 billion in 2026, cash taking big hit" (February 6, 2026)
• →Zylo, "2026 SaaS Pricing Trends Driving Up Enterprise Costs" (2026 SaaS Management Index) (2026)
• →International Energy Agency, "AI is set to drive surging electricity demand from data centres" (Energy and AI report) (April 2025)
• →Data Center Dynamics, "IEA: Data center energy consumption set to double by 2030 to 945TWh" (April 2025)
• →Canary Media, "PJM's capacity costs hit record as grid falls short on supply" (December 18, 2025)
• →Utility Dive, "Data centers were 40% of PJM capacity costs in last auction: market monitor" (January 7, 2026)
• →U.S. Census Bureau, "Large Firms With at Least 20 Employees Biggest AI Users" (Business Trends and Outlook Survey) (May 2026)

---

Opinion piece. Figures verified September 23, 2026 and cited to their original publishers: Futurum Group, CNBC, Zylo, the International Energy Agency, PJM Interconnection and its independent market monitor, and the U.S. Census Bureau. Nothing here is investment, legal, or financial advice, and nothing here is a claim about the conduct of any named company.