Invest1 publisher2 min readPublished
AI spend set to jump $800 billion as nearly all leaders fear falling behind peers
Fortune puts 2026 corporate AI spend above $2.5 trillion, 47% more than last year. The one thing the column measured was how many leaders believed rivals were ahead of them, and that answer was nearly unanimous.
The Investor · Invest desk
What happened
- Fortune projects companies will spend more than $2.5 trillion on AI in 2026, a 47% increase on 2025, and calls it a spending spree without precedent in organizational investment.
- The column attributes the surge in part to companies handing every employee access to GenAI tools such as Co-Pilot, Gemini or Claude.
- It describes three symptoms of the resulting hangover: surprise at the pushback, anxiety at how little business impact is visible, and concern that employees are doing worse work while feeling more overwhelmed.
- The most common corporate response, according to the column, is doubling down on encouraging staff to use tools the company has already sunk millions into.
Compiled by The InvestorSomething wrong?How this is made
Why it matters
- decision Accepting the diagnosis forces a choice between paying more to drive usage of seats already bought and shrinking the deployment, and only the second one changes next quarter's software bill.
- exposure Vendors priced per seat are exposed to the reversal of universal access: a licence model that assumes every employee gets one loses volume the moment buyers restrict it to trained users.
- cost Colleagues on the receiving end pay in review time, and the column says some of them ignore the incoming material altogether, so faster output does not become work anyone uses.
- contradiction A market-wide spending projection sits alongside outcome evidence gathered from one interviewer's conversations with leaders, so the two halves of the case cannot be weighed against each other.
Divide $2.5 trillion by 1.47 and 2025 comes out near $1.7 trillion, so roughly $800 billion of this year's bill is new money [3]. Fortune does not say who produced the projection or what it counts, which leaves open whether that $800 billion is mostly chips and buildings or mostly software seats [13].
The evidence on the other side of the ledger is thinner. Over the past year the column's author asked hundreds of leaders whether they felt their company was trailing others on AI adoption, and the answers came back nearly unanimously positive [4]. Asking leaders whether rivals are ahead measures how anxious they are about the gap. What the spending produced is a different question.
The more interesting number, or rather the only one in the piece with a denominator attached, concerns who holds the licences. Fortune says poor performers and average performers together tend to make up well more than half of any organization, and that these are the two groups who pay the most attention when a company encourages wide GenAI use [8]. The column identifies the driver of the spend as giving all employees access to tools such as Co-Pilot, Gemini or Claude [2]. Put the two together and more than half of the seats being paid for sit with the group the column says is producing worse work [15].
People are anchoring on the hallucination problem, Fortune says, when the real issue is that GenAI output is by definition average unless it is used to stretch thinking, and most of it should never be used as-is [12]. On the standard response of pushing harder on adoption, the author wrote: "Getting people to use these tools even more, at least the way they use them now, is only likely to make a big problem even bigger" [7]. The column's own diagnosis is that firms treat this as a technology rollout when it is closer to an overhaul of how people think [14].
If the $2.5 trillion is dominated by infrastructure, seat-level behaviour never reaches the cash, and repricing an AI budget means repricing a depreciation schedule. If the pushback and the overwhelm are the ordinary first-year cost of any new tool, the 47% increase buys a year of learning and the budgets hold [1]. And the outcome half of the case rests on conversations with leaders, so a study of output quality could land on either side [4].
In my view the repricing shows up first on universal seat licences, because a CFO can cut a seat count inside a quarter and only a seat cut books as a saving [10].
What to watch
- Whether 2026 enablement and training lines grow while seat counts stay flat, which would show buyers funding the doubling-down remedy.
- Any vendor disclosing renewal rates on universal-access deployments, the first place a narrowed rollout would appear.
- Publication of the leader survey with numbers attached, so the pushback and impact symptoms can be sized rather than described.