Spending on AI is easy to count and deployment is not, so the two get discussed as though they were one number. They are rarely close. A firm can sign a nine-figure commitment and still run its operations much as it did last year, because the work that would change that — data plumbing, process redesign, and the people who have to do their jobs differently — moves at nothing like the speed of procurement. These analyses keep the two figures apart: what was committed, what is in production, and which of the obstacles between them are technical rather than organizational.
Committed spend versus production use · integration and data readiness · vendor claims against filings · pilots that did not scale · measured returns
Everyone in the AI spending debate accepts that 95% of enterprise AI pilots never reach production; the fight is over whether that predicts weak demand for AI data centers or just bad customer-side integration.
Key takeaways▼
The strongest case for spending is not optimism about AI — data centers, power contracts and chip allocations take years to secure, so under-building cannot be undone later while over-building can be grown into.
Nvidia sits on both sides of its own AI sales: a $6.3 billion guaranteed-purchase order from CoreWeave, a $2 billion equity stake in CoreWeave, and reported talks to backstop up to $250 billion for OpenAI.
Key takeaways▼
Oracle has booked $638 billion of future cloud revenue, up 363% in a year, and roughly half of it — about $319 billion — depends on one customer, OpenAI, which is itself funded by venture and private capital.
Public pension funds in New York and Pennsylvania hold stakes in the infrastructure funds lending to AI data centers, and no dollar figure for that exposure has been disclosed.
Four US tech giants plan about $725 billion of AI spending in 2026, while just 11% of the S&P 500 shows deep AI integration in what they tell securities regulators — up fourfold from 5% in 2022.
Key takeaways▼
74% of the largest US and global companies told a survey they run AI in production, but only 21% score at production level or above in their own annual filings — a gap of roughly 53 percentage points.
Around 90% of the 6,000 executives in one survey report no measurable productivity gain from their AI investment, and about half of large firms running AI in production say they cannot prove it delivers value.
88% of companies report regular AI use in at least one function, and nearly two-thirds have not begun scaling AI across the enterprise — McKinsey's November 2025 survey reports both figures at once.
Key takeaways▼
The widely-repeated claim that 95% of corporate AI pilots fail comes from an MIT study of 300 deployments; Andreessen Horowitz's April 2026 analysis found 29% of the Fortune 500 paying for one AI startup's product.
Meta's stock fell after its CEO declined to quantify returns on the company's AI spending, in the same earnings cycle Alphabet disclosed specific cloud revenue growth and a large backlog.
Corporate AI pilots typically run on a manually cleaned, one-time data export that reflects neither live permissions nor production volume — which is why a working demo proves little. Cited failure rates run 86% to 95%.
Key takeaways▼
Enterprise AI pilots often don't fail outright — they quietly die after launch when nobody with budget authority is named to own uptime, quality and cost, which is why some companies now create an agent-owner role.
The employees whose daily work an AI system changes are retrained at step eight of nine, after all the technical work is finished, and change management failures are a commonly cited reason adoption stalls.
Microsoft, Amazon, Alphabet and Meta have guided to roughly $725 billion in 2026 AI capital spending, up 77% in a year, while a study of securities filings finds 11% of S&P 500 companies deeply integrated AI.
Key takeaways▼
Capital spending equals about 86% of Oracle's 2026 revenue and roughly half of Meta's, Microsoft's and Alphabet's, on estimates from credit research firm CreditSights, which called those levels "seemingly untenable".
The widely quoted claim that 95% of enterprise AI pilots never reach production comes from a July 2025 report, and secondary sources describe its sample as either 52 or 150 executive interviews.