Make this research yours. Add it to a free WorldbyFlow workbench to run follow-ups, ask questions, and re-check it as events move.

The Trajectory Read

Generative AI is not on one curve: the analyst record places core generative AI in a genuine Trough of Disillusionment on enterprise ROI proof, while capital investment and production-deployment volume are still accelerating — a divergence the classic single-technology hype-cycle model was not built to describe.

Gartner's own 2026 Hype Cycle places generative AI in the Trough of Disillusionment, even as hyperscaler capital spending accelerates toward roughly $725 billion for 2026 — a divergence between narrowing expectations and widening infrastructure commitment. The evidence supports a split-screen read: enterprise ROI proof remains the binding constraint while investment and production-deployment volume keep climbing, which is not the pattern the classic Gartner curve was built to describe.

The clearest fact anchoring this trajectory is that Gartner itself, the author of the hype-cycle framework, has placed generative AI in the Trough of Disillusionment in its most recent cycle. Gartner's own framing states that the Trough of Disillusionment happens when the original excitement wears off and early adopters report performance issues and low ROI, and Gartner's Hype Cycle for Artificial Intelligence documented that despite an average spend of $1.9 million on GenAI initiatives in 2024, less than 30% of AI leaders reported their CEOs were happy with AI investment return. That is the strongest primary-source evidence for the trough thesis: the analyst firm that coined the model applying it to this specific technology, with a disclosed spend figure and a disclosed satisfaction rate attached.

But the adoption data cuts against a simple reading of disillusionment as retreat. A 2026 enterprise survey found that 74% of the world's largest enterprises now run at least one AI solution in production and 93% are piloting or further along, while half of those production-stage companies cannot tell you whether any of it worked. That is not a technology in retreat — it is a technology that has cleared the deployment bar and stalled at the measurement bar, a structurally different failure mode than the classic trough (where users abandon the technology after early disappointment). Separately, MIT's Project NANDA reported that despite $30–40 billion in enterprise investment, 95% of generative AI projects yielded no measurable business return in its original 2025 study — a number that has since become, per one tracking analysis, competing with McKinsey's 17% EBIT-attribution figure as the most-cited single statistic in enterprise AI commentary, and one whose procurement-deck usage a later review argues materially overstates what the underlying survey data supports. Both figures are real, sourced, and in active tension; neither resolves the other.

The capital story is the sharpest argument against a clean hype-cycle read. The four largest hyperscalers — Amazon, Alphabet, Microsoft, and Meta — are guiding to combined 2026 capital expenditure of roughly $725 billion, up about 77% from approximately $410 billion in 2025, according to figures compiled from first-quarter 2026 earnings calls. Classic Trough-of-Disillusionment technologies see capital retreat as expectations fall; generative AI is instead seeing the largest coordinated infrastructure buildout among the companies best positioned to know whether demand justifies it, even as their own share prices sold off around the guidance. That is either evidence the infrastructure layer and the application layer are on different curves entirely, or evidence of a bubble sustained by circular financing — Nvidia investing in OpenAI, which buys cloud from Oracle, which buys chips from Nvidia, which holds a stake in CoreWeave, which serves OpenAI — a structure Bloomberg has tracked as a recurring graphic and that prompted a widely reported CNBC comment likening the pattern to dot-com-era vendor financing.

The honest position is that generative AI is not one technology on one curve. Gartner's 2026 GenAI-specific Hype Cycle itself distinguishes large language models — which it calls the most mature technology on the Hype Cycle — from agentic AI, which a separate April 2026 Gartner Hype Cycle for Agentic AI placed at the Peak of Inflated Expectations, careening toward its own Trough. Layering one curve across foundation models, enterprise chatbots, coding assistants, and autonomous agents obscures more than it reveals; the sub-technologies are years apart on their own arcs even as commentary treats "generative AI" as a single dot on a single chart.

The claims commonly made about this trend, checked against the record — separating the evidenced arc from the narrated one.

Source facts the analysis is grounded in. The → chips after each fact link to the items above that rely on it.

The claims behind this analysis, each with its verification status — including what is contested, unverified, or could not be established.

August 31, 2026

August 31, 2026

August 31, 2026