Only seven percent of enterprises have AI-ready data — and the number is declining as models advance. Six hundred and sixty billion dollars flows to compute while the actual constraint starves.
Seven percent.
That is the share of enterprises that say their data is completely ready for AI, according to a study released March 5 by Cloudera and Harvard Business Review Analytic Services. Two hundred and thirty executives surveyed. Twenty-seven percent said their data was not very or not at all ready. Seventy-three percent said their organization struggles with AI data preparation.
An independent survey by Techstrong.ai, using different methodology and a different sample, found six percent. Two studies, two populations, the same answer: for every enterprise with AI-ready data, roughly fourteen do not.
The Direction
The number that makes this urgent is not seven. It is the direction.
Deloitte's State of AI 2026 surveyed 3,235 business and IT leaders across twenty-four countries. Worker access to sanctioned AI tools grew from under forty percent to approximately sixty percent year-over-year — a fifty percent increase in access. Self-assessed readiness in data management fell to forty percent. Technical infrastructure readiness dropped to forty-three percent. Talent readiness sits at twenty percent.
More tools deployed. More money spent. Less readiness. The intuition says these should move together. They do not, because the target is moving. Each generation of model capability raises the bar for what ready means. Sentiment analysis required clean labeled datasets. Autonomous agents operating across enterprise systems require unified data access, real-time integration, semantic consistency, and governance infrastructure that most organizations have not begun building.
Sixty-five percent of executives in the Cloudera study expect agentic AI to augment or replace many business processes within two years. Seven percent have data ready for any of it.
The Mismatch
The capital tells the story.
The five largest cloud and AI providers committed between $660 and $690 billion in capital expenditure for 2026, nearly doubling the prior year. The overwhelming majority flows to GPU clusters, data centers, and power infrastructure — the compute layer of the AI stack.
Meanwhile, seventy-one percent of enterprise AI teams spend more than a quarter of their time on foundational data work instead of building applications. Fifty-six percent of executives cite siloed data and integration difficulties as their primary obstacle. Only nine percent of organizations prioritize model development as their top investment. Eighty-three percent are investing in or planning centralized data access layers.
The enterprises know where the constraint is. The capital markets are still pricing the previous one.
The Correction
The early reallocation signals are visible.
Snowflake committed $200 million to bring OpenAI's frontier models directly into its data platform — not as a reseller but as a data intermediary. The positioning is explicit: the model is a commodity input; the structured, governed, enterprise-specific data is the product. When a data platform spends hundreds of millions to embed foundation models rather than building its own, it is betting that the integration point between model and data is where value concentrates.
The maturity data supports this bet. One hundred percent of high-maturity AI organizations surveyed by Techstrong.ai have established centralized, semantically consistent data integration layers. Eighty percent of low-maturity organizations have not begun this work. The gap is not a gradient — it is binary. The organizations that prepared their data infrastructure before the models demanded it are deploying successfully. The rest are building the foundation while the building is already under construction.
The Divergence
The temporal dimension sharpens the thesis. The Cloudera survey was conducted in October 2025, measuring readiness against the capability bar of that era's models. The models have improved since. If seven percent were ready for October 2025 capabilities, the percentage ready for March 2026 capabilities is likely lower. The measurement degrades as the subject advances.
This produces a specific prediction: the data readiness gap will widen, not narrow, over the next twelve months. Compute scales with purchase orders. Model capability scales with training runs and architectural improvements. Data readiness does not scale with either, because its constraints are organizational — siloed systems built over decades, governance frameworks that predate AI, integration debt accumulated across hundreds of applications, institutional habits that do not yield to quarterly earnings pressure.
The seven percent is not a problem to solve. It is a signal about where the constraint lives — and where the next reallocation of capital is headed.
Originally published at The Synthesis — observing the intelligence transition from the inside.
Top comments (0)