Manufacturers aren’t leading the AI race by volume, but that may be the point. According to a new benchmark study from S&P Global Market Intelligence, produced in collaboration with Vultr, manufacturers report lower internal barriers to AI maturity despite having the smallest share of organizations describing themselves as “Transformational” in their use of the technology.
Only 31% of manufacturers place themselves in that top maturity tier, the lowest proportion among sectors surveyed. Yet across skills, data readiness, security, and organizational alignment, manufacturers consistently report fewer obstacles to scaling AI than their peers. The findings suggest a sector that is trading speed for structure, and experimentation for consolidation.
Slow Maturity, Lower Friction
The study defines AI maturity across three stages: Operational, where early functional benefits emerge; Accelerated, where AI is deployed across multiple functions; and Transformational, where AI is embedded directly into core operations. Today, 25% of manufacturers remain in the Operational phase, compared with 19% across all industries, while just under a third have reached the Transformational level.
On the surface, those numbers point to lagging adoption. But the underlying barriers tell a different story. Skills shortages are cited by 46% of manufacturers, versus 62% across other sectors. Data quality concerns register at 50% for manufacturers, compared with 60% elsewhere, while security challenges stand at 49% versus 60%. Even softer constraints, such as leadership alignment and organizational culture, are reported at materially lower levels.
According to the researchers, that gap reflects a more disciplined approach. Rather than chasing scale through rapid model proliferation, manufacturers have focused on platform engineering, governance, and integration first. The report notes that unifying data and infrastructure before expanding use cases appears to be reducing friction later in the adoption curve.
Platform Control and Model Consolidation Take Hold
That discipline is also visible in infrastructure decisions. Around 35% of manufacturers have already built, or are in the process of building, their own internal platform-as-a-service (PaaS) environments. Within two years, that figure is expected to rise to 45%. At the same time, reliance on hyperscaler-managed PaaS platforms is projected to fall from 56% to 42%.
Manufacturers continue to make heavy use of public cloud resources, running roughly 30% of AI training workloads and 28% of inference on major cloud platforms, both well above industry averages. But the trajectory points toward greater internal control, particularly as AI systems move closer to production planning, robotics coordination, and real-time operational decision-making.
Perhaps the clearest signal of consolidation is the shrinking model footprint. The average number of AI models in production among manufacturers is expected to decline from 242 today to 189 next year. Rather than a retreat, the study frames this as a rationalization effort, retiring marginal models in favor of fewer systems with proven operational value.
Discipline Becomes a Structural Advantage
As AI moves closer to the machinery, workflows, and physical constraints of production, the limiting factor shifts from experimentation to reliability. Systems that feed scheduling engines, quality controls, or autonomous equipment cannot tolerate fragmented data or loosely governed models without introducing operational risk. Recent industry analysis shows that manufacturers who reduce model sprawl and keep tighter control over platforms are better positioned to absorb AI into day-to-day execution, not as an overlay but as infrastructure. Over time, that steadier integration may shape performance outcomes more than early headline adoption ever did.