Many industries are rushing to scale AI: standing up new pilots, expanding model inventories, and racing toward what they consider "advanced" maturity. Manufacturers are taking a different route. They're engineering their way into AI, not sprinting toward it.
New benchmark data from S&P Global Market Intelligence, in collaboration with Vultr, reveal a striking paradox. While only 31% of manufacturers describe themselves as "Transformational" in their AI adoption – the lowest share among sectors surveyed – they consistently report lower barriers to reaching that maturity. Where others struggle with organizational chaos, manufacturers face engineering constraints they know how to solve.
Solving for infrastructure, not culture
The benchmark study categorizes AI maturity in three stages: Operational (early functional gains), Accelerated (AI across multiple functions), and Transformational (AI embedded into core operations). Currently, 25% of manufacturing respondents remain Operational compared to 19% across all industries, and only 31% have reached Transformational maturity.
But across nearly every organizational barrier – skills shortages, data quality, security, cultural alignment, and more – manufacturers report materially lower severity. The skills shortage register stands at 46% for manufacturers, compared to 62% for all other respondents. Data quality: 50% versus 60%. Security: 49% versus 60%. Even leadership alignment and company culture come in 13 points lower.
Manufacturers identify technical constraints to scaling AI, with compute capacity, storage connectivity, and data pipelines ranking highly. However, in contrast to the challenges facing other industries, these are well-understood engineering limits with well-understood solutions, rather than abstract organizational blockers.
This creates a fundamentally different relationship with AI maturity. Where other sectors wrestle with alignment or building the right culture, manufacturers are asking: Do we have enough GPU capacity? Are our data pipelines fast enough for real-time inference? Can we connect plant-floor systems to analytics platforms without latency?
Those are solvable problems. Manufacturers have spent decades solving these problems in other contexts.
Governance before scale: The rise of internal platform engineering
Perhaps the clearest signal of manufacturing's systems-first approach is the rapid growth of internal Platform-as-a-Service (PaaS) environments. Today, 35% of manufacturers have built or are in the process of creating internal PaaS infrastructure. Within two years, the figure is projected to reach 45%, while reliance on hyperscaler-managed PaaS is expected to decline from 56% to 42%.
This isn't a migration away from hyperscalers; manufacturers still run 30% of their training workloads and 28% of their inference on major public clouds, both figures well above the industry averages. But it does represent a fundamental redesign of how manufacturers govern and orchestrate AI.
Internal platforms give manufacturers unified governance across design, production, and analytics systems. They create consistent model lifecycle management and clean data pipelines connecting plant-floor sensors to enterprise analytics. Most importantly, they establish control points where engineers verify that AI workloads meet the same reliability and safety standards as other production systems.
Robotics, computer vision systems, predictive maintenance workflows, and digital twins all require real-time coordination between edge devices, cloud compute, and on-premises systems. Internal platforms create the architectural coherence to make that integration work.
Consolidating models to scale what works
The average number of AI models in production among manufacturers has declined from 242 today to a projected 189 next year. In an environment where "more AI" is often synonymous with "better AI," manufacturers are moving in the opposite direction.
Manufacturers are consolidating around proven applications: robotic process automation, computer vision for quality assurance, predictive maintenance, energy management, and autonomous vehicle systems. These deliver direct operational improvements in the form of reduced downtime, fewer defects, and lower energy costs.
The pattern reflects an outcome-first mindset. Rather than maintaining sprawling experimental portfolios, manufacturers invest in infrastructure and talent to productionize fewer models at higher quality. That means better monitoring, tighter integration with production systems, and more transparent accountability for business outcomes.
It also points to increased standardization. As internal platforms mature, they create pressure to rationalize model architectures, consolidate MLOps tooling, and reduce technical debt. Fewer models means less fragmentation, making it easier to enforce governance and ensure security across global operations.
When manufacturers say they expect to reach Transformational maturity within two years, this is what they mean: not a proliferation of experiments, but fewer AI capabilities deeply integrated into operations. It’s a shift from volume to value.
The long game
Other industries treat AI adoption as a software problem: Iterate fast, deploy often, learn through experimentation. Manufacturers treat it as a systems problem: design for reliability, integrate across domains, and validate before scaling.
The data shows what that discipline produces: Lower perceived barriers to maturity. Platform engineering that unifies governance before chasing scale. Model consolidation that prioritizes proven value over portfolio breadth.
This isn't the fastest path to Transformational AI, but it may be the most durable. As simulation-driven production, robotics coordination, and real-time analytics become table stakes, manufacturers building AI on integrated platforms will have the foundation to scale sustainably.
For a deeper look at how manufacturers are building AI maturity through platform engineering, GPU optimization, and governance-first infrastructure, explore the full report, “Unlocking the Power of AI in Manufacturing.”