Large companies have pushed hard on the integration of algorithmic models, but budget issues are coming home to roost. According to an analysis by MarketScale, the 74% of businesses has already brought AI solutions into its production environments, and yet approximately half of the latter do not have the tools to demonstrate the real return on investment. The rush to implementation has clearly disconnected the development of internal metrics, governance systems and integration pipelines with pre-existing infrastructures.
The knot is intertwined with one progressive loss of operational control. The integration of decision-making processes and customer information within external platforms transforms companies into simple tenants of their own knowledge architecture. Without stringent clauses on data portability, audit rights and contractual termination, business logic ends up locked into third-party models, making future migration nearly impossible.
At the level of industrial infrastructure the maneuver is even faster. Giants of the caliber of Schneider Electric And Siemens they signed multi-billion dollar acquisitions to embed native AI capabilities directly into their automation and energy management platforms. Those who purchase these systems risk finding the algorithmic functions already packaged and bound by contracts with hardware manufacturers.
Physical infrastructure and regional markets: new logistical challenges
The difficulties also directly affect the ground. Growing community opposition (the so-called NIMBY syndrome) in urban areas across the United States is hindering the construction of traditional data centers. The big cloud providers are therefore moving construction sites to isolated industrial landslike the Permian Basin in West Texaswhere the energy infrastructure linked to the oil sector and the low population density offer a rapid alternative but with inevitable impacts on latency times and network connectivity.
Added to this logistical reorganization is regulatory fragmentation on a global scale. Apple has released official guidelines to allow Mac users in mainland China to connect the model Alibaba’s Qwen to Siri and the platform’s writing tools. A clear sign of how the management of data sovereignty and local compliance is imposing profoundly different artificial intelligence architectures depending on the geographical region in which one operates.
The audit of contracts with technology suppliers must become the priority of IT departments. Without clear business metrics associated with each individual project and without guarantees on the ownership of the knowledge generated, spending on artificial intelligence risks turning into a fixed cost without accounting justification.
