# Tech Show Paris preview: Why AI is forcing retailers to rethink the infrastructure beneath every decision > Source: > Published: 2026-09-14 09:02:24+00:00 **As AI moves from pilot projects into day-to-day retail operations, [Tech Show Paris](https://www.techshowparis.fr/en/) conference manager Jaya Sexton explains why cloud, data, energy and governance are becoming board-level infrastructure questions.** Over recent years retailers have talked about AI as a customer experience story: smarter recommendations, faster service, better forecasting and more personalised engagement. But as the technology becomes more embedded in live operations, the bigger question is shifting from what AI can do to whether retailers have the infrastructure to support it. “Retail has always generated huge volumes of data, but AI changes the expectation placed on that data,” says Sexton. “It now needs to be available,trusted and usable at the moment decisions are being made — whether that is online, in-store, in a warehouse or across the supply chain.” That requirement is already influencing decisions around compute, storage, networking and cloud. Retailers need capacity to run AI models, analyse data at speed and support experimentation as well as live use. They also need storage that keeps data accessible without becoming unmanageable, and networks capable of connecting stores, warehouses, ecommerce platforms and customer systems. ###### From cloud migration to workload strategy According to Sexton, the debate is no longer simply about whether retailers should move more systems to the cloud: “Most retailers are no longer asking:‘cloud or on-premise?’- They are asking: which workload belongs where.” A customer-facing AI assistant, demand forecasting model, in-store analytics tool and supply chain optimisation platform will not have identical requirements. Some will need to scale quickly; some will need real-time response; some will involve more sensitive data; and some will be more cost-effective in one environment than another. “The infrastructure decision is increasingly about matching the workload to the right environment, rather than applying one model across the business,” Sexton says. ###### The real constraint is efficient scale As AI and data-intensive workloads grow, operating constraints are becoming harder to ignore. Sexton says the challenge is not simply whether organisations can scale AI, but whether they can do so efficiently. She explains: “AI increases pressure on compute capacity, network performance, storage architecture and data centre power. At the same time, retailers are operating on tight margins, so infrastructure decisions have to be commercially disciplined.” The pressure to innovate with AI is intensifying, but over-investing in infrastructure can quickly create a cost base that is difficult to sustain. “The question becomes: how do we support AI growth without creating a cost model or energy profile that is unsustainable?” Sexton says. The organisations moving fastest, she argues, are not simply buying more infrastructure. They are becoming more precise about where workloads run, what level of performance they require and how costs are monitored over time. “Some AI use cases justify high-performance infrastructure. Others need lighter-weight models, better data pipelines or more efficient inference,” she notes. “The risk is treating every AI workload as if it needs the same architecture. We need systems that are hybrid by design and bespoke to the needs of each company ecosystem.” ###### Retail examples on the Tech Show Paris stage Sexton sees this shift is reflected across the Tech Show Paris programme, where AI is being discussed less as a standalone technology and more through the infrastructure decisions organisations now have to make around data, cloud, cost, sovereignty, energy and operational resilience. “What stands out is that AI is not being treated as a standalone technology theme. It is being discussed through the infrastructure decisions that organisations now have to make,” she says. A panel discussion from Carrefour is one of the clearest examples. Speakers from its digital factory and data platform will explore how AI-driven retail depends on far more than deploying models. It relies on how organisations structure data platforms, how engineering teams build and ship software, and how AI becomes usable across commercial, operational and customer-facing environments. Another session featuring the global head of analytics at La Redoute will examine why many AI projects fail to reach production. This is where the infrastructure conversation becomes most relevant to retail leaders, says Sexton. “Moving from AI pilots to operational AI requires reliable data flows, scalable platforms, clear governance and systems robust enough to support live commercial decisions,” she adds. Cloud strategy sessions will look at where AI should run, how it should be financed and how organisations avoid turning cloud adoption into long-term dependency. Meanwhile, Data Centre World sessions will bring in the physical realities of AI demand, including colocation, hyperscale, energy, cooling and connectivity. “AI strategy ultimately becomes a set of practical decisions about capacity, power, resilience, cost and control,” Sexton says. ###### Infrastructure becomes a strategic control layer Taken together, the conversations at Tech Show Paris suggest enterprise infrastructure is becoming more intentional. The old direction of travel was relatively simple: move more to the cloud, collect more data and experiment with AI. That, Sexton says, is no longer enough. “Organisations are now asking where each workload should run, how data should be governed, how much dependency they are willing to accept, and how infrastructure choices affect cost, resilience, compliance and energy use,” she says. For retail technology leaders, her advice is to focus on four priorities: data readiness, workload placement, cost and energy discipline, and governance by design. “AI can only scale if organisations have trusted, accessible and well-governed data,” Sexton says. “Scaling AI is not just a capacity question; it is a financial and physical infrastructure challenge, from cloud consumption to power, cooling and data centre availability.” Governance must be built into infrastructure decisions from the outset, particularly as AI moves into live retail environments: “Regulation, cyber risk, sovereignty and resilience need to be built into infrastructure decisions from the start, not added after AI projects move into production,” Sexton concludes. *Tech Show Paris will mark its 10th anniversary on November 18 and 19, 2026, at* *Paris Expo Porte de Versailles. Find out more [here](https://www.techshowparis.fr/en/).* [Click here to sign up to Retail Gazette‘s free daily email newsletter](https://www.retailgazette.co.uk/newsletter-sign-up-retail-gazette?utm_source=Website&utm_medium=Article&utm_campaign=SignUp)