The economics of our current business models are threatened by three systemic fault lines, which are forcing companies to rethink not only how they deploy technology such as Generative AI, but why, says Yves Nicolas, AI Group Program Director and Deputy Group CTO of Sopra Steria.
In 1987, the Nobel laureate economist Robert Solow coined what would become one of the most famous paradoxes of the digital age: “You can see computers everywhere but in the productivity statistics.” Nearly four decades on, the pattern is repeating itself. Generative AI is ubiquitous, yet its tangible impact on productivity remains invisible in macroeconomic data.
This new Solow paradox is, however, only the tip of the iceberg. Below it lies a much deeper problem. The unprecedented convergence of three systemic threats, each of which, taken in isolation, is enough to call into question the very foundations of our economic models.
The Solow paradox, 2026 edition
The figures are stark. According to a study by the National Bureau of Economic Research published in February 2026, based on nearly 6,000 business leaders in the United States, the United Kingdom, Germany and Australia, 89% report no impact from AI on their productivity over the past three years.
Yet 69% of the companies surveyed already use AI, while global investment in data centres reached $500bn in 2024, according to the International Energy Agency.
This gap between promise and reality is not new. The previous major technological wave – the so-called “digital transformation” of the late 2010s – also produced its share of disappointments. According to analysis by McKinsey & Company and Gartner, 70%-80% of digital transformation projects failed to meet their objectives. What if generative AI follows the same trajectory?
Yet reducing the debate to the return on investment of AI alone would be a mistake. What we are witnessing is the unprecedented convergence of three “systemic fault lines” – three structural flaws, each capable, on their own, of derailing our entire economic model.
Fault line No. 1: AI erodes barriers to entry in the IT industry
Financial markets have already priced in a radical assumption – generative AI makes software development so accessible that barriers to entry for IT services companies and software vendors are collapsing. The immediate consequence has been a marked decline in valuations across the sector.
At the same time, procurement departments are exerting tangible pressure. They are demanding price reductions on the grounds of AI-driven productivity gains, even though such gains remain largely unproven. Suppliers, facing commercial pressure, are committing to ever more ambitious targets. What happens when delivery is due?
In reality, barriers to entry remain substantial. Generating code is one thing; integrating complex systems, managing decades of accumulated technical debt and ensuring the reliability of critical infrastructure is quite another. But market perception has already settled the matter – and that perception has very real effects.
Fault line No. 2: shifting alliances fracture the digital supply chain
For decades, Europe’s IT industry has accepted a high degree of dependence on major US software and infrastructure providers. The figures are telling. According to Synergy Research Group, European cloud providers accounted for just 15% of their own market in 2025, while Amazon, Microsoft and Google together control more than 70%.
A report by the European Parliament published in late 2025 estimates the EU’s digital trade deficit at over €100bn a year, with roughly €264bn annually – 1.5% of European GDP – flowing to foreign cloud and software providers.
This dependence was the implicit price of technological convenience but recent geopolitical upheavals are turning it into an operational vulnerability. The most striking illustration came from the International Criminal Court. Following US sanctions, its chief prosecutor found himself cut off from his email account hosted by an American provider. The ICC subsequently announced its migration to a sovereign European suite.
The European market is responding. According to Gartner, European investment in sovereign cloud is expected to rise from $6.9bn in 2025 to $23.1bn in 2027. Beyond cloud, however, every company must now ask themselves, “within my information system, what am I dependent on, and what level of risk am I prepared to accept?”
Fault line No. 3: the underpricing of climate risk – the primary threat
This third problem is, in reality, the source from which the other two flow. It is also the most underestimated in current economic models.
Research by the Network for Greening the Financial System, relayed by the European Central Bank, is explicit. It shows that a sequence of extreme climate events could trigger a contraction in eurozone GDP of up to 5% by 2030 – a shock comparable in magnitude to the 2008 financial crisis. And this scenario does not account for cascading effects from disruptions to global supply chains.
What, then, is the industry doing in response? It is investing heavily in large-scale AI whose energy footprint is staggering. According to the International Energy Agency, global electricity consumption by data centres could double to reach 945 TWh by 2030. That’s equivalent to Japan’s annual consumption.
According to The Shift Project, the digital sector already accounts for 3 to 4% of global greenhouse gas emissions, growing at an average rate of 6% per year.
AI at the scale currently envisaged – based on ever-larger large language models and ever more numerous data centres – is not compatible with a trajectory limiting warming to below 2°C. Just as there is no viable path for aviation without restraint, there is no viable path for AI without digital sobriety.
Three investments for a sustainable model
Faced with this threefold diagnosis, denial is tempting. Yet the convergence of these three fault lines also outlines, by implication, a coherent investment strategy for IT players willing to confront them.
Investing in right-sized AI
Rather than pursuing an arms race in ever-larger models, the priority is to invest in targeted AI systems tailored to real business needs and consuming significantly fewer resources. AI as a knowledge-management tool holds considerable promise – provided efforts are focused where value can be demonstrated, rather than deploying technology by default. Research into frugal models capable of running on lightweight infrastructure thus becomes a strategic necessity.
From consumer to contributor in open source
High-quality, shared open source is the most effective vehicle for independence from critical dependencies. Initiatives are multiplying. In France, La Suite, developed by DINUM, offers a sovereign alternative to major proprietary collaboration environments for the public sector.
In Germany, openDesk, backed by the Centre for Digital Sovereignty (ZenDiS), is already deployed within the Bundeswehr and the federal health ministry. The state of Schleswig-Holstein is saving €15m annually after migrating 80% of its workstations to LibreOffice.
At the European level, Germany, France, Italy and the Netherlands created a consortium for digital commons in 2025, while the OpenBuro initiative, launched at FOSDEM 2026, is working on interoperability between these tools. For IT services companies, moving from passive consumption of open source to active contribution is both an act of sovereignty and an investment in resilience.
From input-based to outcome-based commitments
The move from input-based to outcome-based commitments may be the most profound shift. It entails supporting clients towards right-sized information systems, anchored in real business value. Working on an outcome basis rather than selling billable time implies accepting a degree of contraction in technological scope. Yet it is also what enables a reduction in “run” costs – the primary cost centre for clients – and aligns the interests of provider and customer.
Sovereignty and sustainability: two sides of the same coin
The encouraging news is that the stars are aligning. Digital sovereignty and environmental sustainability converge on the same answers: lighter, more controlled, less dependent systems. A frugal information system is, by definition, more sovereign – less exposed to geopolitical shocks and less costly to operate.
For decades, the industry has been fixated on the “how”. How to deploy a given technology, how to integrate a given tool, how to put AI everywhere. The convergence of these three issues forces a return to a more fundamental question: why. Why this information system? For what real business need? With what acceptable footprint?
The path is not an easy one. But it is the only viable route towards an economic model that does not rest on climate denial, technological overreach or strategic dependencies that recent events have rendered untenable. The companies that move first will not merely be acting responsibly; they will – and are already – defining the rules of the game for the decades ahead.