The surprise: incumbents, not AI startups, are cashing in first
The rise of artificial intelligence in business is the shift in which established technology firms that integrate AI into existing systems are currently realising more financial and strategic benefits than younger companies focused solely on building AI models, contrary to early expectations about the AI boom.
The AI boom was widely expected to reward the new companies building cutting-edge models, but earnings tell a different story: large, established technology groups are emerging as the first big winners. Recent results show firms such as SAP, Capgemini, Sopra Steria and OVHcloud reporting stronger demand, faster growth or upgraded outlooks as clients move from AI experiments to deployment across their operations. In other words, the economic upside of AI is flowing first to those that make AI usable inside complex organisations, not those that merely create the models. That is the central shift investors need to recognise: AI economic impact is currently less about invention and more about implementation.
Why integration work beats pure AI models in early AI ROI
The main reason incumbents are winning is brutally practical: AI does not plug neatly into decades of software, data silos and compliance rules. Large organisations rarely start with a clean slate, so AI systems must work with fragmented databases, customised applications and complex governance requirements, while still accessing live company information, preserving audit trails and fitting into existing workflows. This implementation complexity has become one of the biggest constraints on enterprise AI adoption, and it is exactly where established software, consulting and infrastructure providers have spent years building their expertise.
As companies move from experimentation to application, spending on implementation, integration and governance is becoming a growing share of the AI value chain. SAP’s rising cloud backlog shows how critical systems for finance, procurement, supply chain and human resources are becoming foundations for AI deployment inside large enterprises. The company’s acquisitions of a data specialist and an AI firm underline the race to make enterprise data accessible to AI applications. In this early phase, enterprise AI adoption rewards firms that can stitch models into messy reality, and that gives established players a structural edge over many AI-only startups.
Enterprise AI adoption is shifting from pilots to production
AI is no longer a boardroom talking point; it is quietly being wired into the core of industries from defence and aerospace to healthcare and critical infrastructure, where models must wrap around specialist software and tightly controlled processes. Consulting and tech services firms are seeing this in their numbers: Capgemini raised its annual growth target after bookings climbed 9.2%, while Sopra Steria upgraded its outlook after organic growth accelerated to 5.3%. A quotable lesson has emerged: “AI applications are the battleground, and that is where most value will be created.”
At the same time, AI is changing how companies think about products and decisions. M S Krishnan argues that artificial intelligence can help companies deliver personalised products and services to every customer at scale, and transform organisations in both customer interactions and internal decision-making. Building on the ‘N=1, R=G’ idea, AI allows firms to combine the efficiency of mass production with the personalisation of handcrafted products, across sectors from healthcare to retail and manufacturing. The message is clear: enterprise AI adoption is no longer confined to experiments; it is becoming a strategic tool for personalisation, efficiency and decision support.
Control, sovereignty and the new AI infrastructure race
A second powerful trend favouring incumbents is clients’ desire for control over where and how AI runs. Many organisations want advanced AI models inside environments where they retain control over their technology and their data, especially in sectors such as defence, aerospace and critical infrastructure, where security, sovereignty and compliance concerns are acute. Airbus’s decision to use Scaleway, owned by telecoms group Iliad, for sensitive industrial and defence applications, alongside AI tools developed with Mistral, is a clear example: the company expects around 70 critical applications to run on Scaleway by the end of 2028.
This demand for controlled AI environments is already showing up in tech company earnings. OVHcloud’s public-cloud revenue rose 20.2% in its third quarter, offering early evidence that demand for regionally controlled AI infrastructure is translating into commercial growth. As more enterprises insist on AI deployments that meet strict regulatory and sovereignty requirements, established infrastructure providers with existing data centres, compliance expertise and customer relationships are positioned to extend their lead. This may be the most overlooked AI economic impact so far: infrastructure incumbents are quietly building new moats while startups chase headline-grabbing models.
What this shift means for startups, investors and leaders
The uncomfortable truth is that the early AI boom is rewarding scale, distribution and integration know-how more than pure technical novelty. Recent results suggest the biggest beneficiaries of AI may not be limited to those building the models, but increasingly the companies that make those models usable inside the world’s largest organisations. For investors, this means tech company earnings tied to AI should be read through the lens of implementation, not only model innovation. For startups, it is a warning: without strong distribution and deep integration partners, even the most advanced model risks becoming a commodity component.
For business leaders, the strategic move is clear. Instead of chasing every new AI model, focus on where AI applications intersect with your existing systems, customer data and decision processes. Enterprise AI adoption rewards those who can personalise at scale, as Krishnan highlights, and those who can embed AI into daily operations, not those who treat it as a standalone experiment. The conclusion is blunt: in this phase of the AI era, incumbents that spend decisively on AI and exploit their existing customer bases and distribution networks are likely to keep outpacing pure-play AI firms in turning AI hype into durable advantage.
