For the past two decades, the prevailing narrative in corporate strategy has been clear: digital technologies favor specialization, asset-light models, and ecosystem thinking. Companies were told to focus on their core, outsource the rest, and leverage platforms to scale quickly. And for a time, that approach delivered extraordinary results.
But something curious is happening.
Tech giants that once epitomized lean, modular strategies are now doubling down on physical infrastructure. Meta is building subsea cables and designing its own AI chips. Amazon is investing in nuclear power to secure long-term energy supply for its data centers. Uber, a pioneer of platform-based growth, has spun out a data-labeling operation to secure control over the inputs that feed its AI models.
These moves aren’t isolated anomalies. They suggest a deeper strategic shift. Is vertical integration, long considered a relic of industrial strategy, staging a comeback in the digital age?
Recap: The First Wave of Digital Strategy
In the early years of digital transformation, the logic of value creation was reshaped by two forces: the dramatic reduction in transaction costs and the scalability of cloud-based infrastructure. The result was a strategic playbook that favored openness over ownership.
Platform models thrived. Rather than owning assets, successful firms orchestrated networks—connecting riders to drivers, hosts to guests, and advertisers to users. Value was created at the interface, not in the supply chain. Meanwhile, enterprise architecture followed suit: software development was modularized, compute was outsourced to hyperscale cloud providers, and infrastructure became a utility.
This shift also changed how we thought about competitive advantage. Instead of vertical integration, the emphasis was on ecosystem leverage. Companies competed on speed, data flywheels, and product innovation—not on capital expenditure or control of physical inputs.
It is precisely this context that makes today’s moves so striking. Why are the very firms that scaled through specialization now re-integrating upstream? What has changed in the economics of digital strategy?
The New Moves: Case Studies of Strategic Integration
A growing number of digital-native companies are reasserting control over upstream assets, investing not just in software and services, but in physical infrastructure. These are not minor initiatives; they represent a structural redefinition of what digital firms consider “core.”
- Meta has committed billions to regain control over the infrastructure that supports its AI ambitions. It is building a $10 billion subsea cable network to ensure global bandwidth for its services, an essential move given that Meta’s platforms account for over 20% of global mobile traffic. In parallel, Meta has begun testing its own in-house chips to train large AI models, reducing its dependence on third-party silicon. And in a major signal of upstream integration, it plans to invest $15 billion in Scale AI, a data labeling company critical to the training of high-performing models.
- Amazon, long a leader in logistics and cloud, is now positioning itself in the energy supply chain. It recently signed three deals to support the development of next-generation nuclear reactors. These are not just carbon-free sources of power, they’re designed to secure Amazon’s massive data center needs in a future where compute demand is surging and sustainability targets are non-negotiable.
- Uber, once the textbook case of an asset-light digital platform, has quietly spun out its own data labeling unit, Scaled Solutions. What started as an internal team now provides AI training services externally. The move reflects Uber’s growing reliance on AI models and the need to control the labor-intensive, high-cost process of curating training data at scale.
- OpenAI offers a more indirect case of integration pressure. To reduce its dependency on Microsoft Azure, it has signed a $12 billion, multi-year infrastructure contract with CoreWeave, an alternative cloud provider specialized in AI workloads. While OpenAI isn’t building its own data centers, the structure of the deal underscores how control over compute infrastructure is becoming strategic.
Together, these examples reveal a shift in mindset. For AI-driven firms, access to clean energy, compute power, and labeled data are no longer operational considerations, they are sources of competitive advantage.
What’s Driving the Shift? Strategic Hypotheses
This renewed appetite for integration is pragmatic. Several forces are reshaping the digital competitive landscape, making selective control over infrastructure necessary.
- Infrastructure is strategic again. As AI workloads explode in scale and complexity, performance, latency, and cost are increasingly dictated by infrastructure choices. Chips, data centers, and bandwidth are no longer interchangeable inputs, they define the boundary conditions for what is technically and economically feasible. More fundamentally, the economic profile of digital companies is shifting. Once praised for being capex-light, today’s AI-driven firms are investing heavily in physical assets. Building and owning infrastructure, whether it’s data centers, subsea cables, or power generation, offers both cost stability and strategic independence.
- Data as a bottleneck. While much attention is paid to algorithms, the harder problem is often upstream: acquiring, labeling, and maintaining high-quality datasets. The surge in demand for generative AI has exposed a bottleneck in the data supply chain. Companies like Meta and Uber are reintegrating data labeling to ensure a consistent, scalable source of training material, a labor-intensive task that has become critical to model performance.
- Energy as a new competitive variable. AI models are energy-hungry, and cloud infrastructure is now one of the fastest-growing sources of electricity demand. At the same time, regulatory pressure and public commitments are pushing firms to decarbonize. The solution for many? Secure a direct path to clean, large-scale power. Amazon’s nuclear energy investments exemplify how energy access is becoming a strategic asset in the AI arms race.
- Reducing platform dependency. As digital firms become more reliant on AI infrastructure, they also become more dependent on a small number of hyperscalers. OpenAI’s $12B contract with CoreWeave is a defensive play to reduce its reliance on Microsoft Azure. In this context, infrastructure investment becomes a way to maintain bargaining power in increasingly concentrated supply chains.
Together, these drivers point to a common conclusion: infrastructure, once considered a commodity, is now central to digital differentiation.
Rethinking Vertical Integration in the Digital Age
What we’re witnessing isn’t a wholesale return to industrial-era integration. Today’s moves are more targeted, modular, and high-leverage. Rather than rebuilding the full value chain, companies are reclaiming control over its pressure points, where performance, cost, or reliability have become strategic constraints.
In this new model of integration, the goal isn’t ownership for its own sake. It’s about owning what matters most in a world of AI and cloud: chips and data, energy infrastructure, data labeling and collection.
This logic marks a departure from the early digital playbook. Strategy is no longer just about speed and scale—it’s also about resilience, control, and upstream influence.
Implications for Strategy
The return of integration has deep consequences—not just for tech giants, but for any company operating in an AI-enabled world. The terrain of competition is shifting, and with it, the rules of strategic positioning.
For Digital Companies
- Infrastructure is strategic, not incidental. Owning parts of the stack (energy, compute, labeled data) offers both differentiation and insulation from supply chain risk.
- The cost of entry is rising. Scaling in AI no longer means just building better algorithms; it increasingly requires physical infrastructure investments. This reintroduces capital intensity into business models once admired for their asset-light efficiency.
- New sources of power are emerging. Firms with privileged access to energy or compute infrastructure may gain outsized influence, not just over performance, but over who can afford to compete at all.
For Non-Digital-Native Companies
- Prepare for new forms of competition. Tech firms are no longer staying in their lane. When Amazon signs nuclear energy deals or Meta builds transoceanic cables, the lines blur between digital players and infrastructure incumbents.
- Bargaining power is shifting. If key tech providers integrate upstream, they may become less reliant on traditional partners, and more dominant in shaping value chains.
- Access to infrastructure becomes strategic. Whether it’s cloud capacity, clean energy, or compute, non-digital firms will need to secure reliable and cost-effective access. This may mean forming new partnerships, diversifying dependencies, or co-investing in shared assets.
In short: the battle for value is moving upstream. Firms that ignore the infrastructure layer risk becoming dependent on those that don’t.
Conclusion: A New Phase of Digital Strategy
The reintegration of infrastructure by leading digital firms does not signal a return to old industrial models, it marks the next stage of digital strategy. In an economy increasingly defined by AI, cloud, and data-driven scale, the most strategic assets are no longer just algorithms or networks. They are energy sources, chip designs, data pipelines, and transoceanic cables.
What was once outsourced for efficiency is now being pulled back in for control.
This isn’t a rejection of platforms or ecosystems. Rather, it’s an evolution: platform thinking still matters, but it now coexists with selective integration wherever dependency becomes a risk, or performance a constraint.
For strategists, the implications are clear. The boundaries of the firm are no longer just a question of cost or focus. They are a question of strategic leverage.
So ask yourself: in your value chain, which constraint will define your competitiveness tomorrow? And can you afford not to control it?
Photo de Milad Fakurian sur Unsplash.
Generative AI was used for editing this post.
