Global mergers and acquisitions are entering a new phase—one where the “asset” being bought is less a product line and more the plumbing of the digital economy. Across energy, fibre networks and compute, dealmakers are increasingly treating infrastructure as strategic leverage for the AI era. The result is a market that looks familiar on the surface—corporate combinations, private equity roll-ups, cross-border consolidation—but behaves differently underneath. Transactions are moving faster, due diligence is becoming more technical, and the logic of value creation is shifting from growth forecasts to control of bottlenecks.
In earlier cycles, investors often chased scale: more customers, more distribution, more market share. Today, many of the most consequential deals are about capacity—power that can reliably run data centres, connectivity that can move data with low latency, and compute that can be provisioned at the speed AI systems demand. This is not simply “AI spending.” It is the physical and regulatory reality of building and operating the systems that make AI useful at scale.
Energy: the new constraint is reliability, not just generation
Energy has always been part of industrial strategy, but the AI-driven wave has changed what matters. For years, the conversation around power focused on generation and cost. Now, the centre of gravity is reliability and grid integration. Data centres are not just large electricity users; they are sensitive to interruptions, voltage fluctuations and constrained transmission. That makes the ability to secure firm power—power that can be delivered consistently—an asset in its own right.
Deal activity reflects this shift. Buyers are looking at utilities and grid operators, but also at the ecosystem around them: generation portfolios with dispatchability, transmission assets that reduce bottlenecks, and services that help manage load. In some markets, the most valuable targets are not the biggest plants but the ones best positioned to connect quickly to high-demand zones. The “time to interconnect” has become a hidden variable in valuation models. A project that can come online in 18 months may be worth more than one that is cheaper but delayed by permitting or grid upgrades.
This is why energy deals increasingly resemble infrastructure finance rather than classic corporate acquisition. Buyers want control over the timeline and the operational characteristics of power delivery. They also want optionality: the ability to expand capacity as AI workloads grow, without renegotiating every constraint from scratch.
The energy transition adds another layer. Many jurisdictions are simultaneously pushing for decarbonisation and struggling with the pace of grid build-out. AI demand arrives at the same time as electrification of transport and industry, which means competition for grid capacity is intensifying. In that environment, acquiring assets that can accelerate grid expansion—or that can be paired with storage, demand response and flexible generation—becomes a way to de-risk the AI supply chain.
There is also a geopolitical dimension. Energy security is now tied to digital security. When compute depends on power, and power depends on infrastructure that crosses borders, the strategic calculus changes. Governments that once treated energy as a domestic matter increasingly view it through the lens of national resilience. That can influence approvals, ownership structures and the terms under which foreign capital participates.
Fibre networks: the race is for reach, not just bandwidth
If energy is about reliability, fibre is about reach and performance. AI workloads are data-hungry, but the bottleneck is not only raw bandwidth. It is the combination of bandwidth, latency, redundancy and the ability to scale connections without long lead times. Fibre networks are therefore being valued not merely as telecom assets but as enabling infrastructure for cloud regions, edge computing and enterprise connectivity.
The most interesting deals are often those that look modest compared with headline-grabbing mega-mergers. Instead of buying entire national carriers, acquirers may target regional networks, dark fibre portfolios, metropolitan rings, or wholesale capacity agreements that can be converted into service quickly. The logic is straightforward: AI deployment is uneven. Demand clusters around specific data centre corridors, industrial parks and urban nodes. Networks that already have the physical routes and rights-of-way to serve those clusters can compress timelines.
Another factor is the shift toward hyperscale and multi-cloud architectures. Enterprises and AI developers increasingly distribute workloads across providers, which increases the need for interconnection ecosystems. Fibre assets that sit near exchange points, cloud campuses and carrier hotels can become strategic because they reduce dependency on any single provider’s internal network. In practice, that means buyers are paying for “network adjacency”—the ability to connect compute to customers and partners efficiently.
There is also a subtle but important trend: fibre is becoming a platform for services, not just a conduit. Some acquisitions bundle connectivity with managed services, cybersecurity capabilities, and network orchestration tools. That matters because AI systems are not standalone; they require secure data movement, identity management, and compliance-aware routing. Connectivity that can be integrated into these workflows is more valuable than connectivity that simply delivers throughput.
Regulation shapes the deal landscape here too. Telecom markets are often subject to ownership limits, access obligations and scrutiny around competition. As a result, transactions may be structured with regulatory commitments, open-access requirements or carve-outs. Buyers must demonstrate that they can maintain service quality and fair access while still achieving their investment thesis.
Compute: the asset is capacity, but the real prize is provisioning speed
Compute is the most obvious part of the AI story, yet it is also the most misunderstood. Many people think of compute as chips. But in M&A, the most valuable compute-related assets are frequently the ones that determine how quickly and reliably compute can be delivered to customers.
Data centres are central, but the deal universe extends beyond buildings. It includes power distribution equipment, cooling systems, interconnection facilities, and the software layers that manage capacity allocation. It also includes the supply chain around chips and servers—sometimes directly, sometimes indirectly through long-term contracts and partnerships.
The unique feature of AI compute is that demand is spiky and rapidly evolving. Training runs can be scheduled, but inference and fine-tuning often require continuous capacity. That means buyers want flexibility: the ability to scale up and down, to support different hardware generations, and to integrate new accelerators without major downtime.
This is why compute deals increasingly focus on “future-proofing.” Acquirers evaluate whether a facility can support higher power densities, whether it has sufficient cooling capacity, and whether it can be upgraded to new rack configurations. They also assess whether the site has strong connectivity to multiple upstream carriers and cloud providers. In other words, the value is not only in the square footage; it is in the facility’s ability to remain relevant as AI workloads change.
Cloud capacity and colocation are also converging. Some buyers pursue assets that allow them to offer hybrid services—combining owned infrastructure with contracted capacity. That can be attractive when demand is uncertain but the strategic need for presence is clear. In such cases, the acquisition is less about immediate utilisation and more about securing a position in a market where future demand is expected to be structurally higher.
Then there is the question of chips supply. While chip manufacturing is often dominated by government-backed industrial policy and long-term procurement, M&A can still influence the ecosystem. Buyers may acquire companies that provide packaging, testing, networking hardware, or specialised components that reduce friction in deploying AI systems. Even when the transaction is not directly about semiconductor fabs, it can still affect the speed at which compute becomes usable.
A unique take on valuation: the “bottleneck premium”
Traditional valuation methods—multiples of revenue, discounted cash flow based on historical margins—struggle when the key driver is bottleneck control. In the current cycle, many deals implicitly price a “bottleneck premium”: the value of being able to deliver scarce capacity sooner than competitors.
That premium shows up in several ways. Buyers may accept lower near-term utilisation if they believe they can secure long-term contracts. They may pay more for sites with faster permitting or for networks with existing rights-of-way. They may prioritise assets that reduce interconnection delays even if the purchase price is higher.
This is also why dealmaking can appear irrational to outsiders. A fibre network might not look like a high-growth business on paper, and an energy asset might not look like a pure-play AI beneficiary. But if the asset unlocks the ability to serve AI customers at scale, the economics change. The buyer is not purchasing a telecom or utility business; it is purchasing a lever that converts demand into revenue.
The bottleneck premium also affects how risk is allocated. Sellers may negotiate earn-outs tied to capacity milestones. Buyers may structure deals with staged payments contingent on regulatory approvals or infrastructure build-out. In some cases, the transaction becomes a partnership between capital and engineering execution, with governance designed to ensure that the acquired assets actually translate into capacity.
Why deals are more globally coordinated
The infrastructure race is inherently cross-border. Energy resources, fibre routes, and compute supply chains do not align neatly with national boundaries. Yet the strategic nature of these assets means governments are more involved than in prior cycles.
As a result, global coordination is increasing in two directions. First, capital is moving to where infrastructure can be built or expanded. Second, governments and regulators are coordinating standards and review processes, especially for assets tied to critical digital infrastructure. That can slow some transactions, but it also creates a more predictable framework for others.
Cross-border deals are also shaped by the need for interoperability. Fibre networks must connect to international backbones. Data centres must integrate with global cloud and carrier ecosystems. Compute supply chains depend on components sourced from multiple regions. Buyers therefore seek assets that fit into a broader architecture, not isolated local operations.
This is where the “brave new world of deals” framing becomes more than a metaphor. The deal is no longer just about owning an asset; it is about owning a node in a networked system. That system spans continents, and the value of each node depends on how well it connects to the rest.
Private equity and sovereign capital: different playbooks, same objective
The infrastructure race is not limited to strategic corporate buyers. Private equity firms and sovereign wealth funds are also active, often with different time horizons and risk tolerances.
Strategic buyers