The language of the data centre industry has changed. A few years ago, capacity was measured in racks and megawatts. Today, the unit of measure is gigawatts — and the ambitions attached to that number are reshaping how the world’s largest technology companies think about infrastructure.
The numbers behind the shift are staggering. The International Energy Agency’s 2026 outlook projects global data-centre electricity consumption will nearly double — from 485 TWh in 2025 to around 950 TWh by 2030 — with AI-focused facilities alone tripling their demand over the same period. Every one of those additional megawatts has to be translated into something physical: racks installed, fibre run, cooling systems commissioned, and millions of individual connections tested until they work as one resilient environment.
Power is only the starting point
Ask Verma what it actually takes to turn a powered shell into a live hyperscale environment, and he’s quick to reframe the question. It isn’t a power problem. It’s an orchestration problem.
“A powered shell becomes operational when thousands of racks and millions of physical connections are installed, integrated, tested, documented and commissioned as one resilient environment,” Verma said. “Behind every additional megawatt is a complex physical environment that must be delivered, connected and operated reliably.”
This is the layer where Black Box has built its business — not as a cabling contractor, but as an integrator of the physical systems that make a facility function. “We don’t simply install cable or supply components,” he explained. “We help customers translate a high-density facility into an operational environment — bringing together inside- and outside-plant fibre, structured cabling, network deployment, distributed antenna systems, building management and security systems, equipment logistics and, increasingly, Day-2 operational support.”
The complexity, Verma argued, isn’t technical so much as programmatic. “These are live, fast-changing programmes involving hyperscalers, general contractors, technology providers and a large specialised workforce, where design decisions can keep evolving even while construction is under way,” he said. “A delay or quality issue in one workstream can affect commissioning across an entire building or campus.”
The bottleneck has moved
For years, the industry narrative around AI infrastructure centred on one scarce resource: GPUs. Verma believes that framing is now incomplete.
“Compute availability still matters, but the challenge is increasingly about converting compute into usable capacity,” he said. “Procuring GPUs is one part of the equation; deploying, powering, cooling, connecting and commissioning them at speed is another.”
On the ground, he’s watching programmes scale from single buildings to sprawling multi-campus deployments — and the strain that puts on the physical supply chain. Uptime Institute’s 2025 research backs this up, continuing to flag equipment lead times and workforce shortages as major operational risks industry-wide.
“Millions of connections must be installed and tested with virtually no tolerance for error, while designs, configurations and schedules keep evolving,” Verma said. “The challenge extends beyond access to technology to the ability to operationalise it safely, predictably and at industrial scale.”
What hyperscalers actually want now
Black Box currently works with four of the top six hyperscalers on gigawatt-scale programmes — a vantage point that gives Verma a clear read on how expectations are shifting.
“The biggest shift is from project delivery to business certainty,” he said. “Hyperscalers aren’t simply looking for the lowest-priced contractor; they want partners who can give them confidence that capacity will become operational on schedule, across several buildings and geographies.”
That confidence, he said, is earned through transparency and consistency rather than price. “Customers value engaging one partner who can support multiple infrastructure layers and coordinate delivery across locations, with the transparency that comes from standardised processes, clear governance and real-time visibility into performance.”
Flexibility has also become non-negotiable. “Technology and design requirements can evolve until very late in a programme, so partners must absorb change without losing control of cost, quality or timelines,” Verma noted. “That demands granular visibility, disciplined change management and tight integration between solutioning, commercial and delivery teams.”
But if there’s one word that captures what hyperscalers are really buying, it’s repeatability. “The first project may be won through capability and relationships; every subsequent one is earned through execution,” Verma said. “That repeatability — bringing the same standards and governance to different markets while adapting to local conditions — is what lets us operate as a strategic partner rather than a transactional provider.”
Building a “glocal” delivery machine
Executing that consistency across continents required Black Box to rethink its own operating model — anchored by two Centres of Excellence, one in Minnesota, one in Bengaluru.
“We’ve built an integrated delivery system rather than a collection of individual project capabilities,” Verma said. The Minnesota centre focuses on hyperscale-specific training, cable management, quality assurance, logistics and networking — “a practical environment to test methods, train technicians and standardise the quality expected on live programmes.” Bengaluru complements it with programme management, solution engineering and managed services, “reflecting India’s importance as a source of engineering and operational talent for global programmes.”
Underpinning both is a heavier investment in project controls. “We’ve significantly strengthened project controls, with standardised estimation, planning and delivery processes supported by industry-standard estimation and scheduling platforms and real-time dashboards that give customers detailed visibility across cost, schedule, resources and risk,” he said.
Equally critical, in Verma’s view, is workforce orchestration — blending direct employees, contract resources and certified local subcontractors “under consistent safety, quality and productivity standards, allowing us to scale without compromising accountability.” The result, he said, is a genuinely “glocal” model: “quality and transparency are the same everywhere but delivery still reflects local labour markets, regulations and supply chains.”
AI is rewriting the physical layer
If gigawatt-scale expansion is the headline, AI is the force multiplier changing what actually gets built inside those facilities.
“AI is turning the data centre into a much denser and more interconnected computing environment,” Verma said. “Traditional cloud architecture was designed mainly to move data between servers, storage and users; AI training requires thousands of GPUs to communicate continuously, at extremely high speed and very low latency.”
That shift cascades through nearly every design decision. “Rack power densities are rising sharply, liquid cooling is becoming essential, and fibre requirements are expanding significantly — power distribution, containment, cable pathways and network fabrics must now be designed as one integrated system,” he explained. Customers, he added, are no longer asking for conventional fit-outs — they want “AI-ready infrastructure”: high-density designs, scalable fibre architectures, deeper telemetry, and above all, adaptability. “They want architectures that can absorb changes in server platforms, network speeds and cooling methods rather than being fixed to today’s specification.”
Notably, Verma sees the retrofit opportunity as just as significant as new-build. “We’re also seeing growing demand to modernise existing data centres — retrofitting lower-density environments for higher power, cooling and connectivity without compromising resilience. In many cases, the ability to refresh existing capacity will be as important as building new.”
The next three to five years: synchronisation, not just scale
Looking ahead, Verma doesn’t believe the defining challenge of the next phase will be any single resource constraint. It will be the difficulty of aligning many of them at once.
“The biggest challenge will be synchronising capacity — bringing together power, land, cooling, connectivity, equipment, skilled people and regulatory approvals at precisely the right time,” he said. “Any one of these without the others doesn’t create operational capacity.”
Execution capability, he argued, will increasingly separate the companies that can participate in this growth phase from those that can’t. “The companies that succeed will be those with a repeatable delivery operating system: strong programme and cost controls, global supply-chain visibility, deep engineering and commissioning expertise, scalable workforce models and uncompromising safety and quality standards.”
He’s also clear that no single company — including Black Box — can do this alone. “No single organisation can independently deliver power, cooling, connectivity, compute and operations at this scale — so the winning model is collaborative but built on clear accountability.”
Verma closed with what might be the clearest articulation of how Black Box positions itself in this race: not as a builder chasing headline numbers, but as an operator of complexity at scale. “The market may measure growth in gigawatts, GPUs or square feet, but customers experience success as reliable capacity delivered on time and safely, with the confidence that it can be operated and scaled,” he said.
“Participation in the next phase won’t be determined by who can build one successful data centre — it will be determined by who can deliver multiple mission-critical environments, simultaneously and consistently, across geographies. That is the role Black Box is positioning itself to play.”