FROM THE INDUSTRY
tangible assets gain pricing power. Chips, equipment, power systems and networks cannot be scaled instantly and that scarcity is value.
In this model, hyperscalers depend on communication service providers as infrastructure counterparts, which is not an immediately obvious outcome. OEMs and Chipmakers Signal a Hardware-Led Cycle The supplier ecosystem reinforces this shift. AI infrastructure is pulling forward the entire hardware stack. Chipmakers are expanding accelerator, CPU, and switching roadmaps around density, power efficiency and interconnect. Networking vendors are prioritising AI- optimised fabrics, high-speed Ethernet and optical integration. Optical suppliers are scaling fibre, photonics and transceiver production to meet data centre and transport demand. Systems vendors are redesigning platforms around thermal constraints and fabric integration; even for non-AI workloads. Across the board, roadmaps now assume that hardware limitations shape software outcomes, not the other way around. This is not a short-term spike; what we are witnessing is a multi-year infrastructure cycle anchored in physical deployment.
This is what leverage looks like.
Sustainability in the AI Era: Better Infrastructure, Longer Lifecycles AI’s resource demands are real. Power, water, materials, and cooling capacity are under pressure, and the scale of AI deployment makes those constraints impossible to ignore. Sustainability in 2026, as we embrace AI makes us confront a new truth: less infrastructure is a tempting solution, but shouldn’t we just be deploying what we have more effectively? As AI accelerates hardware turnover, the full lifecycle of infrastructure is becoming a strategic consideration. Hardware sustainability is no longer limited to energy efficiency during operation; it now includes embodied carbon, supply-chain impact, reuse and redeployment.
This is where the circular economy moves from compliance to advantage.
Markets Are Already Reflecting the Shift
Not every AI workload requires the newest accelerator or the densest rack. Inference, edge compute, networking, storage, control-plane systems and many distributed workloads can be supported by existing platforms; if those platforms are properly refurbished, requalified and redeployed. In a market defined by long lead times and constrained supply, refurbished and redeployed hardware can: n Accelerate time-to-capacity when new equipment is delayed n Reduce embodied carbon compared to greenfield builds n Unlock stranded capital from decommissioned data centres, central offices and network sites n Enable localised and sovereign deployments without hyperscale economics Decommissioned now means available capacity, as opposed to obsolete kit we have no use for.
Financial markets are quietly confirming what infrastructure teams see on the ground.
Over the past six months:
Hardware-led segments: n Semiconductor Equipment +60% n Computer Hardware +43% n Electronic Components +41% n Semiconductors +25%
Software segments:
n Software (Infrastructure) –6%
n Software (Applications) –15%
Same broad “technology” category. Very different economics.
Investors are rediscovering a basic truth: when demand collides with physical limits,
68
MARCH 2026 Volume 48 No.1
Made with FlippingBook - Online magazine maker