Op-ed by Francesco Pinci, Boss Energy.
The AI race has largely been framed around speed, investment, and computational advantage. But as demand for energy-intensive data centres rises, the limits of the physical system supporting that growth are becoming harder to ignore. Grid access, power availability, and energy pricing are now moving closer to the centre of the discussion.
Many technology companies have tried to respond by committing to cover the incremental infrastructure costs associated with their growth. These pledges are intended to reassure regulators and the public that AI expansion will not come at the expense of ordinary ratepayers. Yet in practice, electricity systems are interconnected, and it is not always possible to isolate the effects of a major new load so neatly.
When new data centres require transmission upgrades, additional generation, or tighter procurement conditions, the resulting pressure can move through the wider system in indirect ways. Equipment costs can rise, local capacity constraints can intensify, and wholesale dynamics can shift, even when a specific operator agrees to fund part of the build-out directly. That is why the question of who ultimately pays remains politically sensitive.
At the same time, the industry cannot ignore the commercial reality. If access to power becomes slower, more uncertain, or more expensive, it affects the economics of deployment and the pace at which AI capacity can be brought online. In that sense, electricity is no longer just an operational cost. It is becoming a strategic condition of growth.
Building resilience for the AI energy challenge
Ana Jorge Sobrido’s perspective reinforces that argument from a different but complementary angle. As Professor of Sustainable Energy Materials at Queen Mary University of London and Director of the Centre for Sustainable Engineering, her work reflects a broader understanding of sustainability as something that depends on resilience, integration, and long-term system performance rather than isolated technological optimism
One of the strongest ideas associated with her perspective is that next-generation materials and technologies matter not only because they improve performance, but because they can become enablers of more reliable and decentralised energy systems. In the context of AI infrastructure, that matters a great deal. The challenge is not simply to add more electricity, but to do so in ways that strengthen the system rather than exposing new weaknesses.
Sobrido’s thinking also points toward the importance of practical deployment. Innovation alone is not enough if it cannot be translated into solutions that work at scale, under real-world conditions, and across increasingly complex industrial environments. That message fits the AI-energy debate closely. The most compelling long-term technologies may still be years away from broad deployment, while the demand pressures created by AI are already here.
AI is not simply facing an energy shortage; it is facing a systems challenge. The next stage of growth depends on whether infrastructure, policy, technology, and public acceptance can move in step rather than in conflict.
Why companies are exploring new power models
For most data centre operators, the grid remains the preferred source of electricity because it is generally more reliable and more operationally efficient than self-generation. However, where grid connection timelines are too slow, companies are increasingly considering collocated power, private generation, and hybrid supply models to reduce delays and gain more control over project delivery.
This trend is not only technical; it is strategic. Companies that can secure reliable electricity independently, or at least more flexibly, may be able to scale their AI infrastructure faster than those waiting on standard connection processes. In an increasingly competitive market, access to power is beginning to look like a source of competitive advantage in its own right.
Yet these alternatives are not without trade-offs. Private generation can introduce new complexity, higher capital requirements, and environmental concerns, particularly if it relies on more carbon-intensive forms of supply. That is why the long-term solution still points back to the same principle: growth must be aligned with sustainable, resilient system design rather than short-term power acquisition alone.
The bigger test for AI
The AI industry is now reaching a point where its future will depend not just on technical capability, but on the credibility of the infrastructure model behind it.
If companies are seen as expanding in ways that contribute to higher electricity bills, local grid stress, or uneven public costs, resistance will intensify. If, on the other hand, they can demonstrate a more responsible alignment between digital growth and energy planning, the political environment becomes more manageable.
This is why the debate around sustainable scaling is becoming so important. The issue is no longer simply whether the technology works, but whether the wider system can absorb it in a way that is fair, resilient, and economically defensible. That is a much more demanding test than product performance alone.
The future of AI infrastructure will not be driven by a single factor. It will depend on the ability to bridge innovation with real-world deployment, align ambition with resilience, and scale within an energy system capable of sustaining it.
Learn more at www.bossprofessional.com.
