By Atul Roy
Recent market turmoil in utility stocks, situated near major data centre hubs has sparked debate about the future of AI-supporting infrastructure. While this market reaction initially suggested scepticism about AI's power demand, a deeper analysis reveals a more complex transformation — not a reduction in power demand, but a fundamental shift in how and where that power will be consumed and, consequently, where it is generated.
The DeepSeek Catalyst
DeepSeek's breakthrough in training a competitive AI model for just $5.6 million — a fraction of established players’ typical spend — marks a pivotal moment in AI development. This comes as the industry grapples with mounting concerns over AI training costs and concentration risk in specialised data centres. The timing is particularly significant, as growing geopolitical tensions underscore the strategic importance of geographically diversified infrastructure.
The implications of DeepSeek's innovation extend beyond cost savings: it challenges the conventional wisdom that AI computation must be concentrated in massive, proprietary facilities operated by tech giants.
Instead, we're witnessing the emergence of a more distributed model that could fundamentally reshape the data centre landscape.
This shift manifests itself in three ways. Firstly, the increasing accessibility of AI development is accelerating adoption across industries, driving demand for distributed computing resources that support practical applications rather than training.
Secondly, power demand isn't diminishing — it's dispersing across a broader network of facilities that support diverse, production-level AI applications.
Finally, colocation and interconnection facilities serving mature use cases are uniquely positioned to thrive in this evolving landscape, offering stable returns and reduced exposure to AI training cost fluctuations.
Market Evolution and Strategic Considerations
The post-DeepSeek era points toward a future dominated by smaller, geographically distributed data centres rather than concentrated AI training facilities. This transformation particularly benefits edge computing and regional data centres that support mature AI applications, as the market emphasis shifts from training to practical deployment.
This evolution in infrastructure requirements creates compelling opportunities for facilities that serve diverse applications and established technologies. As AI development costs decrease and adoption accelerates, data centres positioned at the application end of the AI value chain stand to benefit from increased demand for distributed computing resources.
DeepSeek's emergence — with its selective open-sourcing and timing coinciding with significant political events — raises important considerations about the future of global AI infrastructure. The potential for regional blocks, regulatory constraints, and increased competition in AI platform development could further accelerate the trend toward distributed infrastructure.
More importantly, the reduced cost of AI training could catalyse the rapid development of new AI applications. This proliferation will likely drive demand for data centre infrastructure that prioritises deployment over training — a positive development for facilities focused on production-level services and mature use cases.
Conclusion
DeepSeek’s innovation has changed the AI landscape and heralds a significant transition in computing infrastructure, favouring distributed, diverse facilities over concentrated AI training centres. This evolution particularly benefits data centres positioned to support mature AI applications and diverse use cases.
For infrastructure investors, the opportunity lies in facilities that enable the practical implementation of AI technologies rather than their development. As AI training costs decrease and adoption increases, data centres serving these use cases across diverse industries are poised to capitalise on the value from the expanding ecosystem of AI applications.
The future may not be fewer data centres, but rather more distributed, diverse, and democratised ones — a shift that creates compelling opportunities for well-positioned investors who understand these evolving dynamics.
Atul Roy is the Managing Director of Cordiant Digital.