From cloud to compute: Why AI is rewriting the rules of digital infrastructure assets
This article was originally published in The Alternative Investor (November 2025)
Christopher Yoshida, Executive Chairman and Co-Founder, Nexus Core Systems
Everyone talks about AI as the rise of a revolutionary business tool. It's more than that. In truth, it's an industrial revolution, one built on compute.
Every technology cycle creates its own kind of infrastructure. The cloud era gave us the warehouses of the internet - built to store, move and monetise data. The AI era is different. It's about transforming that data into intelligence, decisions and action.
And that shift changes everything. These workloads are hungry, pulling power at levels the digital infrastructure market has never seen. Cooling, density, grid access – all of it must be re-engineered.
This is no longer about capacity. It's about performance: how efficiently a facility can turn a watt of power into intelligence. That's the new equation.
When technology becomes the yield curve
Digital infrastructure lives and dies by how well it keeps pace with the technology it enables. In AI, each new generation of computing hardware – most visibly led by NVIDIA's accelerator chips – resets the benchmark for performance and efficiency.
Think of them as AI factories, not data centres.
In this world, infrastructure behaves less like real estate and more like an industrial process. Output isn't measured in rent or racks, it's measured in throughput – how much intelligence you can produce per watt, per dollar, per second.
From dollars to tokens
Listen to NVIDIA's Jensen Huang and you'll notice he rarely talks about dollars. He talks about tokens – the real unit of productivity in AI.
Every model trained, every query answered, every decision generated consumes tokens. The velocity of those tokens – how fast energy becomes intelligence – now defines competitiveness. For investors, the focus shifts from occupancy and yield to time-to-token.
It's also a sustainability story. The more efficiently tokens are generated, the more value – and less carbon – per unit of energy consumed.
AI is redrawing the map
Of roughly 5,500 US data centres, only a fraction can handle AI-scale workloads. The rest of the world is even further behind and massively under-represented, with limited access to the compute required to train and deploy advanced models.
Power is now the defining factor in digital geography, and it's creating one of the biggest rebalancing opportunities in modern infrastructure.
New corridors of compute will emerge where energy, stability and policy align, unlocking growth far beyond the traditional hyperscale hubs. That's why new entrants, including Nexus Core Systems, are looking beyond traditional zones to combine long-term power security, renewable supply and global connectivity.
Why it matters
This isn't another speculative cycle. It's a durable, income producing asset class rooted in physical capability – power, hardware and performance.
Yes, the diligence is tougher and the capital requirements higher, but the fundamentals are stronger. Those who secure efficient power, build modularly and stay aligned with the technology curve will endure through market cycles.
In the end, value will be defined by how efficiently power, technology and capital combine to deliver lasting compute capacity for intelligence.
