Baseload power with combined heat and power (CHP) applications does not have to be built specifically for artificial intelligence or data centres.
But I believe the power architecture should increasingly be designed with AI-era requirements in mind.
Why?
Because AI infrastructure is forcing us to reconsider what we expect from a modern power system: high availability, firm capacity, power quality, rapid response, efficient cooling and much better utilisation of energy that would otherwise be rejected as heat.
Those requirements are equally valuable for many industrial applications.
Power should no longer be designed in isolation
The conventional approach begins with electricity:
How many megawatts must the plant generate?
A more integrated approach asks:
What useful outputs can we obtain from the total energy entering the system?
Electricity is one output.
Steam is another.
Industrial heat is another.
Cooling can be another.
Water may become another.
And, increasingly, these outputs should be designed as one integrated energy architecture.
CRT provides an interesting example
In Carbon Recycling Technology (CRT), we capture CO₂ and react it with renewable hydrogen to regenerate methane:
CO₂ + 4H₂ → CH₄ + 2H₂O
This methanation reaction is strongly exothermic.
Therefore, when we establish the CRT mass balance and methanation stoichiometry, we simultaneously establish a substantial thermal-energy stream.
That heat is not simply an unfortunate loss that must later be removed.
It is part of the architecture.
The engineering question becomes:
Where should that thermal energy go?
Depending on the application, it could contribute to steam generation, process heating, absorption cooling, water treatment or other useful thermal duties.
This is why I increasingly view CRT not simply as a power-generation technology, but as an integrated power–heat–molecule system.
Then consider AI
A data centre may primarily require electricity, but electricity alone does not describe its energy system.
High-density computing creates a substantial cooling requirement. As rack densities increase, thermal management becomes increasingly important.
This creates an interesting opportunity.
Instead of designing:
Power plant → electricity
we can begin designing:
Primary energy → firm power → compute + useful heat integration + cooling + water management
The objective should be to maximise useful output from the entire energy system, not merely electrical efficiency at the generator terminals.
This principle extends far beyond AI
The same CRT plant architecture could support:
- industrial processing;
- green iron and steel;
- chemical manufacturing;
- food processing;
- district energy;
- desalination and water treatment; or
- data centres.
The customer does not have to be an AI company.
But designing the energy system with the demanding requirements of AI infrastructure in mind can produce a more flexible and resilient architecture for everyone.
Start with the complete energy balance
My preferred engineering sequence remains:
Mass balance → energy balance → heat integration → equipment efficiency → dynamic optimisation
Do not begin by deciding that heat is “waste heat.”
First calculate how much thermal energy the chemistry and equipment actually produce.
Then ask where that energy has the greatest value.
In CRT, particularly because of methanation, this could materially change the economics of CHP and trigeneration.
The AI era may therefore teach us something much broader than how to power data centres.
It may teach us how to design integrated energy systems in which every megawatt — electrical or thermal — has a useful destination.
