Azeem Azhar

The reason is that a gigawatt data center is probably worth tens of billions of dollars or $10 billion of revenue, reckon semi-analysis, over its lifetime.

So you don't really want to sit around waiting for a grid connection.

And you're happy to pay for natural gas, which is more expensive now than solar in most parts of the world.

But it's a blip against the overall paradigm, which is that

Electricity is turning into a manufactured technology because of the learning curve that attaches to solar and why we are seeing this dramatic growth in solar deployment, not just in the rich parts of the world, but also in places like India.

India's coal consumption in 2025 was lower than in 2024, in large part because of the growth of solar.

And we've obviously know the story in Pakistan and we see what's happening in sub-Saharan Africa.

What will continue to happen year in, year out will be the doubling of cumulative production will continue to apply that

learning rate on solar.

And production is growing rapidly.

Pakistan imported 17 gigawatts of solar modules in 2024.

That was more than all but four countries.

Africa imported, the sub-Saharan Africa, two gigawatts of panels in one month last year.

last year.

This is people-led, market-led, and it's policy-led, but all of those panels increase production volumes and the learning rate applies.

Historical learning rates have been 20% per doubling of cumulative production, which means that the panels get 20% cheaper.

New analysis suggests we might be in a regime change and then that learning rate might be as high as 40%, maybe even higher.

I'm a bit sceptical, frankly, about it being 40%, but what it means is that the

reduction in cost will get even faster.

Now, of course, there are other questions like balance of systems costs and how do you build out a utility scale solar cheaply?