Last week, Microsoft signed up as a partner of the World Nuclear Association in what both sides – in this rare case with some justification – trumpeted as a game-changer because of the amount of money it could put behind rolling out nuclear energy.
Just about at the same time, we learnt that Rolls-Royce is thinking about a possible IPO for its small nuclear business “amid growing investor excitement about the nascent technology”, as the Financial Times put it.
What everyone is actually excited about is, of course, AI.
Amazon and Google are trying to buy small modular reactors (SMRs) that they can plug straight into the many new data centres they need to feed AI demand. That way, they don't have to worry about extending the electricity grid, or putting up huge storage to deal with variable renewables output.
Sam Altman chaired Oklo, a fast-neutron start-up, until he had to step aside because it is a bit dodgy to be creating both the disease and the cure, as it were, in the electricity market.
Models are getting more efficient fast, but demand is rising much faster, so every projection for future electricity demand is basically an arrow pointing straight up. Meanwhile, nobody wants new pylons running through their patch of countryside, and nobody wants to wait 10 years for new power either.
So, everyone wants to buy SMRs. What they won't be buying is South African PBMRs.
To be fair, pebble bed modular reactors aren't hot right now; there's one plant in commercial operation in Shidaowan, China, and that's about it. Integral pressurised water reactors are the flavour of the month, with the high-temperature gas-cooled PBMR approach still the Betamax equivalent, just without even the niche success that Betamax found.
But just listen to the language the hot nuclear companies such as Oklo are using: fast, small, inherently safe, stackable, versatile, efficient. There was a time South Africa was all but alone in that kind of strategic approach to nuclear. Unfortunately, that time was about a quarter of a century too early.
SA remains a strange and wonderful place that must confront problems nobody else is really interested in right now.
Eskom – then a visionary company with the appetite and resources for serious R&D – started buying in PBMR tech in 1999, just as everyone else was losing steam on that technology, and a bit on nuclear in general. The Germans offloaded their assets to the British, which had already taken over the Americans, and it all ended up in Japan not long after, with the French also in there.
It's not that there was no value in IP that had been built up over a long time, it was just that there wasn't a clear market, or even use-case.
South Africa was different, and nearly unique in its confluence of circumstances. It wanted to leapfrog the need to build big grids; it wasn't afraid of nuclear; it had some money to work with; it had a stable, long-term political vision; and it wanted to be a trail-blazer, not a mere consumer of foreign technology.
And so, by 2005, the South African government was very nearly alone in funding something very close to all the serious PBMR work being done on the planet.
By 2010, of course, it was all over. The money tap was closed and the mothballs came out.
Could South Africa have held on, financially and politically, for another 15 years? Would it have done so, had the Zuma administration not set its heart on hugely expensive Russian reactors for reasons never adequately explained? Is there a parallel dimension where SA is powering the AI revolution with a similarly revolutionary approach to making electricity?
Maybe, maybe not. We certainly can't fault SA for failing to foresee the AI boom back in 2010.
We can – and very much should – think long and hard about what the whole episode tells us about sustainable funding of blue skies research, about shared public-private risk, and about pre-incubation resources in South Africa.
It is probably too late to de-mothball the PBMR now, but SA remains a strange and wonderful place that must confront problems nobody else is really interested in right now, but which could have global application down the line.
Notably, in AI.
Rolls-Royce and Microsoft will win this one, but that does not have to be the case next time.
Share