- How are we going to power AI?
- AI’s poor sense of timing is a historic opportunity
- The shovels to AI’s golden nuggets
It’s often said that selling shovels was the best way to profit from a gold rush. If so, what is the equivalent investment play on artificial intelligence (AI) today?
My old friend Sam Volkering, who first told me about bitcoin around 2012, reckons he has found the answer. You can sign up to his investment briefing on that topic here.
From what I’ve seen, it’s to do with using AI instead of investing in it. That’s a classic variation on the “shovels instead of nuggets” argument.
But I have another one…
You see, AI is frighteningly energy intensive. The sort of computing power we are going to need to make AI accessible in our everyday lives is truly enormous.
We’re talking about adding a mid-sized country to the global electricity grid. And that’s under a conservative estimate. It all depends on how much we actually use AI.
The world’s data centres already use more than a percent of the global energy use. If Google ran its search engine using AI tech, it’d need to invest $100 billion to build an energy grid with the capacity of Ireland’s…
Strangely enough, a lot of the energy demand from AI is up-front, before you can even use it. AI programs need to go through what’s known as a “training phase” before they begin being useful. And that training needs a lot of energy in and of itself. Not to mention actually using the program thereafter.
Then there’s cooling the data centres running AI, which is also notoriously energy intensive in most places.
I’m sure you know where I’m going with all of this…
AI’s poor sense of timing
The problem with our AI future is that we are currently busy trying to cut energy consumption because of carbon emissions. Quite frankly, I don’t think we can have both AI and anything resembling net zero.
Some countries have managed to reduce their energy consumption substantially already, depending on how you measure things. The average German is using less energy than they did in 1970!
Now, a great deal of that drop is due to the deindustrialisation of energy intensive industries. Instead of producing energy intensive goods, the Germans import them from developing countries. And tell themselves this has lowered emissions…
While this may also be a plausible strategy for AI – just pollute and generate electricity elsewhere – it’s probably a bit too cynical for even today’s politicians and tech moguls.
While AI is taking off in a limited number of contexts, like generating text, audio and visual content, we’d need a vast rollout of AI for it to begin impacting other parts of our lives. And, unlike some, I suspect that impact is more likely to be good than bad.
But the world is not exactly matching its coming AI boom with the corresponding energy boom that’s needed to make AI a reality. There aren’t enough shovels to make the gold rush happen in the first place.
Not yet…
If we begin to accept that technological revolutions which can improve our standard of living dramatically tend to need a lot more energy, then the question becomes where that energy will be coming from. I don’t think people will like AI if it makes their lights go out…
For once in the energy sector, the free market isn’t just sitting around waiting for the government to provide sufficient subsidies. The sorts of subsidies needed to help companies overcome the impossible red tape governments also create for energy producers.
Instead, the private sector seems to have found an intriguing work-around…
Descalable solutions
The most important feature of technology today is its scalability. If you wanted to hear music 300 years ago, you need a musician in person. And that musician could only be heard as far as they could be heard.
Today, anyone can come up with a song, upload it online and get played for pretty much every human being’s ear within 24 hours. That’s scalability – the ability to multiply the amount of your product rapidly. It’s largely a digital phenomenon.
The problem with scalability is that, when it goes awry, it does so on a large scale. Everyone gets a virus instead of the song they wanted.
Our energy grid today is an enormous, interconnected behemoth that can move power around the country or even internationally. But it’s not exactly doing a good job of it. We keep experiencing issues.
Today, I want to show you what the reverse can look like: what if our energy grid became a series of smaller, less interconnected and more independent blocks?
People have questioned nuclear power from small modular reactors (SMRs) for quite some time. They are considered inefficient and too small compared to their (properly built) larger cousins. SMRs just don’t make sense when compared with the huge nuclear power plants we’re used to.
But what if SMRs solve an entirely different problem to supplying a vast energy grid to power the country? What if the SMR revolution is all about making electricity a private and free market good that allows people to provide the power they need independently from governments, large corporations, grid interconnections and so much more that’s going wrong today?
What if SMRs are cheap and versatile enough to make it possible for non-energy focused industries and firms to power their own operations in a reliable way even while the country’s grid is falling apart?
Right now, the diversified and national energy grid is failing industries and households in many countries. That’s why industry is leaving places like Germany and why people are leaving places like California.
What if companies and industries could solve this problem, not at the ballot box, but by building their own solutions?
In Surry, Virginia, which is not to be confused with Surrey, UK, plans are underway to try this. They want to solve the problem of unreliable power from the network, and vast power demand from the sorts of data centres that can drive AI, by becoming self-reliant on power.
I suspect this could become a model for the future which circumvents all the problems we are seeing on the power grid today. We’re talking about an Uber and Airbnb type of solution solving the ridiculous problems of the taxi and hotel industries, but for electricity. The secret is to circumvent what’s failing rather than solve it.
The idea for Surry is to build SMRs that provide enough nuclear power to supply 30 data centres and a hydrogen production facility.
What makes this project so interesting is the decision to provide its own power for itself – that’s what SMRs seem to be revolutionising. They might be more expensive, less efficient, etc, than their larger counterparts, but at least they circumvent the problems created by being reliant on the grid.
What SMRs promise is the ability for companies like German chemicals giant BASF to provide their own power sources for their own operations on their own terms, without government involvement and without the capacity of the government to meddle in the future.
The same goes for mining companies, which have been trying to figure out how they’ll make their incredibly isolated operations go green.
In fact, mines are probably the best example of what I’m trying to get at. These remote locations need stable, 24/7, reliable, clean and safe power on a moderate scale. That rules out pretty much every possible energy tech we have today, despite miners’ best efforts to try them. But SMRs are the perfect solution, presuming you can get electrification to work.
Independently powered AI
AI could of course function on the back of the national grid. But I rather suspect that the companies driving AI today are not going to want to rely on politicians getting their energy policy right in order for the AI programs to function. Not after the last few years…
I suspect that the tech moguls of the world who are exploring AI will want their own reliable, independent, clean and cheap energy.
So, if you’re as bullish on AI as Sam Volkering and I, you need to have some understanding of what that means for energy markets.
My prediction is this: unless you have an in-depth understanding of AI’s specific opportunities, the best way to profit from the AI boom in coming decades will be found in SMR technology. SMRs are the shovels of the coming AI gold rush.
Alternatively, you could gain access to some of the brightest minds out there on the topic of AI itself, and investing in AI. They can give you the advantage you need to invest in the AI boom directly. Find out what they’re planning, here.
Until next time,
Nick Hubble
Editor, Fortune & Freedom
