Key Takeaways
- AI will need massive amounts of power to sustain itself in the next few years.
- Big AI companies are turning to nuclear energy as a cleaner and more reliable energy source.
- Clean energy can accelerate AI development which, in turn, will advance research into making sustainable energy more efficient.
AI this, AI that — it seems AI chatbots, AI PCs, AI generators, and AI large language models (LLMs) have hijacked the entire tech conversation. Not a day goes by without coming across news of a new ChatGPT LLM or a new ChatGPT competitor. Everyone wants a piece of the fresh-out-of-the-oven AI pie, and tech companies are scrambling to be the first ones to get the next big thing out the door.
But, what you might not have noticed as much is the brewing conversation around AI’s energy demands. As companies break new ground in AI models and applications, the world’s existing energy infrastructure will soon prove woefully inadequate to satiate the associated rising power requirements. However, tech giants might be one step ahead of all of us. They are already hoping that nuclear energy will be the magic pill that saves the day.
Related
Forget about software, because the race to AI dominance is really about chipmaking
All the biggest names in AI and technology are fighting behind the scenes to make the best AI chips.
Why does AI need so much energy?
Gotta power those sweet GPUs
Computers need power. The one you’re using right now might not need heaps of it, but the servers that power the apps and websites we use daily run on serious amounts of electricity. The same is true with the AI data centers running the core hardware (GPUs and NPUs) behind the scenes. Data center solutions like the Nvidia DGX are composed of multiple Nvidia GPUs and x86 CPUs. AI companies often invest in multi-million-dollar DGX SuperPODs, each of which has up to 72 of Nvidia’s latest GPUs. Such expansive setups naturally require loads of power to sustain them 24/7.
The computational power needed to support the AI surge is currently doubling approximately every 100 days.
Plus, AI searches online can require almost 10 times as much energy than that needed for a standard search. But what’s really troubling is that the computational power needed to support the AI surge is currently doubling approximately every 100 days. In fact, according to a New Yorker report, ChatGPT alone consumes over 500,000 kWh every single day — this is roughly 17,000 times more than an average US household consumes in a day.
And unlike the ones keeping the internet alive, the data centers which are at the heart of new-age AI applications are proliferating at a much faster rate. According to the World Economic Forum, by 2028, we could be looking at AI gobbling up the equivalent of Iceland’s entire power consumption in 2021. What will this mean for the power needs of the general public? For instance, the US alone is expected to experience a rapid rise in energy requirements, after more than a decade of nearly flat demand.
Related
OpenAI wants to make its own AI hardware
The company recently struck a deal to get chips made.
Big AI is betting big on nuclear energy
Clean, mean energy of the future
The concerns about the burgeoning energy demands of AI haven’t sprung up out of the blue. The cryptocurrency wave already showed us how organizations as well as consumers chasing a trend contributed to a chip shortage and rising electricity demand. It seems the biggest players on the board have anticipated the impending energy shortage due to their AI ambitions, as we’ve seen multiple big-ticket nuclear deals in the recent past as proof of this.
Microsoft, the biggest investor in OpenAI, signed a deal with nuclear fusion company, Helion Energy, to source fusion-powered electricity in about five years. While practical nuclear fusion is still elusive today, we’ve also seen landmark nuclear fission deals go down, clearly signaling toward the future-looking outlook of the biggest tech companies.
Nuclear fusion offers a utopian energy source for the future — carbon-free, waste-free, and independent of environmental factors.
For instance, Sam Altman of OpenAI is planning to raise hundreds of millions for Oklo, a company working on nuclear micro-reactors. Plus, Amazon Web Services (AWS) recently acquired a 960MW data center campus in Pennsylvania from Talen Energy. Google also announced a joint collaboration with Microsoft and steelmaker Nucor to purchase clean energy from geothermal, clean hydrogen, battery storage, and nuclear technologies.
Nuclear fusion offers a utopian energy source for the future — carbon-free, waste-free, and independent of environmental factors like wind and solar energy. Unlike nuclear fission, which is the current viable source of nuclear energy and accounts for 18% of the US electricity capacity, nuclear-fusion-powered energy might still be decades away from being realized. Still, it seems promising that there is growing interest from big tech to get us there eventually.
Related
On-device AI processing is the breakthrough we’re still waiting for
Artificial intelligence is everywhere, but most of it happens in the cloud. The next big shift is bringing all that processing on-device.
AI and sustainable energy will be a symbiotic relationship
AI will help renewable energy help itself
As big tech pours millions into nuclear energy and other forms of renewable energy, the dividends could benefit them in more ways than one. Providing clean energy for AI data centers will mitigate pollution concerns over AI advancement, paving the way for rapid development. This will inevitably advance renewable energy, as AI research makes it more efficient and reliable. And this, in turn, will enable more and more clean energy for AI data centers, completing the loop.
As it stands today, AI is already doing that by predicting the weather, monitoring melting ice, tracking pollution, and mapping deforestation. A recent BCG report claimed that AI can reduce the amount of planet-heating pollution by 10%. It could also predict instabilities in nuclear fusion reactions, helping accelerate commercialization of the technology.
Perhaps the biggest upside of big tech’s nuclear ambitions will be the renewed interest in nuclear energy, allowing heightened scientific and public investment into the technology. The way things look with respect to climate change and the growing geo-political instabilities, seems to indicate that our reliance on conventional energy sources needs to reduce as soon as possible.
Related
3 ways AI will change gaming forever
AI is doing things humans cannot, and it’s only getting better.
Nuclear energy might just make the AI wave worth it
The unprecedented rate at which AI is advancing today raises genuine concerns over its environmental impact. We can now directly link its growing carbon footprint to its rising adoption, not in terms of empty arguments but with hard numbers. So, if nuclear fission and later, nuclear fusion can help alleviate some of the AI burden from fossil fuels, that’s something to cheer on.
Of course, things could take a turn for the worse where we might only witness more energy consumption as a result of even more AI adoption. Let’s tackle one problem at a time though, and hope that nuclear energy does come through for us.