Artificial Intelligence poses an unprecedented threat to energy security in the Global North which could seriously derail decarbonization goals, place immense strain onto power grids, and cause volatility in energy markets, which will inevitably extend to the economy writ large. After decades of plateaued energy demand, the energy required by data centers is now surging, and is expected to continue to rise at a breakneck pace.
By 2030, AI is expected to represent 3.5 percent of the global electricity consumption, and 9 percent of electricity generation in the United States (a sharp increase from the country’s current rate of around 3.5 percent – already a hefty amount). Put together, electric vehicles and AI are on track to add 290 terawatt hours of electricity demand to the United States energy grid by the end of the decade according to projections by Rystad Energy. This will put their collective electricity consumption at about the same level as the entire country of Turkey, the world’s 18th largest economy.
“When you look at the numbers, it is staggering,” Jason Shaw, chairman of the Georgia Public Service Commission, a U.S. electricity regulator, told the Washington Post earlier this year. “It makes you scratch your head and wonder how we ended up in this situation. How were the projections that far off? This has created a challenge like we have never seen before.”
In the face of this rapidly growing problem, public and private sector leaders are scrambling to come up with new ways to feed the tech sector’s newly insatiable energy demand without seriously compromising energy security or climate outcomes. “This growth is a race against time to expand power generation without overwhelming electricity systems to the point of stress,” said Rystad analyst Surya Hendry.
Slowing the growth of AI, perhaps the most logical solution to the puzzle, seems to be completely out of the question. In the United States, the technology has rare and strong bipartisan support, as maintaining a leadership position in the emerging sector is viewed as a critical strategy for national security, the economy, cybersecurity, and governance of the tech sector. There is no putting the genie back in the bottle.
With unquestionably massive energy growth just around the bend, the scale of the problem of powering AI in the near future is so immense that solutions are relying more on futuristic tech approaches than existing technologies. Big Tech bigwigs like Bill Gates and Sam Altman are calling for increased investing for nuclear fusion research as a potential way to unlock massive amounts of clean energy. Others are investigating not only how to produce more clean energy efficiently, but how to make AI consume less.
One potential solution for the latter approach may be found through quantum computing. While normal computers run on binary, with 1s and 0s serving as on- and off-switches, quantum computing runs on qubits, which can be both on and off simultaneously, like a heads-or-tails coin flip before it lands. This state of being both on and off at the same time is called superposition, and it could completely revolutionize computing as we know it.
In certain cases, quantum computers could be up to 100 times more energy efficient than a standard supercomputer. This could have enormous implications for AI, for which quantum computing could be especially well-suited.
“For the things that quantum computing is good at — such as AI processing — there's no way for any GPU to compete against us. Those workloads are ultimately going to go to quantum and current technology just literally can't compete,” Peter Chapman, President and CEO of quantum computing company IonQ, was recently quoted by Forbes. “Quantum computing — our next generation chip — to simulate what it's doing, you would need something like two and a half billion GPUs and it runs off a two standard wall sockets,” he added. Chapman says his company will likely have prototypes of such a chip ready in just six to nine months.
While the scalable use of quantum computing would be an enormous step in the right direction for the tech sector, the country, and the world, it should not be viewed as a silver bullet. According to Barclays’ Will Thompson, who recently co-authored research on AI power consumption, solving the AI energy puzzle “will require an above-all-approach of expanding and modernizing electric grid infrastructure, integrating renewables with utility-scale storage, utilizing our existing nuclear capacity, and scaling new forms of carbon-free energy. That will include geothermal, advances nuclear small modular reactors (SMRs) and fusion technology.”
Furthermore, as quantum computing still has quite a ways to go before it becomes a commercial reality, a broad approach to boosting clean energy as well as energy efficiency is the order of the day.
By Haley Zaremba for Oilprice.com