By John Q. Hosedrinker 
OpenAI, faced with the immense energy demands of its AI models like Sora 2 and ChatGPT, has secured another major power deal. This agreement, totaling 10 gigawatts, reflects the significant energy consumption required to train and run large language models and video generators. The deal underscores the rapid growth of AI and the substantial infrastructure needed to support it. This ongoing expansion highlights the increasing pressure on energy resources as the field of artificial intelligence continues to advance.
Read the original article here
Sora 2 and ChatGPT are consuming so much power that OpenAI just did another 10 gigawatt deal, and it’s a bit like watching a runaway train. These AI models, along with their brethren, are demanding an absolutely staggering amount of energy. We’re talking about a deal that could potentially make a significant dent in the energy grid, and the implications are, frankly, a bit frightening. The scale is almost incomprehensible; it’s like a tech-fueled hunger that demands more and more power, with little regard for the consequences.
For some perspective, imagine the potential impact. If AI’s energy demands continue on this trajectory, we could see a scenario where a substantial portion of the nation’s total energy consumption is dedicated to running these models. The thought of dedicating a substantial percentage of the nation’s energy use to creating… well, a whole lot of internet slop, is a rather sobering thought.
There’s a curious shift happening in how we measure this power consumption. While the computing power of these AI models could be expressed in more traditional terms, like PFLOPs, the focus seems to be shifting to power consumption as the primary metric. This might reflect the practical realities of the situation. It’s not just about the abstract computing ability; it’s about the concrete need for more and more electricity.
A potential, if somewhat ironic, outcome is the accelerated adoption of nuclear power. Historically, corporations have often lagged on green energy initiatives. But now, facing the insatiable energy needs of AI, these massive companies might become the driving force behind infrastructure upgrades. This shift could change the energy landscape, but the irony is palpable – these companies are the very ones who ignored the cries for greener sources, but the politicians are less likely to ignore their massive needs.
This feels like a race with potentially devastating consequences. AI, despite its shortcomings, seems to be employed in ways that devalue labor, and it’s starting to look like a propaganda tool. This isn’t about innovation that benefits everyone; it feels more like a situation where a few benefit, and the many suffer. The environmental impact of all this is a growing concern, and there’s a legitimate question about whether corporate energy needs should outweigh the needs of ordinary citizens.
What happens when the average person’s access to power is restricted because of these demands? There’s a growing sense of urgency and unease, especially as we consider the massive resource consumption of the AI industry. We should start to think about the broader picture; the vast energy expenditure needed to run these systems, especially when considering the vast quantities of resources required to keep them running. This includes water consumption, which will likely become another crisis, as the industry expands.
This current trajectory is not just about energy. The amount of resources this industry demands has reached a level that could be considered unsustainable. The wealth generated and held is more easily measured in energy, according to one theory. This applies to everything from how much energy is required to heat a loaf of bread in Rome, to how much energy is used by today’s technology. Private planes and jet fuel, for instance, are ridiculously energy-intensive.
The fact that such vast resources are being consumed for something that arguably doesn’t provide equivalent value feels like a failure of global capitalism. The “look at my BIG THING getting BIGGER” approach isn’t sustainable, and it’s hard to ignore the fact that OpenAI, despite generating revenue, seems to lack a clear path to profitability. It’s a model that can’t last forever.
The charging of electric vehicles raised concerns about the existing power grid’s capacity, and the same can be said for the AI data centers. Given that data centers already use a substantial percentage of all electricity produced, the added burden from AI models is concerning. It does not paint a positive picture for our future, with the need to meet all of these demands.
One gigawatt is a genuinely astonishing amount of energy. It is increasingly apparent that electricity bills have skyrocketed due to the data centers that are the lifeblood of AI. This represents a significant shift in the energy economy, and its impact is already being felt by consumers. Moreover, the sheer waste of resources, like drinking water, for AI is alarming.
The fact is, these deals are not real. There is no concrete plan for all of this, and the timeline for developing power infrastructure is extremely lengthy. Gas turbines have a multi-year backlog, and nuclear plants take years to build. This all feels like hype, and a move to artificially inflate the company’s public reputation in order to sell it off to the public. All of this is an indication of how much demand there is for AI, and how much consumers will have to pay for it.
The potential impacts extend beyond electricity bills and environmental concerns. AI is already being used to generate vast amounts of low-quality content, creating an online landscape polluted with junk and adding very little value to society. Even worse, the energy is being used for brute-force model training, with no real innovation beyond “run the same training, but 10x more”.
There’s a sense of déjà vu, as this feels like the dot-com bubble all over again. It’s time for a serious examination of the situation, and how it’s affecting everything. There is a huge amount of energy being allocated to AI, and the only way for it to work is to be eco-friendly or to stop the machine. There are those who do not even care about the human and environment costs. And, of course, there is the risk of AI being used for very malicious purposes, like misinforming and deceiving people.