There have been quite a few pundits, futurists, and normal people weighing in on how AI might end the world. Skynet is a popular villain. Ultron if you’re younger, HAL9000 if you’re of a certain age. The plot is always the same: AI goes bad and attempts to destroy the world until an unlikely human hero saves us. It makes for a good story, but the real threat posed by AI isn’t that obvious, nor is it anywhere near as engaging – which is why it is extremely dangerous.
AI Addiction and General Use
Humans (especially those who use technology to compete with each other for a living) are about to become seriously addicted to AI. We were already well on our way: social media algorithms (AI) for news and entertainment, recommendation engines (AI) for shopping and entertainment, Waze (AI) for way-finding, etc. But in the post-ChatGPT era (aka The Generative Era), almost everyone who is serious about knowledge work or content creation has started to use some kind of generative AI tools to increase their productivity. More generally, office workers across the world are learning that they need to be proficient with Microsoft Copilot just to “meet expectations” on their performance reviews. Stating the obvious, AI is being woven into the fabric of our lives at an alarmingly rapid rate. While the technology will evolve (more rapidly than anyone can cope with), there is no sign that this adoption trend will ever slow down or decrease.
The More AI We Have The AI More We’ll Need
Satya Nadella may be the greatest CEO of our time. He is a master strategist and has taken Microsoft to places few thought it could ever go. So it should not surprise anyone to learn that Microsoft and OpenAI are reportedly planning to build a massive US-based supercomputer called “Stargate,” with a projected cost of around $100 billion (about 100 times more expensive than the largest data centers currently in operation). The project would be the centerpiece of a five-phase plan focused on a series of supercomputer installations over the next six years. Stargate, which would be phase 5 of the plan, could launch as soon as 2028. Why invest $100 billion in a souped-up data center? Because Microsoft has done their best to calculate future demand.
You can be sure that Stargate will not be the only AI super-data center to be built. Russia, China, and India will build their own versions. As will dozens of other governments and commercial enterprises.
As you can imagine, massive AI-focused supercomputers will require millions of chips and huge amounts of electricity. Where will this come from?
Historically, Humans Fight For Resources
As humanity becomes more and more addicted to AI, the need for access to AI-focused supercomputers (“compute”) will slowly, but surely accelerate. As a society, we won’t really notice it. AI will be incorporated into all our tech. Some AI will run on our local devices, but the velocity of our need for compute will always increase. Which leads to a fairly scary logical conclusion. We’re going to fight for what is likely to become a scarce resource.
Massive Chip Shortages
The semiconductor industry is currently facing a major challenge: demand for advanced 5nm and 7nm chips. These chips are essential for AI compute. The complexity and cost of manufacturing these advanced chips means few companies dominate the market. For some AI-specific chipsets TSMC is practically a sole source. This concentration of manufacturing capability has led to concerns about supply shortages. With fabs operating at full capacity, any disruption will have ripple effects across the globe, affecting everything from automobile manufacturing to consumer electronics​​.
The industry’s response has been a flurry of construction, with new semiconductor fabs being built around the world, aimed at increasing the production capacity for these advanced chips. For example, TSMC’s Arizona Fab 21, focusing on 5nm technology (the types of chips needed for the best-in-class AI systems), is a part of this expansion but has seen delays, pushing its opening to 2025 due to challenges like finding skilled workers​​. These expansions are critical in meeting the growing demand but also underscore the fragility of our current technological ecosystem, heavily reliant on a handful of advanced manufacturing hubs.
You have to ask: Will demand exceed supply? And, if so, when? It’s hard to predict the answer to these questions, but I’ve heard estimates in the 60 month range. Assuming none of the current fabs are damaged in any way, progress will be limited by the availability of chips – people will fight for this resource.
Lack of Power
The exponential growth of AI has a less discussed but potentially more dangerous implication: a massive increase in power consumption. Forgetting the environmental impact (which will be significant), where will this power come from? There won’t be enough to go around. In the US, AI’s power consumption needs will directly compete with EV (electric vehicle) mandates. We’re either going to discover a way to do cold fusion (star in a jar) or start building a lot more power plants. Will they burn coal? Other fossil fuels? Are we getting back into fission reactors? The power will have to come from somewhere. People will fight for this resource.
This doomsday scenario resembles Nick Bostrom’s paperclip maximizer thought experiment. Not in the specifics but in the underlying principle: humanity inadvertently creates a system that consumes overwhelming resources, in this case, electricity and semiconductor manufacturing capacity, to sustain AI’s growth. And so the world ends not from a malevolent AI system hell-bent on destroying humanity, but by humanity destroying itself by fighting over the resource to serve its addiction to AI.
Addressing the Challenges
The challenges of chip shortages and power demands highlight the need for a balanced approach to AI development. This includes investing in renewable energy to meet the growing power demands sustainably and improving the efficiency of AI algorithms to reduce their environmental footprint. BTW, a scientific breakthrough that reduces the need for exponential increases in chip production or energy consumption would make the thesis of this essay meaningless.
That said, the real risks AI poses to the world may not come from super-intelligent machines with malevolent intentions but from the very human oversight of not accounting for the environmental and infrastructural costs of rapidly advancing technology. Ensuring that AI serves the common good requires not just technical and ethical guidelines for AI development but also a comprehensive approach to managing the resources it depends on.
Author’s note: This is not a sponsored post. I am the author of this article and it expresses my own opinions. I am not, nor is my company, receiving compensation for it. This work was created with the assistance of various generative AI models.
ABOUT SHELLY PALMER
Shelly Palmer is the Professor of Advanced Media in Residence at Syracuse University’s S.I. Newhouse School of Public Communications and CEO of The Palmer Group, a consulting practice that helps Fortune 500 companies with technology, media and marketing. Named LinkedIn’s “Top Voice in Technology,” he covers tech and business for Good Day New York, is a regular commentator on CNN and writes a popular daily business blog. He's a bestselling author, and the creator of the popular, free online course, Generative AI for Execs. Follow @shellypalmer or visit shellypalmer.com.
