The AI Bubble: A Complicated Problem
Cover image by Valerie Tea
Like it or hate it, artificial intelligence (or as most know it, AI) is here to stay. Accompanying the AI boom back in 2022, things that were once seen as nothing more than advanced chatbots have become almost integral to your daily life. Nowadays, AI runs everything from analytics to content creation — what a human can do, an AI can do better and more quickly. Technology at its finest, something that both serves us while surpassing us at the same time. At least, that’s what people would like you to think. While there are plenty of benefits to integrating AI into fields such as software development and the biomedical industry, there are also equally terrible downsides, especially in creative fields. And behind the scenes, the bills continue to pile up as we unknowingly pay costs we might not be willing to. Today, we’ll be going over the ever-inflating AI bubble.
It began with programs like ChatGPT and DALL-E — chatbots and tools that most considered harmless entertainment. And then, with the rise of large language models and similar forms of generative AI began to grow ever-popular, an uncomfortable truth started to sink in; this was the future. And it looked nothing like what we expected. Companies like Microsoft, OpenAI and an ever-increasing number of large corporations began to sink time, money and technology into the development and growth of AI models. And why wouldn’t they? A thinking, learning machine that could do better than a human ever could, without having to pay them a salary? A businessman’s dream. And so began years of big companies convincing people that yes, AI was the future and yes, it was everything. Worst of all, people started to believe it.
Now, there are benefits to utilising AI as a tool in many industries. At its core, most AI models are essentially more advanced computing systems, far better than humans at processing large amounts of data, filtering and catching details much more quickly and efficiently than your average human being ever could. Industries that rely heavily on processing such data (such as the medical industry) can benefit greatly from using such technology to automatically screen through and flag vast numbers of reports, menial and time-consuming tasks. In a similar vein, AI can be used to write code more quickly and offload the boring, tedious parts of coding (that is, writing the code itself) while the humans do the editing and debugging needed to turn that code into something worth using. This is something most people can agree on — when it comes to tasks that involve large amounts of data, humans simply cannot match up to AI.
However, we’ve also seen just as many issues with the ever-increasing usage of AI. From an individual’s perspective, AI has become a crutch so important that people have forgotten what it was like to walk without it. For instance, many fields in the creative industry are under attack from the spread of generative AI. People would rather ask an AI model to generate a picture and pretend they had a part to play in it rather than actually pick up a pen. Much worse is the fact that such models are, in essence, trained on large amounts of data from existing artists and writers…most of which never actually consented to having their work be fed into a machine. For instance, something like this — the use of an em-dash, while once just another writing technique, now becomes one of the indicators of AI-generated writing. Because when it comes down to it, generative AI only does one thing; consume everything it’s fed, and reproduce it mindlessly. Or, in other words, AI slop.
Even ignoring the fact that people are growing increasingly reliant on machines to do their thinking for them, AI is wreaking havoc on the environment. As large language models grow, an equally large cost is necessary to continue “improving” them. Data centers required to train such LLMs require an obnoxious amount of electricity to function and be maintained, along with vast amounts of water to use for evaporative cooling. Even simple tasks like generating an answer to a prompt consume an excessive amount of energy to do so. Simply put, AI is not energy efficient. And the more we feed it, the more we destroy any sort of hope that we can sustain the environment so that future generations aren’t rewarded with a planet on the verge of death.
Even if you consider the energy consumption of AI somewhat reasonable considering the amount of data processing that goes behind the scene, no one can deny the fact that AI is power hungry (literally) and more people should really take that into account considering how people treat AI models like convenient chatbots at this point. Considering ChatGPT is the fifth most visited website in the world, standing shoulder to shoulder with apps that are probably older than the majority of people reading this, it should be reasonable to think about what goes on behind the scenes of these LLMs.
Still, countries like the US and China continue to sink time and money into the AI industry, continuing to produce AI models and data centers and chips to feed the hungry machine. Think of the Stargate initiative, which aims to spend 500 billion USD (more than the Apollo space program!) over the next couple years with the intention of building even more data centers, each of which could require five gigawatts to power. You can see their influence from the way every social media app nowadays needs an AI chatbot you can ask for assistance, or how computers come with AI programs pre-installed for free. How very generous.
Perhaps there are those that truly believe that AI is the future, that all it’ll take for humanity to eventually never have to do anything ever again is a few billion more dollars into a stagnant industry. Or perhaps it’s another sign of big companies jumping onto the bandwagon of whatever’s most popular at the time. We’ve seen the signs before. The rise of social media platforms and their optimisation. The way said platforms began to curate their content to factor in the ever-decreasing attention span of their consumers, all style and no substance. And now, perhaps just another step in that process, how everything needs to have some form of AI in some way. It’s the same as an economic buzzword; people see AI-integration the same way they might have seen better processors for their computers and better graphics for their screens not even a decade ago. From a logical, economically efficient standpoint, it’s easier for companies to keep pushing the agenda. Why stop what’s already working?
Regardless of what AI may be doing to the environment (and our wallets), the AI industry continues to grow. As of right now, the worldwide market continues to exhibit a steady annual growth rate, projected to be about 18% per year for at least the next five years. Even countries in Southeast Asia have jumped onto the bandwagon, with Vietnam attempting to establish itself as a regional AI hub, or Malaysia sinking their own funds into data-center investment. Whether it be because countries are afraid to fall behind or they truly think that this is a good investment for their future, no one can say for certain. But regardless of what we might think, governments continue to promote and invest in AI, throwing their support behind its continued growth in order to ensure that their countries remain relevant both in the region and on a global stage.
Despite an ever-increasing number of reasons and evidence to point towards the stagnation and eventual downfall of the AI industry, it remains bigger (unfortunately not better) than ever. So begs the question, when will the AI bubble pop? If an industry that openly demonstrates its lack of profitability and is largely funded by debt can continue to grow, when will enough finally be enough? Will the overevaluation of the AI industry eventually lead to another financial crisis people will remember for the history books? Well, contrary to what AI-bros want you to believe, it might be happening sooner than you think.
OpenAI is projected to run out of money by mid-2027, and expects to report annual losses in the billions even if it somehow manages to survive. And if the company that arguably started it all is struggling this badly, it’s not too far of a stretch to imagine that other similar AI tech firms will crash and burn just as spectacularly in the next decade. Because, truthfully, the AI industry has stagnated. That, coupled with the fact that we could very well be paying more and more to keep the power grid running as it feeds the energy monster, along with the fact that companies continue to hide how much investment is going into this modern age black hole, there’s a good chance that we might be standing at the end of it all with nothing to show for it but depleted resources we will never get back. Unless something revolutionary happens soon, it is not ludicrous to consider whether the AI bubble will burst — it’s a matter of when, rather than if.
Still, AI is (unfortunately) here to stay. Yes, we can continue to benefit from using it as one of our many technological tools, but those who over-rely on it and those who continue to follow it blindly may have to consider if they still remember how to function in a world before AI. If we plan to keep AI as part of the foreseeable future, using it for everything and anything probably isn’t the way to go about it, especially as more and more people do so — all of this adds up behind the scenes, and blinding ourselves to the fact of the matter will do nothing to change it. Most of all, we must remember that it has and will continue to cast a dark shadow over the future. Perhaps the ramifications may not come in our generation, or the next, but the costs have and are still being paid. And perhaps it’s about time we consider what sort of world we want to leave for those who come after us. Yes indeed, the AI bubble will burst. What you do with that, however, is up to you.
References
Barrientos, A. (2026, July 10). OpenAI could run out of cash by Mid-2027 – bleeding $1.69 for every dollar earned. Yahoo Finance. https://www.gadgetreview.com/openai-could-run-out-of-cash-by-mid-2027-bleeding-1-69-for-every-dollar-earned
McCurdy, W. (2025, December 7). ChatGPT now has more monthly visitors than these top platforms. PCMag UK. https://uk.pcmag.com/ai/161892/chatgpt-overtakes-amazon-x-reddit-whatsapp-and-wikipedia-in-visitors
McMahon, G. (2024, March 8). From buzzword to game-changer: Adapting to the AI revolution. Forbes. https://www.forbes.com/councils/forbesbusinesscouncil/2024/03/08/from-buzzword-to-game-changer-adapting-to-the-ai-revolution/
Reitmeier, L., & Lutz, S. (2025, September 12). What direct risks does AI pose to the climate and environment? - Grantham Research Institute on climate change and the environment. Grantham Research Institute on Climate Change and the Environment. https://www.lse.ac.uk/granthaminstitute/explainers/what-direct-risks-does-ai-pose-to-the-climate-and-environment/
The Economist. (2026, July 30). How to spot AI writing. The Economist. https://www.economist.com/culture/2026/07/30/how-to-spot-ai-writing
Comments