Last year, I wrote a blog on the growth of AI, where I broke down the public hype around the technology and the barriers preventing it from scaling further. Earlier this year, Gartner projected that worldwide AI spending will reach $1.5 trillion by the end of 2025, up $490 billion over last year.
Despite growing billion-dollar investments, many businesses in 2024 were not getting the value they anticipated from AI. Now that almost a full calendar year has passed, I wanted to check back in on this. In summary, not much has changed.
On the optimistic end (yes, really), McKinsey reported that 80% of companies have seen “no significant bottom-line impact” from their AI investments. Meanwhile, MIT proposed this figure to be as high as 95%. This value gap has garnered substantial attention, with many sources calling this an “AI bubble.”
The truth is that AI does pose significant economic risks.
AI Risks to the Global Economy
Let’s start the risk conversation by looking at semiconductor industry growth. During this year’s SEMICON West Market Symposium, Christopher Thomas from Integrated Insights highlighted that if we were to remove Nvidia from the equation, semiconductor industry growth has actually been flat for the last three years. This is true on a national economic scale as well. Without data center investments, U.S. economic growth was just 0.1% in the first half of 2025.
Circular deals between AMD, OpenAI, Nvidia, CoreWeave, Oracle, and others have also raised concerns about the true value of AI.
Open AI now has a 10% stake in AMD, and Nvidia is also investing $100 million in Open AI. At the same time, CoreWeave – a cloud computing company – has agreed to sell Nvidia systems to Open AI, and Nvidia is also one of CoreWeave’s top investors. Finally, Oracle is spending $40 billion on Nvidia chips for Open AI’s new U.S.-based data center.
Given the interconnected nature of these companies, if one fails, it could trigger a domino effect across the entire industry. In addition, Open AI isn’t profitable yet, but it appears to be laying the foundation for a $1 trillion IPO.
The True Cost of AI
The global economy has bet substantially on AI, so there’s a massive push to make the technology work. Right now, AI spending is largely attributed to the following:
- Data center buildouts: Hyperscalers are rushing to build data centers to scale AI, with the goal of securing a competitive advantage and ultimately driving down compute costs.
- Integration into businesses: As more businesses adopt AI, more industry-specific use cases are needed. More use cases require more compute power, which will eventually lead to more robust AI applications.
- Government spending: Governments around the world are investing in AI infrastructure to bolster their national security and economic growth.
AI may offer a ton of promise, but only if it can address profitability concerns due to ongoing scaling challenges. Right now, the plan is to spend heavily in the hopes of realizing its potential, but even with Large Language Model (LLM) improvements and mass infrastructure buildouts, AI’s reliance on expensive resources still prevents it from fully scaling. In addition, costs soar when efforts to strengthen the technology are made.
For example, Epoch AI reported that spending for training LLMs is growing at a rate of 2.4x per year, half of which is on GPUs alone. While we can argue that the price of AI tokens is at least dropping, the rising number of tokens needed to execute complex tasks negates such progress.
AI’s aggressive power demands are also fueling massive spending. The U.S. alone plans to allocate a staggering $1.1 trillion over the next five years to prepare its energy grids to handle more advanced AI, which won’t be an easy task.
Final Thoughts
All of these factors paint a complicated picture for AI’s future profitability. I think Derek Thompson summed this up best in his blog, where he said that “AI will rise first, crash second, and eventually change the world.”
But what’s the true purpose behind all of this spending? From what I gathered, companies are pouring vast amounts of resources into AI in the hopes of one day achieving Artificial General Intelligence (AGI), or an AI system that can match or exceed the capabilities of a human brain. The New York Times reported that “no one” has articulated how companies will make money from this, classifying the hundreds of billions spent on the potential of AGI as “a leap of faith.”
Although we can point to examples like autonomous vehicles, ChatGPT, and others as proof that AI has already begun to change the world, we still can’t ignore the massive value gap between the lofty promise of AI and what the technology is currently capable of. The problems I outlined in last year’s blog – electricity concerns, interconnect limitations, and manufacturing bottlenecks – still stand today, not to mention other concerns like worker shortages, materials constraints, and more.
Hyperscalers and government bodies will keep throwing money at these problems to make them go away, but even with billions already spent, new concerns arise anytime improvements are pursued. To be honest, I’m not sure that investors fully understand what they’ve gotten themselves into, but at this point, I think it’s too late to turn back. AI is propping up the global economy, and halting progress will likely lead to a crash.
All of that to say that the future profitability of AI is largely unknown for now, but that’s not to say the technology will never get there. I think it’s too soon to tell exactly where AI is headed and how it’ll get there, but we know it’s here to stay.
