Imagine a technology so transformative that companies are willing to bet billions�even trillions�on its future? That’s exactly what’s happening in artificial intelligence right now, as OpenAI CEO Sam Altman defends the sector’s “bubbly” characteristics while insisting the underlying value is real? But is this massive investment wave sustainable, or are we witnessing history repeat itself with another technology bubble?
The Compute Gold Rush
OpenAI’s recent partnership with AMD represents just one piece of a staggering infrastructure build-out? The deal involves 6 gigawatts of compute capacity�enough power to rival Singapore’s entire electricity consumption�with AMD granting OpenAI a warrant for up to 160 million shares, equivalent to 10% of the chipmaker’s outstanding stock? This follows OpenAI’s $100 billion partnership with Nvidia and a $300 billion commitment to Oracle Cloud Services, creating what some analysts call a “circular structure” of inter-company deals?
Altman argues this isn’t irrational exuberance? “We can still monetize every GPU we get our hands on super well,” he told reporters at OpenAI DevDay 2025? “The degree to which we could have 10x of compute, we could build so many more products and offer so many more services people would love?”
The Bubble Warning Signs
However, financial analysts see troubling parallels with past technology busts? According to Absolute Strategy Research, AI stocks exhibit classic bubble characteristics: skyrocketing share prices, excessive index concentration, and inflated valuations through vendor financing? The analysis draws direct comparisons to the late-1990s TMT bust, where Microsoft fell 65%, Apple 80%, Oracle 88%, and Amazon 94% from their peaks?
William Janeway, author of “Doing Capitalism in the Innovation Economy,” offers a nuanced perspective: “Periods of bubble behaviour�and especially excess capex�are central to the adoption of new technologies? The hype around them drives down the cost of capital, allowing the rapid build-out of the new technology?”
The ROI Reality Check
While the investment frenzy continues, the return on investment remains uncertain for most enterprises? An MIT study found that 95% of companies attempting to harness AI aren’t seeing measurable results in revenue or growth? Gaurav Gupta, Gartner’s VP Analyst in Emerging Trends and Technologies, acknowledges this gap but sees it differently: “We have been talking about the AI bubble, where enterprises are finding it hard to achieve ROI beyond initial productivity gains with AI? On the other hand, you can see hyperscalers, frontier labs, and advertising companies continue to spend to get access to more compute?”
The numbers are staggering: OpenAI’s total committed capacity across all deals reaches 23 gigawatts, estimated to cost over $1 trillion to develop? Meanwhile, the company reported $4?3 billion in revenue with $2?5 billion in cash burn during the first half of 2025, highlighting the massive gap between investment and current returns?
Competitive Pressures Mount
The investment surge isn’t happening in a vacuum? Europe is leading on AI regulation with the European AI Act, while competing models like Deep Seek from China and K2 from the United Arab Emirates use less computing power, potentially offering more efficient alternatives? These developments create additional pressure on Western AI companies to justify their massive compute investments?
Altman remains optimistic about the long-term trajectory? “People will overinvest in some places,” he concedes? “There will be numerous bubbles and corrections over that period, but what I don’t think [is that] this is totally divorced from reality�there’s a real thing happening here?”
The Infrastructure Endgame
The voracious appetite for computing power isn’t just about running existing AI models? Recent product releases continue to highlight the need for more capacity? Greg Brockman, OpenAI’s Vice President, noted that the number one lesson from the Sora video generator release was the need for additional compute power? Similarly, the new Pulse feature�which provides personalized news digests�is limited to Pro subscribers due to computational demands?
As Gupta explains, “This tells us that a lot more work still needs to be done on large language models and a race towards artificial general intelligence�hence all the crazy demand for compute?”
Balancing Optimism with Caution
The fundamental question remains: Are we building the infrastructure for AI’s future or creating the next great technology bubble? The evidence suggests both might be true simultaneously? While excess capital expenditure drives technology adoption and eventual ubiquity, it also leads to creative destruction where late investors often face significant losses?
For businesses and investors, the key lies in distinguishing between sustainable infrastructure development and speculative excess? As the AI capex endgame approaches, the companies that survive the inevitable corrections will be those building real value, not just riding the investment wave?