AI’s $600 billion dilemma: Where’s the real value?

The artificial intelligence sector is rapidly approaching a pivotal $600 billion moment, driven by unprecedented investments in infrastructure and soaring company valuations. However, a deeper examination reveals a concerning disparity between the vast sums being invested and the actual value generated. In his recently updated analysis titled 'AI’s $600B Question', David Cahn revisits his previous forecast from September 2023, which highlighted a staggering $125 billion annual gap between AI infrastructure spending and the revenue it produced. Alarmingly, that gap has now expanded fourfold. Cahn emphasizes the urgency of this issue, particularly in light of Nvidia's ascent to become the world's most valuable company. "If you run the numbers today, that $200 billion question has evolved into a $600 billion dilemma,” he states. Cahn's latest calculations are informed by Nvidia's anticipated revenue growth. By adjusting the figures to account for the total costs associated with AI data centers, along with the profits from companies reselling computing power, the disparity between infrastructure investment and realized revenue has widened significantly. He notes that the race for AI infrastructure has shifted from supply shortages to a surplus. "Startups were scrambling for GPUs in late 2023, but now they are readily available with shorter wait times,” he remarks. Major cloud providers, such as Microsoft, have reportedly amassed substantial GPU stocks, leading to an increasing surplus in the market. Among the few exceptions, OpenAI has distinguished itself by generating considerable revenue, reportedly achieving $3.4 billion annually—up from $1.6 billion just months prior. However, most consumer-oriented AI applications continue to struggle with providing consistent value akin to services like Netflix or Spotify. Cahn estimates that even with significant contributions from tech giants like Google, Meta, Apple, and Microsoft, along with firms like ByteDance and Tesla, a staggering $500 billion gap would still persist. Cahn warns that this situation should serve as a critical alert for the industry. He contrasts the current AI investment landscape with historical railroad developments, arguing that the analogy falls short due to fundamental economic differences. "Railways had pricing power, which AI compute lacks," he explains. As GPUs for data centers become increasingly commoditized, new AI cloud providers are entering the market, leading to diminishing profit margins and rapidly depreciating hardware. The depreciation rate is also intensifying, as evidenced by Nvidia's newly announced B100 chip, which offers 2.5 times the performance of its predecessor at only a 25% higher cost, effectively making older hardware obsolete more quickly than anticipated. Despite the cautionary narrative for investors, Cahn remains hopeful about the future for innovators. "Falling GPU compute prices are advantageous for startups, making experimentation and development more affordable," he asserts. "Founders who focus on delivering genuine user value are likely to succeed. This is how significant companies emerge." While speculative enthusiasm can attract capital, it often skews expectations. "AGI isn’t arriving tomorrow, and GPUs aren't gold bars," he concludes. As the excitement surrounding AI grows, Cahn's analysis calls for the industry to confront the widening financial chasm and prioritize the creation of sustainable, long-term value. The next phase of AI's journey may hinge on a more refined focus rather than merely expanding data center capacities.