The $600 Billion Focus on Artificial Intelligence

The artificial intelligence industry is rapidly approaching a $600 billion turning point, driven by increased investments in infrastructure and unprecedented valuations. However, a detailed examination shows a significant difference between the funding received and the actual value generated.

Revisiting AI’s $600B Question

In a refreshed analysis entitled AI’s $600B Question, David Cahn revisits his September 2023 prediction, which had previously identified a $125 billion annual shortfall between infrastructure investment and the resulting AI revenue. This gap has now increased fourfold.

“Nvidia’s emergence as the world’s most valuable company has intensified this question,” notes Cahn. “If we assess the numbers today, the $200 billion concern has transformed into a $600 billion dilemma.”

The revised estimation is grounded in Nvidia’s anticipated run-rate revenue. By doubling this figure to account for the total expenses of AI data centres and doubling it once more to reflect the gross margins for companies reselling computation, the infrastructure expenditure now significantly surpasses generated revenue.

The Shift in AI Infrastructure

Cahn observes that the race for AI infrastructure has shifted past supply shortages. “In late 2023, startups were reaching out to anyone who could supply them with GPUs. Now, GPUs are readily accessible with relatively short delivery times,” he indicates. Additionally, cloud services such as Microsoft have reportedly accumulated large stockpiles, contributing to an increasing surplus.

Identifying Industry Leaders and the Expanding Gap

OpenAI emerges as one of the few entities achieving notable revenue. Current reports indicate it has reached an annual revenue of $3.4 billion, an increase from $1.6 billion just a few months prior. However, few others are keeping pace. Most consumer-focused AI applications still lack the recurring value realized by services such as Netflix or Spotify.

Cahn calculates that even if tech behemoths like Google, Meta, Apple, and Microsoft each generate $10 billion in new AI income, alongside other significant players like ByteDance or Tesla contributing $5 billion, a $500 billion deficit remains. “This should be a crucial alert,” he cautions.

Critique of the Railroad Analogy

Some have compared AI investment to the construction of railroads, with revenue and applications expected to emerge later. However, Cahn states that this analogy overlooks essential economic distinctions.

“Railways enabled pricing control. AI computation lacks this,” he clarifies. “Data centre GPUs are rapidly becoming commoditized, with new AI cloud providers consistently entering the market. This scenario is conducive to diminishing margins and depreciating hardware.”

Furthermore, depreciation is occurring at a more accelerated rate. Nvidia’s recently unveiled B100 chip offers 2.5 times the performance of its predecessor for merely 25% additional cost, which effectively makes existing hardware obsolete more swiftly than investors might anticipate.

Long-Term Opportunities Ahead

While Cahn paints a cautionary picture for investors, he retains optimism about the long-term prospects for innovators.

“Falling prices for GPU computation are beneficial for startups. The expense of experimentation and creation is decreasing,” he adds. “Entrepreneurs who concentrate on genuine user value will succeed. This is how significant companies come into being.”

While speculative enthusiasm may channel investment, it also skews expectations. “AGI isn’t around the corner, and GPUs should not be viewed as gold bars,” he concludes. “We are amidst a tech wave that will define generations, but we must remain realistic.”

As the enthusiasm surrounding AI grows, Cahn’s analysis encourages the sector to confront the expanding financial gap and prioritise the delivery of measurable, long-term value. The future phase of AI’s progression may not centre on larger data centres, but rather on a more focused approach.