AI’s $600 Billion Gamble
David Carr, a partner at Sequoia Capital, has shared an intriguing post titled "AI is a $600 billion question." It presents a thought-provoking calculation about the AI industry's current investment landscape.
The post begins with a simple yet crucial question: How much revenue do companies buying AI chips need to generate to recoup their investment? Carr starts with the assumption that most companies are purchasing from Nvidia, whose data center business has a $250 billion run rate.
Considering that chip costs represent about half of the total data center expenses (including infrastructure and energy costs), the total spending reaches $300 billion. Assuming these companies aim for a 50% profit margin, they would need to generate at least $600 billion in revenue to justify their investments.
However, when Carr examines the actual revenue generated from these AI investments, the picture looks quite different. OpenAI, the market leader, has an estimated $3.4 billion in annualized revenue. Microsoft is generating about $10 billion from its Copilot product, while Google and others are also seeing some returns. In total, Carr generously estimates the industry's current revenue at around $100 billion - far short of the $600 billion target.
This discrepancy has led many to conclude that we're witnessing a significant AI bubble. Given current valuations and estimates, it's likely that many companies will be worth less in two years than they are now.
However, there's also a more optimistic perspective. Some argue that the potential value created by AI is so immense that these current losses are justifiable. In Silicon Valley, many believe that achieving Artificial General Intelligence (AGI) or Artificial Super Intelligence (ASI) could create nearly infinite value, making any investment worthwhile.
My view falls somewhere in between. I see parallels with the late 1990s tech boom. While there is indeed a bubble, it may not be as severe as the dot-com crash. We'll likely see some companies fail, but others, like Google or Amazon, will emerge stronger and more valuable than ever.
It's also worth noting that many companies investing heavily in AI chips have substantial cash reserves. Unlike the startups of the early 2000s, if a company like Microsoft spends billions on chips without immediate returns, it's not an existential threat to their business.
Moreover, these AI chips have versatile applications. For instance, if Meta doesn't succeed with its Llama project, the chips could be repurposed for other services like improving Instagram's recommendation algorithms.
While Carr's calculation is fascinating and provides valuable insights, I don't believe it necessarily indicates an imminent, catastrophic bubble burst. Instead, three big winners and the rest left holding the bubble.
What I think it means
Startups: Stay agile, like Amazon don't be afraid to take a leap when the opportunity comes and every bubble burst is high opportunity - but high risk.
Small / Medium businesses: look for utility outside the hype. AI is real, AI is here to stay if you create value. Build a brand out of your core utility that will last.
For people like me, honestly, not much. AI is real; we will keep using it no matter what. It just means there will be fewer options to choose from.
