
The Worst is Almost Over
Audio Summary
AI Summary
The current financial landscape surrounding Artificial Intelligence (AI) is showing significant cracks, suggesting that the unsustainable surge in investment may be reaching its limit and could potentially destabilize the global economy. Evidence of this is emerging, such as reports of declining DDR5 prices worldwide, a key component in AI infrastructure. While many AI leaders maintain that the market is not in a bubble, OpenAI CEO Sam Altman himself has previously acknowledged the irrational flow of money into AI and the possibility of investors getting burned. His credibility has been questioned following a reported attempt to secure 40% of major DRAM producers' capacity, only to retract the commitment, a move likened to a broken promise.
Understanding what constitutes a bubble is crucial. Investopedia defines it as an asset's price skyrocketing far beyond its intrinsic value due to a rush of buyers hoping for quick profits. The historical example of "tulip mania" in the 1600s illustrates this, where tulip bulb prices soared to astronomical levels before a dramatic crash, wiping out significant market value and leaving many investors in debt. While the asset itself doesn't become worthless, its value can become disproportionately low compared to other assets.
Several factors point to AI potentially being in a bubble. The valuations of leading AI companies are a primary concern. Companies like Anthropic ($380 billion), XAI ($200+ billion), and OpenAI ($852 billion) are approaching trillion-dollar valuations. While high valuations aren't inherently problematic, they typically correlate with substantial revenues. For comparison, Samsung, a major player in AI hardware manufacturing, has annual revenues exceeding $220 billion. Although OpenAI has reported revenues of over $2 billion per month, a remarkable feat, revenue does not equate to profit. Samsung, for instance, generates over $28 billion in annual profit, while OpenAI, despite its revenue, is reportedly burning billions annually after more than a decade of operation with no clear path to profitability.
OpenAI's strategy appears to involve massive infrastructure investment to make AI as ubiquitous as electricity, with the eventual goal of monetizing locked-in customers. However, with 900 million users already, it's questionable whether acquiring the remaining global population will bridge the gap between losing and making billions. Enterprise solutions present another avenue, but businesses are likely to be critical of AI that falls short of true artificial general intelligence. The integration of AI into enterprise has been inconsistent, with personal experience suggesting it can be more of an irritant than a help. Even with widespread adoption, companies may opt for locally hosted, open-source models over outsourcing to OpenAI, especially considering competitors like Anthropic and Google.
OpenAI's performance is often seen as a proxy for the broader AI market, similar to how Bitcoin's movements influence the cryptocurrency industry. Recently, OpenAI has faced setbacks. Reports indicate they have significantly reduced their spending expectations, targeting only $600 billion by 2030, down from previous ambitions. This could impact their ability to secure large amounts of memory, potentially easing pressure on competitors.
Energy costs represent another significant expense for AI companies, second only to hardware. These costs have escalated over the past five years, exacerbated by current global geopolitical instability. This may have contributed to OpenAI's abrupt cancellation of its Sora 2 video generation app, reportedly due to prohibitive running costs. Oracle, despite announcing a partnership with OpenAI to build AI infrastructure and boasting about exceeding demand, recently laid off 18% of its global workforce. While Oracle's stock has seen a temporary rise, its overall trajectory has been downward, and it's currently trading at about half its all-time high. Microsoft is experiencing a similar correction, with its stock only slightly above its November 2021 peak.
Not all AI companies are struggling. Nvidia continues to thrive by supplying essential hardware. Google and Meta have experienced smaller dips, more in line with general market corrections during times of geopolitical uncertainty. Despite these challenges, AI is not disappearing. Meta, for example, is heavily investing in integrating AI across its vast product ecosystem, aiming to have its 3 billion-plus users interact with AI daily. However, the speaker doubts that companies like OpenAI will own this future infrastructure.
OpenAI is compared to a "Dropbox-like company" – one that might survive the bubble pop due to its brand recognition and ubiquity but could struggle to convert free users to paid ones, potentially settling for a niche role. Meta and Google, in contrast, are seen as more resilient because their valuations are grounded in existing product usefulness and revenue streams outside of AI. This existing foundation shields them from the brunt of the bubble's deflation. As OpenAI falters, competitive pressure on infrastructure investment may lessen, allowing companies like Meta and Google to scale their investments more prudently as they discover practical applications for AI technology.
The author believes that OpenAI's DRAM commitment issue has rightly made multi-billion dollar investors question the certainty of future revenues. This realization is expected to bring financial pragmatists back into control, leading to more reasonable AI investment levels and a focus on generating actual revenue from AI services. The process of this correction is anticipated to be gradual, akin to the slow unwinding of tulip mania, suggesting a light at the end of the tunnel for consumers. The video also briefly mentions ProtonMail as a privacy-focused email alternative and a t-shirt available for purchase.