Wall Street’s love affair with artificial intelligence stocks might be heading for a sobering moment. The S&P 500’s meteoric rise — a 91% rally over five years — has pushed valuations into territory that’s making even seasoned investors nervous. Initially fueled by pandemic-driven tech reliance and later by AI advancements, this growth suggests a potential drop in US stock prices could be approaching.
Predicting the unpredictable: The “Magnificent Seven” tech giants have long dominated markets, with market value influenced more by AI hype than current earnings. Despite strong profits, their recent increases in AI spending have led to a broader market pullback. Meanwhile, their billion-dollar investments in AI technology have yet to deliver the anticipated returns, hinting that the industry’s pace of innovation may be slowing. This raises questions about whether the optimistic, “priced-in” valuations can withstand the realities of slower technological progress and escalating costs.
- The S&P 500’s cyclically-adjusted price-to-earnings (CAPE) ratio now stands at 38 times earnings — well above the 20-year average of 27x.
- Nvidia, the face of the AI boom, trades at 48x its 2024 earnings — though analysts project that multiple to fall to 16x by 2029, assuming earnings growth continues.
Byte-Sized Battles
Even with rapid advancements, companies like OpenAI, Anthropic, and Google are struggling to meet expectations as AI model improvements slow. OpenAI’s latest model, Orion, has fallen short of expectations, particularly with coding tasks, compared to the dramatic leaps seen between GPT-3 and GPT-4. Similarly, Google’s next-gen Gemini model and Anthropic’s awaited 3.5 Opus have hit similar stumbling blocks.
- Difficulties in acquiring high-quality training data and rising development costs are complicating progress — challenging the belief that simply boosting computing power and data automatically makes AI systems better.
- The cost-benefit equation is becoming increasingly problematic, with Anthropic’s CEO warning that training expenses could balloon to $100B in the coming years, casting doubt on the economic viability of future developments.
Reality checkpoint: Recent reports suggest that Large Language Models (LLMs) have reached a point of diminishing returns, with Marc Andreessen noting that more computing power isn’t yielding proportional intelligence gains. Companies are shifting their focus from raw computational power to enhancing reasoning abilities and creating specialized applications, like AI agents designed to book flights or manage emails. The coming months will reveal whether AI can sustain its lofty market valuations or if Newton’s laws still govern Wall Street’s fluctuations. For now, the industry faces a critical moment as it balances high expectations with the realities of innovation and cost.
