The piece anticipates a significant shift in science and technology following the current hype around large language models (LLMs). The author argues that while LLMs have their place, true advancements will come from leveraging massive GPU infrastructure, potentially available for personal and institutional use. This shift could lead to a new wave of supercomputing that empowers fields like medicine and engineering, making complex tasks faster and more efficient. The author envisions a future where commodity GPUs enable breakthroughs in areas like genome sequencing and real-time vaccine development. The author references Geordi La Forge from Star Trek as a metaphor for the relationship between expertise and technology. Just as La Forge's skill allows him to effectively use advanced systems, human expertise is essential for maximizing the potential of AI tools. There's a clear distinction between the capabilities of machines and the nuanced understanding that trained professionals bring to their fields. As AI technology matures, the need for skilled users who can interpret and challenge AI outputs becomes more pronounced. Financially, the article emphasizes the unsustainable nature of current investments in data centers, citing a cost of about $80 billion to build a one-gigawatt facility. As firms chase aggressive growth in AI, the cumulative investment could reach $8 trillion, with returns looking increasingly unlikely. The author suggests that the market may soon face a glut of unused or underutilized computing resources as companies struggle to recoup these massive investments. The focus shifts to who will take advantage of these distressed assets when the bubble bursts, hinting at a future where repurposing existing infrastructure becomes a key strategy.