From bits to atoms: AI is shifting tech's center of gravity

For the past twenty years, the tech investment scene has predominantly favored digital enterprises, often referred to as 'bits' businesses, such as software and online platforms. The reasoning was straightforward: software can scale easily, is distributed without delay, and yields high margins once established. In contrast, 'atoms' businesses, which involve tangible assets like factories and logistics, were perceived as slower, requiring more capital and thus considered less profitable. However, this perspective is undergoing a transformation. An increasing number of investors and entrepreneurs are now suggesting that the advent of artificial intelligence (AI) is changing the narrative, making businesses reliant on physical assets more appealing than their digital counterparts. A notable example is Travis Kalanick, co-founder of Uber, who is making a comeback with a new venture named Atoms, focusing on sectors like manufacturing and food logistics. Kalanick pointed out that while software has automated many tasks involving language and mathematics, the full automation of physical operations—what he calls autonomy—remains largely unexplored. This shift is attributed to AI's ability to disrupt supply dynamics across various sectors, leading to an increase in the availability of digital goods such as code and multimedia content. Joe Fath, a partner at Eclipse who invests in traditional industries, describes this as a significant structural change. Historically, software companies thrived because they required less capital to achieve sustainability, whereas asset-heavy enterprises faced higher risks and longer timelines. But now, Fath observes, this trend is rapidly evolving. AI is not only streamlining digital tasks but is also making physical industries more programmable. Innovations in robotics and systems that integrate visual recognition with action are enabling businesses to adopt efficiencies typical of software-driven models in areas like manufacturing and energy production. Fath states, 'For the first time, AI is making physical industries meaningfully programmable, paving the way for faster scaling with reduced capital and labor requirements.' Simultaneously, traditional software business models are feeling the pressure. With AI enabling anyone to produce decent code or applications, maintaining a competitive edge has become increasingly challenging. This trend is already evident in the stock markets, where software valuations are experiencing downward pressure. Michael Bloch from Quiet Capital recently questioned the safety of investing in software, especially as even companies focused on AI face uncertainties stemming from open-source alternatives and global competition. With the core technology becoming commoditized, the value is shifting towards infrastructure, the essential physical foundation of AI. Major tech firms like Microsoft, Google, Amazon, and Meta are investing heavily in data centers and hardware, indicating a shift toward 'atoms' even among traditionally 'bits' companies. As a Google employee noted in an internal memo, 'We have no moat, and neither does OpenAI,' reflecting a growing awareness that the fastest and most reliable delivery of AI models may depend more on physical infrastructure than on software alone. This trend is evident in various industries. Companies such as Tesla and SpaceX have long integrated software with real-world operations, and newer firms are emerging to exploit AI for enhancements in areas ranging from defense to manufacturing. While AI complements rather than replaces these physical operations, the demand for skilled trades in areas like construction and maintenance is rising, driven by the expansion of AI infrastructure. Elon Musk emphasized that while AI excels in digital tasks, it cannot fully replicate the hands-on work required in physical trades. Jobs that involve manual labor, such as welding and plumbing, are likely to persist longer, whereas digital roles are increasingly vulnerable to automation. Nevertheless, the transition to a more physical economy is not without its challenges. Scaling physical businesses is complex and requires substantial capital and expertise. Fath acknowledges this difficulty, stating, 'You're not just solving a technical problem; you must also scale it, which is incredibly challenging when it comes to physical production.' Despite these hurdles, the trajectory is clear: as AI diminishes the cost and value of digital work, scarcity—and thus value—is returning to the physical realm. Rohan Pandey, a former OpenAI researcher, encapsulated this sentiment by asserting, 'As AI reduces the cost of digital tasks to nearly zero, the focus and capital will revert to the physical world.' The future of technology, therefore, may evolve into a hybrid model blending both bits and atoms, with a renewed emphasis on the latter.