Ben Horowitz on AI Infrastructure, Economics and The New Laws of Software

The United States faces a critical need to rebuild its entire infrastructure, including shortages in rare earth minerals, electricity, and manufacturing capacity. While companies like Nvidia may produce enough chips, other components like memory and electricity remain significant bottlenecks. This situation presents a stark contrast to China's rapid growth in these areas, raising concerns about the US's future technological standing. The history of technology shows a consistent trend of improvement, with humans demonstrating an incredible ability to innovate and meet new needs. However, the current technological shift, particularly with the advent of AI, is introducing fundamental changes that challenge long-held assumptions in the business world. One significant change is the ability to "throw money at the problem." Previously, it was understood that simply hiring more engineers couldn't accelerate a product's development if it was already years behind. This "mythical man-month" concept is no longer entirely true; with sufficient funding and data, companies can acquire enough GPUs to solve complex software problems. Another fundamental shift concerns customer lock-in. In the past, companies with a customer base benefited from various lock-ins, such as migration pain, data lock-in, and user interface lock-in. These are now largely diminished. Code is easier to replicate, data is more portable, and with AI interacting with software, traditional user interfaces become less critical. CEOs must recognize these changes and redefine their value proposition beyond these disappearing advantages. Pricing, for instance, must now be based on a much more distinct and unique value. The pace of innovation has also dramatically accelerated. What once offered a five to ten-year product lifecycle might now be reduced to mere weeks. This rapid disruption poses a significant challenge for legacy companies or those founded five to ten years ago, especially those that are "pre-AI." These companies face financial market skepticism and the risk of rapid obsolescence. While staying private longer might offer some protection during existential crises, the "SAS apocalypse" highlights doubts about terminal value, as companies risk being worth zero if they wait too long to adapt. CEOs must be brutally honest about their company's true assets. While some companies are correctly facing disruption, others are being unfairly undervalued. The key question is whether a company is strengthening or degenerating during this period of change. Companies that find their customers shifting to other products face a serious problem, likely requiring deep cuts and a pivot. However, some companies, despite valuation challenges, remain strong due to inherent complexities in their operations. For example, a travel company, despite the "SAS apocalypse," relies on extensive, explicit relationships with airlines and hotels globally, and a specialized sales channel to travel managers—factors that are difficult for AI-first companies to replicate quickly. This highlights that the impact of AI is highly company-dependent, and a "Brave New World" requires abandoning old frameworks. The current environment also makes it difficult to distinguish between a "feature," a "product," and a "company." The ease with which features can be created with AI blurs these lines, making it challenging to understand where true economic value lies. The venture capital landscape has also transformed significantly since the 2009 financial crisis. Funds have grown exponentially, with a much broader and more international investor base. This growth is partly driven by the immense capital needed to rebuild America's infrastructure and address critical shortages in areas like rare earth minerals, electricity, and manufacturing capacity. The demand for AI-related infrastructure, such as power transformers, is vertical, while the capacity to build it is not, leading to bottlenecks across the supply chain. This means investment is needed in fundamental areas that haven't seen significant innovation in decades. The rapid consumption of resources like RAM by AI initiatives exemplifies the latency problem. While the "cure for high prices is high prices" might eventually spur new production, building new factories takes years, creating immediate shortages. This situation is different from the dot-com bubble, where much of the fiber infrastructure remained "dark." Today, almost everything is a bottleneck, with electricity potentially becoming a limiting factor even before chip production. Venture capitalists must meticulously study the supply chain and invest in alleviating these bottlenecks, mirroring an approach seen in individuals like Elon Musk who tackle multiple bottlenecks simultaneously. AI also creates significant challenges that crypto solutions might address. The proliferation of AI-generated content raises concerns about authenticity, making it difficult to discern between human and bot interactions, or real and fake information. Cryptographic keys and blockchain technology could provide a verifiable source of truth, establishing identity and content authenticity. Furthermore, crypto could offer a solution for universal basic income (UBI) distribution, providing a secure and traceable "address" for sending money, addressing the inefficiencies and fraud seen in past government stimulus programs. Finally, crypto could enable AIs to become independent economic actors, allowing them to earn and spend money without human intermediaries, which is crucial for their integration into the economy. The future of venture capital itself is uncertain, with two main scenarios emerging. One possibility is a consolidation into a few gigantic companies, similar to the industrial revolution where early venture capitalists evolved into major banks. In this scenario, venture capitalists would move upstream with these dominant companies. The other possibility is a future where intelligence becomes a utility, potentially nationalized, allowing everyone to build upon it. This would lead to a very different venture capital landscape. The electricity shortage could further influence this, either empowering large companies that monopolize resources or pushing computing to the edge with more efficient, smaller models. Ultimately, venture capital could become much larger and more exciting, with a world of entrepreneurs, or new companies could become harder to establish. Despite the uncertainties and anxieties surrounding these changes, the historical macro trend suggests that technology consistently improves human lives. Just as electricity transformed society, AI holds the promise of a future where everyone lives better than in previous eras, with increased access to luxury, information, and new experiences. The transition is inherently scary due to the shift in job types and societal structures, but the long-term outlook is one of significant improvement, as humans continuously create new wants and needs that drive further innovation and economic activity.