In response to AI's potential to commoditize software, investors are shifting capital to "HALO" businesses like industrial manufacturing and aerospace. These sectors feature heavy physical assets and complex operations that are difficult for AI to replicate, promising lower obsolescence risk.
The defensibility of complex hard tech companies doesn't rely on a single patent or technology. Instead, their moat is "novel in the aggregate"—the difficult-to-replicate integration of dozens of complex systems across design, manufacturing, supply chain, and regulation. This holistic execution is the true barrier to entry.
As developers increasingly use AI coding assistants like Claude Code, they flood public repositories like GitHub with high-quality, AI-generated outputs. This effectively turns the internet into a massive, unavoidable training dataset for competing models, making it difficult to police "distillation" as a violation of terms.
The primary bottleneck for creating powerful foundation models in biology is the lack of clean, large-scale experimental data—orders of magnitude less than what's available for LLMs. This creates a major opportunity for "data foundries" that use robotic labs to generate high-quality biological data at scale.
The argument that AI will eliminate software jobs by making coding easy overlooks a key reality: most existing software is buggy and frustrating. The demand for better, more reliable products is practically infinite, suggesting AI will augment developer productivity to meet this demand rather than replace developers wholesale.
For hard tech startups, the decision to vertically integrate and build a factory shouldn't be automatic. It's a strategic imperative only when "cadence"—the speed of iteration and delivery—is the primary competitive advantage. In such cases, the in-house capability to move fast outweighs the high capital cost.
The "Citrini" essay caused a market sell-off not because it was more technically sound than other AI analyses, but because it framed abstract AI risk in the concrete language of finance (SaaS multiples, credit risk), making it resonate powerfully with a Wall Street audience.
Unlike the Y2K bug or the 2012 apocalypse, which were largely fringe concerns, the idea that AI could end humanity is held by over 30% of Americans. This marks a significant shift in public consciousness, where technological anxiety has moved from niche communities to a widespread societal concern.
The primary threat AI agents pose to platforms like DoorDash or Uber isn't that they can "vibe-code" a replacement app. It's that they can eliminate the friction of price shopping, thereby commoditizing the demand side of the marketplace and destroying the customer lock-in that constitutes the company's core value.
Despite technologies like Zillow seemingly making them obsolete, real estate brokers have remained resilient due to market inertia and regulatory capture. This serves as a powerful counter-example to predictions of rapid, friction-less AI-driven job displacement in other white-collar professions.
Many dot-com era predictions, like the demise of physical retail, were directionally correct. The primary forecasting error was "timeline compression"—assuming a multi-decade societal transformation would happen in just a few years. This serves as a cautionary tale for the current AI boom, where the "when" is as important as the "what."
The "Battle of Seattle" protests during the dot-com boom raised political awareness and subtly shaped trade policy for years. Similarly, today's local protests against AI data centers, while smaller, introduce political friction that can act as a significant, often underestimated, brake on the speed of technological infrastructure deployment.
For asset-heavy hard tech companies, debt is most effective not as a bridge to the next equity round, but to finance long-lived assets (e.g., machinery) that are directly tied to contracted revenue. This approach de-risks the loan and supports scalable growth without excessive equity dilution, a sharp contrast to SaaS venture debt norms.
A significant disconnect exists between AI's market valuation, which prices in massive future GDP growth, and its current real-world economic impact. An NBER study shows 80% of US firms report no productivity gains from AI, highlighting that market hype is far ahead of actual economic integration and value creation.