Early-stage startups, unburdened by legacy systems, adopt the most efficient new technologies. Observing their infrastructure choices (e.g., which semiconductors they use) provides a powerful, early signal for identifying dominant players in the public markets long before mainstream recognition.
The next leap in AI's value will come from "agents" that work on problems autonomously without direct, real-time user commands. This shift from a reactive, search-engine-like model to a proactive, problem-solving one will drive a 5x increase in compute consumption and unlock new applications.
Unlike mature industries, technology's winner-take-all dynamic stems from a virtuous cycle. The market leader generates more capital, which it reinvests into R&D to widen its technological lead, making it increasingly difficult for competitors to catch up. This compounding advantage is key.
Fears of an AI bubble fueled by over-investment are currently misplaced. Real-world bottlenecks in the supply chain, from specialized memory to semiconductor manufacturing capacity (TSMC), are constraining the buildout. This forces a more measured pace, keeping demand well ahead of supply for the foreseeable future.
Despite a significant head start via its OpenAI partnership, Microsoft initially "fumbled" its advantage. The poor performance of its early AI integrations created a market opening that competitors like Anthropic successfully exploited. Microsoft is now playing catch-up in a more competitive landscape.
To find investment opportunities, don't try to cover an entire industry. Instead, concentrate capital and research where change is happening most quickly and dramatically. This is where incumbents are most vulnerable and new value is created, as disruption equals opportunity.
Beyond financial analysis, a key principle from Julian Robertson's Tiger Management was an unwavering focus on the integrity of leadership. Any question about a CEO's or CFO's integrity was an immediate deal-breaker, regardless of the perceived opportunity.
The idea that GPUs are the ideal architecture for AI is not new. Bill Joy, a Sun Microsystems founder, identified engineers using GPUs for AI calculations back in the mid-2000s. This long-gestating insight highlights that major technological shifts are often recognized by experts decades before they become mainstream.
A common mistake is applying new technology to solve old, already-automated problems. True value comes from using new tech, like AI, to tackle entirely new challenges. Incumbent systems that work well, even if based on older technology like mainframes, are rarely swapped out, creating opportunities when the market overestimates their demise.