Historical tech cycles show that 95-99% of companies fail. For most current AI startups, the next 12-18 months represent a value-maximizing moment to sell before their technology is commoditized or outcompeted by foundation models.
Pay attention when a major tech company abandons a project or market. These strategic retreats, like Google shutting down its Maven defense project, create a vacuum and signal a prime opportunity for a startup (like Anduril) to enter and capture the market.
Tech giants like Meta aggressively bidding on AI talent has created a wealth event for 50-200 top researchers, similar to a collective IPO. This enriches them as a class, not just as employees of a single company, altering their career trajectories and focus.
The most defensible AI companies don't just have superior models; they embed themselves deeply into customer workflows. The primary barrier to adoption is change management, so overcoming that hurdle creates a durable competitive advantage that is difficult to displace.
While venture capital often praises contrarian thinking, during moments of fundamental technological shift like the current AI boom, the most rational strategy is to be consensus. The market is so open and growing so fast that betting on the obvious winners is the right move.
The "great product wins" narrative often omits the aggressive distribution tactics that scaled today's tech giants. Google spent hundreds of millions bundling its toolbar, and Facebook bought ads against users' names—proving that distribution is as critical as product.
A two-year constraint on high-bandwidth memory (HBM) prevents any single AI lab from buying enough compute to pull significantly ahead. This enforces a temporary parity among giants like OpenAI, Google, and Anthropic, creating a short-term oligopoly.
The AI explosion wasn't just due to better models like GPT-3, but the shift to a simple, generalizable API. This eliminated the need for complex, in-house ML Ops teams, allowing any developer to access vast knowledge and reasoning with just a few lines of code.
The fundamental shift with generative AI in B2B is selling "work product" or "human labor equivalents," not just software tools. This reframes the value proposition and opens up historically difficult markets, like law firms, that were resistant to buying traditional SaaS products.
Founders should view board members as long-term relationships akin to in-laws, since they're difficult to remove once appointed. Prioritize a high-quality, helpful board member you can work with for a decade over a slightly better valuation from a less suitable partner.
Despite the rise of remote work, physical location is more critical than ever in hyper-competitive fields. For AI, the San Francisco Bay Area is the undisputed global hub, concentrating 91% of all private AI market capitalization, making it a mandatory presence for serious players.
To break a decades-long stalemate with Pepsi, a Coca-Cola CEO reframed their market from "share of soda" to "share of all liquids." This shifted their market share from 50% to 0.5%, unlocking new growth avenues like bottled water (Dasani) and ultimately dominating the beverage industry.
While diligence is extensive, the decision to make a late-stage investment ultimately hinges on a single core question or belief about a company's unique advantage. If you need to believe more than one or two things for it to be a 10x outcome, it's too complicated and likely won't work.