The race for dominant large language models is over. OpenAI, Anthropic, Google, Meta, and potentially X are the winners. Their massive, ongoing spend on compute (up to $100B/year) creates an order-of-magnitude advantage that new entrants, even with billions in funding, cannot overcome.
While NVIDIA currently holds a stranglehold on AI compute, this dominance won't sustain. The industry will move towards specialization, with new architectures and ASICs designed for specific tasks like inference (e.g., Cerebras) or with neural network weights baked in. This will fragment the market.
Success creates a "reinforcement learning" loop, codifying a firm's methods. When a paradigm shifts, like the move to AI, this reinforced playbook becomes a liability. The more successful a firm was in the prior era, the harder it is for them to adapt to new, foundational business assumptions.
For senior investors, past success creates a comfort zone that is hard to break. To stay relevant with young founders and new technologies, they must be willing to tear down their existing knowledge base and approach conversations as equals, a process that can feel deeply uncomfortable but is essential for growth.
Traditional brand-building tactics like sponsoring conferences are "off-key" to young entrepreneurs. A modern VC brand is built organically through word-of-mouth, based on providing valuable inside knowledge, connections, and being genuinely helpful, not through loud marketing efforts.
Algorithmic improvements alone are not enough for a new AI lab to challenge incumbents, who are also researching next-gen architectures. The only viable path is to focus on domains where proprietary data can be generated and is unavailable to the big labs, such as robotics or specialized life sciences.
As AI lowers software creation costs, the high-margin "product" business is splitting. Companies will either be low-cost providers or offer customized solutions via forward-deployed engineers. This "professional services" model, once a red flag for VCs, is now a viable, high-value strategy.
In AI, companies can reach massive valuations quickly and still offer venture-like returns (e.g., 10x+). This makes traditional stage definitions (early, growth) irrelevant. Investors should ignore stage and focus on the magnitude of the opportunity, whether it's two founders or a $60B company.
Venture's long feedback loops mean a firm's current fundraising success is based on prior generations' work. The most common reason firms fail to transition is when legacy partners, who are no longer active investors, take a disproportionate share of the economics from new funds.
Venture capital has become a scaled, specialized business with large teams. The future, however, belongs to compact firms of well-rounded individuals who can source, exercise judgment, sell, and help companies. Over-specialization where one person sources and another helps is an inefficient model.
Every VC firm claims to help with recruiting and marketing. The real, defensible value proposition is being the go-to person who has an answer when a founder gets stuck with a unique, complex problem, from co-founder disputes to financing strategy. This value can only be conveyed through references.