The most defensible long-term business for foundation model labs is not their commoditizing API but the end-user applications they will inevitably build on top of their own tech.
Deep AI engineering expertise is a more scarce and critical determinant of success for AI application companies than deep domain expertise.
The venture capital industry is structurally bifurcating into small, specialist firms and large, multi-stage platforms, rendering mid-sized funds increasingly uncompetitive.
AI agents will create companies of unprecedented scale because they compete for human labor budgets, a much larger market than existing software spending.
The potential exit outcomes for premier technology companies have expanded from tens of billions to the multi-trillion dollar range, fundamentally changing the scale of venture capital investment.
▶The Precarious Economics of Foundation ModelsMar 2026
Bucky argues that despite their technological significance, AI model providers face severe business model challenges. Their API-driven revenue is subject to inevitable and rapid price decay due to competition, while high capital expenditures on compute lead to margin compression and cash burn.
This suggests that the long-term value in the AI stack may not accrue to the model providers themselves, but rather to the application layer or vertically integrated players who can build moats beyond the raw API.
▶Venture Capital's Structural BifurcationMar 2026
Bucky posits that the VC industry is splitting into two successful models: small, specialized firms and large, multi-stage mega-platforms. He believes mid-sized funds will struggle to compete as LP capital concentrates and investment sizes for top companies swell to unprecedented levels, reaching up to a billion dollars.
This trend implies that both founders and LPs will face a more polarized landscape, needing to choose between deep, niche expertise and massive scale and network effects, with less room for generalist, mid-market funds.
▶The Primacy of AI Engineering and AgentsMar 2026
Bucky asserts that deep AI engineering expertise is the most critical and scarce resource for building successful AI application companies, trumping traditional domain expertise. He believes AI agents will create larger companies than SaaS by replacing human labor budgets, not just existing software spend.
This signals a fundamental shift in talent acquisition and market sizing for investors; companies with elite AI engineering teams are better positioned, and the total addressable market for AI applications should be calculated based on labor costs, not just IT budgets.
▶The Inevitable Move to Vertical IntegrationMar 2026
Bucky predicts that all major AI labs will be forced to build their own end-user applications to escape the commoditization of their model APIs. He sees this as their most compelling long-term business, citing OpenAI's implicit move into enterprise search as an example.
This foreshadows a future of intense competition between former partners, where infrastructure providers become direct competitors to the application companies they once supplied, reshaping the entire AI ecosystem.