Jack Altman & Martin Casado on the Future of VC
Martin Casado, General Partner at a16z, discusses the evolution of venture capital in the age of AI, covering the firm's shift to specialization, the critical value of AI infrastructure, the fierce competition for talent, and the strategic importance of media and open source.
The Evolving Role of Media and Distribution in Venture Capital
In the past, top-tier venture capitalists often maintained a low public profile. However, the landscape has shifted significantly. A primary driver for this change is that traditional media has become increasingly hostile towards the tech industry, making it “very dangerous” for founders to engage with reporters. To counteract this, VCs must now “go direct” by building their own media platforms. This allows them to control the narrative and provide a channel for their portfolio companies to get their message out.
The nature of content consumption has also become highly “episodic,” revolving around major events and launches. A firm with an in-house media capability can help its portfolio companies navigate this environment and draft on the “current zeitgeist” to gain a voice. Ultimately, the goal of a VC’s media presence isn’t personal branding; it’s to create a distribution channel that helps founders solve the “bootstrap problem of of zeitgeist understanding and brand.” While a VC firm can be an accelerant, the companies with the best brands are almost always built by the founders themselves.
The Transformation of a16z: From Generalists to Specialists
a16z’s structure has evolved dramatically, moving from a small group of generalist partners to a scaled-up firm with specialized platforms. This transformation was a necessary response to the maturation and expansion of the technology market. When tech was a niche market, a generalist approach made sense. Today, markets are so vast that a single individual can build an entire career investing solely in a sub-field like databases.
This market growth necessitates specialization to remain competitive. As firms grow their assets under management (AUM) to cover more stages (seed, venture, growth), the traditional, consensus-driven partnership model of venture capital fails to scale. Specialization allows a firm to structure its approach, ensure comprehensive market coverage, and recruit top-tier GPs by offering them significant autonomy over their specific domains.
Investing in AI Infrastructure
Martin Casado defines infrastructure as the foundational technology sold to technical buyers—developers, database administrators, etc.—who use it to build applications. This includes compute, networking, storage, databases, developer tools, and now, AI models.
His core thesis is that infrastructure is where the most durable value and differentiation in software are created.
“In software the true differentiation is is technical… that almost always comes from the actual infrastructure that the software is built on.”
While there are fewer infrastructure companies than application companies, they tend to command higher multiples because they are the source of value for everything built on top of them. Each major platform shift—cloud, mobile, and now AI—ushers in a new set of essential infrastructure companies.
When evaluating infrastructure investments, the threat of large incumbents like AWS or OpenAI entering the market is a common concern. However, it’s very difficult for large companies to execute with the focus and agility of a startup. If a market is viable and expanding, a focused independent company can fill that expanding space and succeed.
Market Dynamics: Conflicts and the War for Talent
The rapid expansion of the AI market has created unique challenges, particularly around portfolio conflicts. The most common conflict arises when one portfolio company pivots into another’s space. A more difficult, modern conflict emerges when an established, non-“AI-native” company wants to adopt AI, putting it in competition with a new, AI-native startup in the same domain.
However, the most intense competition in AI today is not for market share, but for talent.
“The market is so big and it’s growing so fast. Even companies that seem like they’re competing end up in totally different places… But they’re all competing like totally different companies faces are competing with the same talent. This is the first time I can remember where the actual talent competition is like way more fierce.”
This is especially true for roles requiring experience in training large-scale models, a skill set possessed by a very small number of teams globally. This scarcity of experienced talent is a primary driver behind the large “aqua-hire” acquisitions seen in the industry.
The State of Play in AI: What’s Working
When assessing the current AI landscape, different use cases show varying levels of maturity and economic viability:
- Clearly Working: Diffusion models and content creation (images, music, speech). The economics are simple and powerful: these models reduce the marginal cost of content creation by several orders of magnitude, making them immediately valuable.
- Working, but Fragmented: Companionship, emotional support, and entertainment. These use cases have proven user engagement and willingness to pay, but from an investment standpoint, the market tends to be long-tailed and fragmented.
- Working, but Economics Unclear: Enterprise automation and “agentic” workflows. While companies are successfully implementing these solutions, they often require significant “bespoke work,” making the economic model more complex and less straightforward than pure content creation. The distinction is between a model creating content (which works well) and a model automating a human task (which still needs a lot of work).
The Future of Code and the Importance of Open Source
AI is fundamentally changing the discipline of software engineering. While current AI coding assistants are often described as “dazzling” but not yet consistently useful, they excel at specific tasks like writing documentation, handling boilerplate code, and navigating complex frameworks. We are in the early stages of developing best practices, but a 10x improvement in developer productivity is likely on the horizon.
Open source is a critical component of a healthy AI ecosystem. It prevents monopolies, fosters innovation, and ensures broad access to powerful technology. A year ago, there was a concerning trend of VCs, academics, and founders—traditional champions of open source—arguing that it was dangerous in the context of AI. This “lopsided debate” was largely influenced by a pre-existing intellectual legacy around AI risk. The conversation has since become more balanced, with the right voices re-entering the room to advocate for the importance of an open ecosystem.