Marc Andreessen at a16z LP Day: the co-founder's view on where the firm and tech are heading

Summary

Marc Andreessen argues the current AI moment is structurally different from the dot-com era in one key respect: the products work, and users are showing up in numbers the internet didn't support in 1999. He points to seed-stage companies already past $10M ARR as evidence that the underlying businesses are real, not just speculative. History rhymes but doesn't repeat — the more useful lesson from the dot-com era is that the technology kept growing even after the crash, and incumbents who wrote it off in 2001 were wrong by 2005.

Open source and the geopolitics of AI

Andreessen draws a direct line from the Linux era to the open-source AI debate. Just as Linux commoditized the server operating system market — wiping out companies like Sun that were making serious money on proprietary Unix — open-source AI models could do the same to today's frontier labs. He thinks it's entirely plausible that DeepSeek, Llama, or something that comes after them becomes the default global standard.

But he's explicit that the origin of the model matters as much as its openness. Weights encode values — you can't separate them from the training process. DeepSeek literally includes Marxism as a scored benchmark category, hitting 100 out of 100. Andreessen's framing is blunt: AI will intermediate legal systems, education, and healthcare. Whether those systems run on American-trained or Chinese-trained models is not a neutral infrastructure question.

a16z's stated position is to align its foreign policy with U.S. foreign policy. The firm never invested materially in China, and Andreessen treats that as a decision that has aged well — though he acknowledges the US-China relationship is fluid and publicly reported talks are underway.

American dynamism

The firm's American dynamism thesis targets GDP sectors that technology has largely bypassed over the past 50 years: education, healthcare, housing, defense, and law. Andreessen's sharpest example is defense — the Pentagon spends close to a trillion dollars annually, and most of its hardware predates the modern software era. The F-16 is from the 1970s. The U-2 spy plane is still flying. He sees bipartisan political momentum now, with the current administration embracing the thesis alongside Democrats who are receptive to the build-more framing.

How the firm has evolved

a16z raised its first fund in 2009, in the wreckage of the financial crisis. Andreessen describes being advised by a senior VC to treat LPs like mushrooms — keep them in the dark and don't open the box for two years. He and Ben Horowitz took the opposite approach, treating LPs as full partners. Sixteen years in, the firm has moved toward funding companies later in their lifecycle at larger check sizes, a direct response to the IPO market failing to reform. Companies staying private longer was a structural shift, not a temporary one, and a16z adapted rather than waited for it to change.

What Andreessen is still waiting for

On AI-native companies, he says he's starting to see a generation of founders using tools like ChatGPT deep research to operate at capability levels that weren't possible before — synthesizing information and writing code at a pace that used to require large teams. But he's candid that he hasn't yet seen the company he's really waiting for: the one with 1,000 employees where 999 are bots. The real unlock, in his view, isn't bringing AI products to market — it's using AI to fundamentally change how companies themselves operate.

Edge in an AI world

On whether accumulated knowledge still confers an individual edge when any query can surface near-PhD-level information, Andreessen argues depth is still valuable in narrow technical fields — AI foundation model research, biotech — but that for most domains, the advantage shifts to breadth. The best future entrepreneurs will be skilled across six or eight disciplines and able to synthesize across them, using AI to go deep on demand. He points to great tech CEOs as the existing model: they're simultaneously strong on product, sales, marketing, legal, finance, and investor relations. That multi-disciplinary range has always been the entrepreneurial burden, and AI amplifies rather than replaces it.

Brand and cultural moment

The a16z rebrand, led by designer Greg Trudell with close input from Andreessen, is framed as a deliberate break from what he describes as roughly 15 years of cultural and aesthetic contraction — ultra-minimalist design, speech caution, and pervasive negativity about institutions and national identity. He argues that cycle peaked around 2021–22 and has since reversed, replaced by renewed appetite for ambition, scale, and achievement. The new identity, including a figure named Technomedes, is meant to reflect that shift.