Bill Maris: How Google Could Crush AI Competitors, Why Small Funds Win, and AI's Atari Stage

Bill Maris opens with his origin story, founding a web hosting company in 1997 out of his Vermont apartment using credit cards, with servers sharing space with his living quarters. He uses this experience — including tarring a roof during a thunderstorm — to illustrate his first two lessons: glimpsing the future through a 'keyhole' and the necessity of seeming a little insane to act on unconventional insights.

His third lesson comes from building Google Ventures, where he and co-founder Rich Miner used machine learning (Google forbade the term 'AI' at the time) to design optimal portfolio construction and fund sizing. He claims GV achieved approximately 4.1x returns overall, with his personally led investments performing significantly higher, validating his thesis that data science applied correctly to the right problem yields right answers.

His fourth and central thesis is that small funds structurally outperform large ones — not as opinion but as mathematical reality. He cites data showing funds under $750 million average 4.76x DPI returns versus 2x for funds over $1.242 billion, with sub-$750M funds representing 95% of top decile performers. He walks through simple math showing that a $7 billion fund would need $210 billion in exits to return 3x — exceeding total venture-backed M&A and IPO exit value in most years.

On AI, Maris argues we are currently at the 'Atari command line stage,' analogizing today's AI to the text-based game Zork versus modern photorealistic gaming. He predicts the transition will happen in roughly five years and says he is not interested in investing in larger models, but rather in the infrastructure — physics engines, controllers, GPUs — that will enable that transition, similar to how better controllers and graphics cards transformed gaming rather than better storylines.

Maris makes a pointed argument about Google's ability to crush OpenAI and Anthropic by cutting token prices by 80%, using their revenue-generating businesses to subsidize AI as a competitive weapon. He calls this 100% probable and draws a parallel to Uber's capital-as-weapon market share strategy.

He raises a sharp critique of late-stage private companies staying private longer, arguing this forces overpriced equity onto retail investors through passive ETFs and 401k plans while keeping early value creation for elite insiders — and that companies claiming to be building for 'humanity's benefit' while doing so are being hypocritical.

On life sciences, Maris expresses enthusiasm for computational biology and longevity but cautions that human clinical trial requirements mean the sector won't accelerate as exponentially as hoped until in silico cell simulation becomes realistic. He also expresses concern about the US losing scientific talent due to H1B visa restrictions, gutting of NIH/CDC funding, and an 'anti-science vibe,' while China is actively recruiting top global scientists.

The conversation closes with a discussion on venture fund incentive misalignment: large fund GPs can make more money returning 1.01x on $5 billion than small fund GPs returning 3x on $500 million, and entrepreneurs are incentivized to take inflated valuations from large funds — all of which Maris argues will eventually cause the pendulum to swing back toward smaller, more focused venture funds.