Bill Maris says Google could destroy OpenAI and Anthropic overnight by cutting token prices 80% — and argues it's probably already inevitable.
All-In with Chamath, Jason, Sacks & Friedberg
Bill Maris says Google could destroy OpenAI and Anthropic overnight by cutting token prices 80% — and argues it's probably already inevitable.
TL;DR
Bill Maris, founder of Google Ventures and Section32, delivers four hard-won lessons from a career spanning data-center entrepreneurship, AI-powered portfolio construction, and fund management. He argues that small VC funds (under $750M) dramatically outperform large ones — averaging 4.76x DPI versus 2.42x [1] — Bill Maris "$7B fund needs $210B in exits to 3x: A $7B fund needing a 3x return requires $210B in exits — more than total venture-backed M&A and IPO ex…" 24:27 — and that Google could crush OpenAI and Anthropic simply by slashing token prices 80% [2] — Bill Maris "The math is unambiguous: VC funds under $750M average 4.76x DPI versus 2.42x for funds over $1B, and represent 95% of all top-decile perfor…" 21:40 . His clearest takeaway: AI is currently at its "Atari stage," and the real investment opportunity lies in the infrastructure layer — physics engines, controllers, and GPUs — not bigger models [3] — Bill Maris "Today's AI is at the Atari command-line stage: brittle, turn-based, lacking memory and consistency — just like Zork in the 1980s. The leap …" 18:50 .
Bill Maris, founder of Google Ventures and Section32, shares four career lessons: glimpsing the future early, the insanity required for vision, the power of data-driven investing, and why small VC funds outperform large ones. He also discusses Google's pricing power over OpenAI, AI's 'Atari stage,' and the US scientific brain drain.
The episode opens with a quick profile of Bill Maris — former founding CEO of Google Ventures, now running the smaller and deliberately lean Section32 fund. The stage is set for a conversation about his four career lessons, with sponsor reads from EY, NYSE, and Plaud bracketing the open.
Maris opens his keynote with a story that is equal parts comedy and prophecy. Fresh out of college with a neuroscience degree, he ended up on Wall Street — miserable in a suit — until the day he found a server hiding in the office closet. It was a revelation: if a company's email and website could live in a closet, how many websites could live in his? He quit the next day. What followed was a quintessentially scrappy origin story: three servers (small, medium, and large), a Vermont apartment doubling as a data center, a Home Depot rug for a bed, and glasses of water that would ice over by noon. The setup was absurd, but Maris had glimpsed something real through that keyhole. [1] — Bill Maris "Bill Maris quit his Wall Street job in 1997 after spotting a server in an office closet and realizing he'd seen the future of the internet.…" 00:35
The roof of Maris's Vermont apartment-cum-data-center started leaking during a thunderstorm, and with water and servers a lethal combination, he made the only rational (if irrational-seeming) decision: climb up with a mop and a bucket of tar. He tarred himself into a corner while lightning flashed around him, shoes forever stuck to the roof. It's a story he tells with genuine self-deprecating warmth, but the payoff is pointed. Using a series of inauguration photos — from 1989 to 2005 to 2009 — he illustrates how the person in the crowd with a laptop livestreaming, when everyone else just had a camera, is always the one who 'knows a secret about the future that most of us don't believe.' That's what he looks for in founders. [1] — Bill Maris "When the roof of his apartment-cum-data-center leaked during a thunderstorm, Maris climbed up with a bucket of tar — and tarred himself int…" 05:58 [2] — Bill Maris "At a 2009 inauguration where everyone else had a camera, one person livestreamed on a laptop. Maris uses this image to define what he looks…" 08:10
In 2007, Google handed Maris a challenge: build a coherent venture fund from scratch. He partnered with Rich Miner, co-founder of Android, trekked up and down Sand Hill Road, and devised a plan to collect every piece of historical venture data they could find. Then they applied machine learning — Google wouldn't allow the word 'AI,' insisting it was '100 years away' and would 'freak people out' — to design the ideal portfolio construction through millions of simulations. [1] — Bill Maris "When tasked with creating Google Ventures in 2007, Maris gathered every piece of venture data he could find and ran millions of portfolio s…" 09:30 Executives thought the plan was crazy. But between 2009 and 2018, Google Ventures is estimated to have returned approximately 4.1x, with the investments Maris personally led performing even better. His third lesson distils into one rule: don't bet against computer science applied to the right problem at the right time.
When Maris left Google in 2017 to start his own fund, the conventional wisdom was unanimous: raise as much as possible. The management fees alone would be worth it. He ignored that advice entirely. Section32's six funds have averaged approximately $400M in size, and every single one has landed in the top decile. The underlying math explains why: funds under $750M average 4.76x DPI, funds over $1B average 2.42x, and sub-$750M funds represent 95% of all top-decile performers. [1] — Bill Maris "The math is unambiguous: VC funds under $750M average 4.76x DPI versus 2.42x for funds over $1B, and represent 95% of all top-decile perfor…" 21:40 The logic is simple and brutal — a $7B fund targeting 3x returns needs $210B in exits, which exceeds total venture-backed M&A and IPO value in most years. Size is the enemy of returns, not the friend.
The sharpest moment in the conversation arrives when Maris raises what he considers Google's obvious strategic play: slash token prices by 80% and watch OpenAI and Anthropic's business models go 'super critical.' If an enterprise can get a basically identical product from Gemini at a fifth of the price, why wouldn't they switch? The panel explores whether OpenAI is burning investor cash Uber-style to grab market share, and whether the public markets will eventually be left holding the bag. Maris is blunt: with approximately $1 trillion in spend commitments and only $60 billion in revenue, OpenAI's IPO math relies on retail and passive funds absorbing valuations that the S&P 500 rules are already being bent to accommodate. [1] — Bill Maris "Google could arbitrarily cut token prices by 80%, making its model functionally identical to OpenAI's at a fraction of the cost — and Maris…" 24:50
The conversation turns darker when Friedberg raises the question of whether China and other countries are becoming more attractive destinations for scientific investment and talent. Maris doesn't hedge: the gutting of the CDC and NIH, combined with H-1B visa restrictions and what he calls an 'anti-science vibe,' has made it easier for scientists to just go elsewhere. [1] — Bill Maris "The gutting of the CDC and NIH, H-1B restrictions, and a pervasive anti-science cultural shift in the US are driving scientific talent to C…" 23:00 Friedberg adds that China has its own 'paperclip model' — aggressively recruiting top scientists from Europe and India, talent pools that used to flow to the United States. Maris is not triumphalist about this; he simply notes it's damaging, and calls for the 'neurological reserves' the US needs to maintain scientific leadership.
Chapter 2 · 00:33
Maris opens his keynote with a story that is equal parts comedy and prophecy. Fresh out of college with a neuroscience degree, he ended up on Wall Street — miserable in a suit — until the day he found a server hiding in the office closet. It was a revelation: if a company's email and website could live in a closet, how many websites could live in his? He quit the next day. What followed was a quintessentially scrappy origin story: three servers (small, medium, and large), a Vermont apartment doubling as a data center, a Home Depot rug for a bed, and glasses of water that would ice over by noon. The setup was absurd, but Maris had glimpsed something real through that keyhole. [1] — Bill Maris "Bill Maris quit his Wall Street job in 1997 after spotting a server in an office closet and realizing he'd seen the future of the internet.…" 00:35
Bill Maris quit his Wall Street job in 1997 after spotting a server in an office closet and realizing he'd seen the future of the internet. He built a data center startup from his Vermont apartment — on credit cards, with servers in one room and a Home Depot rug for a bed in the other.
Chapter 3 · 05:58
The roof of Maris's Vermont apartment-cum-data-center started leaking during a thunderstorm, and with water and servers a lethal combination, he made the only rational (if irrational-seeming) decision: climb up with a mop and a bucket of tar. He tarred himself into a corner while lightning flashed around him, shoes forever stuck to the roof. It's a story he tells with genuine self-deprecating warmth, but the payoff is pointed. Using a series of inauguration photos — from 1989 to 2005 to 2009 — he illustrates how the person in the crowd with a laptop livestreaming, when everyone else just had a camera, is always the one who 'knows a secret about the future that most of us don't believe.' That's what he looks for in founders. [1] — Bill Maris "When the roof of his apartment-cum-data-center leaked during a thunderstorm, Maris climbed up with a bucket of tar — and tarred himself int…" 05:58 [2] — Bill Maris "At a 2009 inauguration where everyone else had a camera, one person livestreamed on a laptop. Maris uses this image to define what he looks…" 08:10
When the roof of his apartment-cum-data-center leaked during a thunderstorm, Maris climbed up with a bucket of tar — and tarred himself into a corner while lightning struck nearby. The lesson: visionary entrepreneurs always look unhinged to the people around them.
At a 2009 inauguration where everyone else had a camera, one person livestreamed on a laptop. Maris uses this image to define what he looks for in founders: they know a secret about the future that most people don't believe yet.
When tasked with creating Google Ventures in 2007, Maris gathered every piece of venture data he could find and ran millions of portfolio simulations using machine learning — though Google refused to let him call it 'AI.' The result was a disciplined, data-driven fund strategy that Google executives thought was crazy.
Chapter 4 · 09:51
In 2007, Google handed Maris a challenge: build a coherent venture fund from scratch. He partnered with Rich Miner, co-founder of Android, trekked up and down Sand Hill Road, and devised a plan to collect every piece of historical venture data they could find. Then they applied machine learning — Google wouldn't allow the word 'AI,' insisting it was '100 years away' and would 'freak people out' — to design the ideal portfolio construction through millions of simulations. [1] — Bill Maris "When tasked with creating Google Ventures in 2007, Maris gathered every piece of venture data he could find and ran millions of portfolio s…" 09:30 Executives thought the plan was crazy. But between 2009 and 2018, Google Ventures is estimated to have returned approximately 4.1x, with the investments Maris personally led performing even better. His third lesson distils into one rule: don't bet against computer science applied to the right problem at the right time.
Bill Maris estimated Google Ventures' returns at approximately 4.1x using publicly available data over the 2009–2018 period.
Google Ventures' investment in Climate Corp — David Friedberg's company — resulted in a $1 billion exit to Monsanto.
Chapter 5 · 14:36
When Maris left Google in 2017 to start his own fund, the conventional wisdom was unanimous: raise as much as possible. The management fees alone would be worth it. He ignored that advice entirely. Section32's six funds have averaged approximately $400M in size, and every single one has landed in the top decile. The underlying math explains why: funds under $750M average 4.76x DPI, funds over $1B average 2.42x, and sub-$750M funds represent 95% of all top-decile performers. [1] — Bill Maris "The math is unambiguous: VC funds under $750M average 4.76x DPI versus 2.42x for funds over $1B, and represent 95% of all top-decile perfor…" 21:40 The logic is simple and brutal — a $7B fund targeting 3x returns needs $210B in exits, which exceeds total venture-backed M&A and IPO value in most years. Size is the enemy of returns, not the friend.
Today's AI is at the Atari command-line stage: brittle, turn-based, lacking memory and consistency — just like Zork in the 1980s. The leap to a PlayStation 10-level AI will come in 5 years, driven not by bigger models but by the same kind of infrastructure that transformed gaming: physics engines, controllers, and GPUs.
Chapter 6 · 19:09
The sharpest moment in the conversation arrives when Maris raises what he considers Google's obvious strategic play: slash token prices by 80% and watch OpenAI and Anthropic's business models go 'super critical.' If an enterprise can get a basically identical product from Gemini at a fifth of the price, why wouldn't they switch? The panel explores whether OpenAI is burning investor cash Uber-style to grab market share, and whether the public markets will eventually be left holding the bag. Maris is blunt: with approximately $1 trillion in spend commitments and only $60 billion in revenue, OpenAI's IPO math relies on retail and passive funds absorbing valuations that the S&P 500 rules are already being bent to accommodate. [1] — Bill Maris "Google could arbitrarily cut token prices by 80%, making its model functionally identical to OpenAI's at a fraction of the cost — and Maris…" 24:50
The transformation gaming underwent from text-based Zork to photorealistic modern games will happen to AI within the next 5 years, per Maris.
Bill Maris compared today's AI to the Atari command-line era and predicted it would reach PlayStation-equivalent sophistication within 5 years.
Maris founded Calico, invested in Flatiron and New Limit, and remains deeply interested in longevity — but says therapeutic clinical trials are a specialist area he's stepping back from. The real prize is computational biology: simulate a human cell in silico and drug discovery accelerates exponentially.
The math is unambiguous: VC funds under $750M average 4.76x DPI versus 2.42x for funds over $1B, and represent 95% of all top-decile performers. Maris's six Section32 funds have averaged ~$400M in size and all landed in the top decile.
Bill Maris's Section32 has raised 6 funds averaging ~$400M in size, with all 6 performing in the top decile.
Funds smaller than $750M averaged 4.76x DPI returns versus 2.42x for funds over $1B, according to Bill Maris's analysis.
Funds below $750M represented 95% of top decile venture performers, with discontinuous return compression above that threshold.
The gutting of the CDC and NIH, H-1B restrictions, and a pervasive anti-science cultural shift in the US are driving scientific talent to China and elsewhere. China is now actively recruiting top scientists from Europe and India — a talent pool that previously flowed to America.
Bill Maris reminded the audience that the 75th percentile of venture funds loses money, making top-quartile persistence critical.
Deep tech — historically slow, capital-intensive, and risky — is becoming more tractable because AI and physics engines are compressing development timelines. Maris sees healthcare and the full infrastructure stack of the AI revolution as the two primary deep tech opportunity sets.
A $7B fund needing a 3x return requires $210B in exits — more than total venture-backed M&A and IPO exit value in most years.
Google could arbitrarily cut token prices by 80%, making its model functionally identical to OpenAI's at a fraction of the cost — and Maris says this isn't speculation, it's the rational play. The compression on OpenAI and Anthropic's business models would go 'super critical.'
Bill Maris argued Google could unilaterally cut token costs by 80%, putting OpenAI and Anthropic under existential pricing pressure.
Bill Maris identified human biology and healthcare as probably the largest total addressable market in the world for deep tech investment.
Chapter 7 · 25:23
The conversation turns darker when Friedberg raises the question of whether China and other countries are becoming more attractive destinations for scientific investment and talent. Maris doesn't hedge: the gutting of the CDC and NIH, combined with H-1B visa restrictions and what he calls an 'anti-science vibe,' has made it easier for scientists to just go elsewhere. [1] — Bill Maris "The gutting of the CDC and NIH, H-1B restrictions, and a pervasive anti-science cultural shift in the US are driving scientific talent to C…" 23:00 Friedberg adds that China has its own 'paperclip model' — aggressively recruiting top scientists from Europe and India, talent pools that used to flow to the United States. Maris is not triumphalist about this; he simply notes it's damaging, and calls for the 'neurological reserves' the US needs to maintain scientific leadership.
OpenAI is heading toward a public market with roughly $1 trillion in spend commitments against only $60 billion in revenue. Maris is skeptical retail investors and passive funds can absorb these valuations — and notes the S&P 500 rules are already being bent to accommodate them.
Bill Maris highlighted the mismatch between OpenAI's ~$1 trillion in spend commitments and only $60 billion in revenue, questioning the IPO math.
VC incentives are broken at every level: LPs write big checks to big funds to avoid blame, GPs earn more in management fees at 1.01x on $5B than at 3x on $500M, and naive founders take inflated $4B valuations over disciplined $100M ones. The pendulum will swing back — but not yet.
A $5B fund returning just 1.01x still lands in the 75th percentile of VC, illustrating the broken incentive structure of large fund management.
The fee structure means a GP managing a $5B fund at 1.01x makes more money than one running a $500M fund at a 3x return.
No indexed bits in this chapter.
Google could arbitrarily cut token prices by 80%, making its model functionally identical to OpenAI's at a fraction of the cost — and Maris says this isn't speculation, it's the rational play. The compression on OpenAI and Anthropic's business models would go 'super critical.'
The math is unambiguous: VC funds under $750M average 4.76x DPI versus 2.42x for funds over $1B, and represent 95% of all top-decile performers. Maris's six Section32 funds have averaged ~$400M in size and all landed in the top decile.
Today's AI is at the Atari command-line stage: brittle, turn-based, lacking memory and consistency — just like Zork in the 1980s. The leap to a PlayStation 10-level AI will come in 5 years, driven not by bigger models but by the same kind of infrastructure that transformed gaming: physics engines, controllers, and GPUs.
This episode
Factual claims made this episode, and whether a source was named.
VC funds smaller than $750M average 4.76x DPI returns, while funds larger than $1B average only 2.42x.
Funds below $750M represent 95% of all top decile VC performers.
75% of venture capital funds lose money.
A $7B VC fund targeting 3x returns would need $210B in exits — exceeding total venture-backed M&A and IPO exit value in most years.
Google Ventures' returns between 2009 and 2018 are estimated at approximately 4.1x based on publicly available data.
Section32 has run 6 funds averaging approximately $400M in size, with all 6 performing in the top decile.
If Google cut token prices by 80%, companies would switch from OpenAI to Gemini for a basically identical product at a fraction of the cost.
OpenAI has approximately $1 trillion in spend commitments against only $60 billion in revenue.
The transformation of the gaming industry from text-based Zork in the 1980s to today's photorealistic games will be replicated in AI within the next 5 years.
A GP managing a $5B fund returning 1.01x earns more money than one managing a $500M fund returning 3x, due to management fees.
A $5B venture fund returning 1.01x still qualifies as a 75th percentile fund and can raise its next fund successfully.
China is actively recruiting top scientists from Europe and India who previously would have come to the United States.
Finding a drug compound represents only about 5% of the work in bringing a therapeutic to market, with titrating and safety testing dominating the rest.
This episode
Discussed as the entity with the pricing power to crush AI competitors by slashing token costs 80%, and as Maris's former employer.
The venture fund Bill Maris founded at Google in 2009, which he used machine learning to design and which he estimates returned ~4.1x.
Discussed as vulnerable to Google's pricing power and questioned for its valuation with ~$1T in spend commitments against $60B revenue.
Bill Maris's current venture fund, with 6 funds averaging ~$400M in size, all performing in the top decile.
Cited alongside OpenAI as vulnerable to Google's pricing power in an AI token price war.
A longevity-focused biotech company co-founded by Bill Maris inside Google, and a marker of his interest in life science.
David Friedberg's company that Google Ventures invested in, resulting in a $1 billion exit to Monsanto.
Cited by David Sacks as a benchmark large VC firm that smaller funds shouldn't aspire to emulate.
Cited as an example of a fund poised to generate enormous paper returns from companies like SpaceX, illustrating the liquidity challenge of large venture positions.
The National Institutes of Health, cited by Maris as being gutted in a way that is driving US scientific brain drain.
Cited as an example of a deep tech company facing the public market test for its valuation.
The Centers for Disease Control, cited alongside NIH as an institution being gutted in ways harmful to US science.
An AI company and Section32 portfolio investment cited as a notable fund holding.
One of Section32's notable portfolio companies cited as evidence of top-decile fund performance.
One of Section32's notable portfolio companies cited as evidence of the fund's top-decile performance.
An oncology data company cited as a past Section32 life science investment.
A longevity biotech company backed by Blake Byers and Brian Armstrong in which Section32 is an investor.
Autonomous vehicle company incubated by Bill Maris during his time as Google's VP of Special Projects.
Google's AI model, cited as the vehicle through which Google could cut token prices 80% relative to OpenAI.
Rich Miner, the co-founder of Android, became Bill Maris's first partner in building Google Ventures.
Stats
We use essential and analytics cookies to run Vuci. To understand how the site is used: Privacy Policy.
Install Vuci on your phone
Add it to your home screen for a faster, app-like experience.
Install Vuci on your phone
Tap the Share button, then “Add to Home Screen”.
A new version is available
Reload to get the latest Vuci.