ATRIUMsearch → argument graph
Audio · 2026-06-09 · 29m · 11 moments

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

(0:00) Bill Maris joins the Besties! (0:33) Four critical lessons from a career in technology (5:58) Building Google Ventures with data and machine learning (9:51) Why small VC funds beat big ones on average (14:36) OpenAI's valuation problem and the AI price war (19:09) AI's "Atari Stage": what comes next? (25:23) VC's broken incentives and the future of deep tech Thanks to our partners for making this possible! EY - Agentic AI is introducing a new investment discipline. As AI s ✦ AI generated

Apple Podcasts ↗
GUEST Bill Maris

timeline · colored by role

01
Claim◆

Small venture capital funds outperform large funds, as shown by the data.

Bill Maris presents data showing that VC funds under $750M significantly outperform larger funds in DPI, arguing the math is simple and objective.

transcript

Bill Maris: Smaller funds, you can have more focus. I mean, I've already managed A multi-billion dollar fund with hundreds of employees. It's distracting. You cannot give the attention to founders that I would like to give. There are many reasons for this. And if we look at top decile performance of DPI. Funds smaller than 750 million, average return of 4.76x, and funds larger than a billion, 2.42x. Funds below 750 million across that time period represented 95% of top decile performers with discontinuous return compression above 750 million.

02
Claim◆

Small VC funds outperform large funds; funds below $750M averaged 4.76x DPI returns vs. 2.42x for funds over $1B, and 95% of top-decile performers were sub-$750M.

Maris argues that small VC funds dramatically outperform large ones, citing data that sub-$750M funds averaged 4.76x DPI versus 2.42x for billion-dollar funds, and that 95% of top-decile performers have been smaller funds.

transcript

Bill Maris: Smaller funds, you can have more focus. I mean, I've already managed A multi-billion dollar fund with hundreds of employees. It's distracting. You cannot give the attention to founders that I would like to give. There are many reasons for this. And if we look at top decile performance of DPI. Funds smaller than 750 million, average return of 4.76x, and funds larger than a billion, 2.42x. Funds below 750 million across that time period represented 95% of top decile performers with discontinuous return compression above 750 million.

03
Mechanism

A large VC fund's math often doesn't work because the required exit values can exceed the total available market for venture-backed exits.

Maris runs a thought experiment showing that a $7B fund targeting a 3x return would need $210B in exit value, which exceeds total annual venture-backed IPO and M&A value in most years.

transcript

Bill Maris: If you have a $500 million fund, and let's say on average these days you can own 10% of a company, you need $5 billion of exits to get your money back, let's just remind ourselves that the 75th percentile of venture loses money. and there is persistence of performance of the top quartile. So if you need 5 billion to get your money back and if you want to be in this business for the long term, let's say you set your goal at 3x, you need to return $15 billion of exit value in your companies. Now, if you have a $7 billion fund and we do the same math through, you know, you've got to return 210 billion. 7 billion to 70 times 3x is 210 billion, which exceeds the total venture-backed M&A and IPO exit value in most years.

explains mechanism · 2

04
Claim

Companies that stay private for too long, wrap themselves in 'public benefit' language, and then force overpriced shares onto retail investors through 401ks and ETFs are making the public the bag holders while saying they're benefiting humanity.

Maris criticizes the trend of companies staying private longer, raising at enormous valuations, and then forcing ordinary Americans' retirement accounts to buy overpriced shares — all while claiming to be doing it for the benefit of humanity.

transcript

Bill Maris: I have an observation that a bit of an objection to companies that wrap themselves up in public benefit language and then keep the value creation to themselves and an elite group of investors through a big part of the curve and then say, well, we're here to benefit humanity. Well, what humanity needs is money. ... We're going to force overpriced products on the 401k holders of America who didn't get to participate early. This is your position that this is profoundly in there and creates more wealth creation for the people who don't need it. And it makes the people's retire accounts the bag holders. ... My objection is don't say you're doing this for the benefit of humanity and do the other thing. Make the public's retirement accounts the back holders. Or just say, this is how we're running our business and this isn't for the benefit of humanity.

provides context · 1supports · 1

05
Prediction◆

Google could crush AI competitors like OpenAI and Anthropic by arbitrarily cutting token prices, given its massive war chest.

Maris argues that if Google decided to offer tokens at 80% less, it would put critical pressure on OpenAI and Anthropic's business models, and he believes this is clearly what Google will do.

transcript

Bill Maris: However, if I'm Google and I don't speak for Google and I decide to arbitrarily cut the cost of, you know, tokens to 80%. I'm going to cut them in. What happens to the business models of OpenAI and Anthropic at that point? ... Well, if you're a company and you can go to Google and Gemini and you can pay 80% less for that basically identical product, why wouldn't you do that? And then the compression and the pressure on those other businesses goes super critical.

06
Prediction◆

If Google arbitrarily cuts token prices by 80%, it would destroy the business models of OpenAI and Anthropic because customers would switch to a cheaper, essentially identical product.

Maris argues that Google could use its cash war chest to undercut AI competitors like OpenAI and Anthropic on pricing, putting them under existential pressure — and that this is the rational thing for Google to do.

transcript

Bill Maris: If I'm Google and I don't speak for Google and I decide to arbitrarily cut the cost of, you know, tokens to 80%. I'm going to cut them in. What happens to the business models of OpenAI and Anthropic at that point? ... If you're a company and you can go to Google and Gemini and you can pay 80% less for that basically identical product, why wouldn't you do that? And then the compression and the pressure on those other businesses goes super critical. ... If I were Google, that's what I'd do.

rebuts · 1supports · 1

07
Claim

Companies that stay private longer and then go public force retail investors and 401k holders to become bag holders of overpriced shares they were excluded from during the value-creation phase.

Maris sharpens his critique of extended private market timelines, arguing that retail investors — particularly those saving for retirement through 401ks — will be forced to buy overpriced shares in companies whose value was created while only elite investors had access.

transcript

Bill Maris: We're going to force overpriced products on the 401k holders of America who didn't get to participate early. ... It creates more wealth creation for the people who don't need it. And it makes the people's retire accounts the bag holders.

rebuts · 1

08
Claim

Companies that stay private too long force overpriced products onto retail 401k holders, contradicting their stated mission of benefiting humanity.

Maris argues that large companies staying private longer and receiving exceptions to public market rules ultimately force retail retirement accounts to become bag holders, while the companies cloak themselves in public benefit language.

transcript

Bill Maris: And where do you think we are on that curve of value creation? Could they go 3X from here? Sure. But they... So the, just to say it as plainly as possible, we're going to force overpriced products on the 401k holders of America who didn't get to participate early. This is your position that this is profoundly in there and creates more wealth creation for the people who don't need it. And it makes the people's retire accounts the bag holders. There's a lot of risk in that. And My objection is don't say you're doing this for the benefit of humanity and do the other thing. Make the public's retirement accounts the back holders. Or just say, this is how we're running our business and this isn't for the benefit of humanity.

09
Prediction◆

AI today is in an 'Atari command line stage' analogous to 1980s text-adventure games like Zork — brittle, no memory, session resets — and will reach a 'PlayStation 10 stage' within five years through platforms (physics engines, controllers, GPUs), not just bigger models.

Maris compares today's AI to the text-adventure game Zork — brittle, lacking memory and consistency — and predicts that within five years the industry will reach a photorealistic, ambient-computing stage driven by infrastructure platforms, not just larger models.

transcript

Bill Maris: I'm going to make an analogy to the gaming industry. We all get asked and we all think about, well, what does the future look like when AI is everywhere? ... Let's look at the gaming industry. So I used to play this game, Zork. There's one called Planet Fall back in the 80s. And it was very brittle. It was turn response, turn response, grab a lamp. Oh, I didn't, it's a lantern. I should have said lantern. Go north and you wait for the computer to respond. Let's show the most sophisticated retail available AI system out there today on the next slide and tell me how different it looks. ... What's happened to the gaming industry from the 80s to today is going to happen in AI, but in the next like 5 years. ... On the AI side, there will be ambient computing. There will be the problems that Zork had will be solved for AI. Lack of memory, lack of consistency, session resets, and so forth. ... I don't plan on investing in kind of larger models, right? Just like it wasn't better stories that would make better games. It was controllers and physics engines and GPUs. And those are the parts of the AI cycle that I'm interested in, which is all the platforms that need to be built to ... make this reality real in the next five years. And it's not just bigger models. I think we're at the Atari command line stage of AI, and we're going to get to the PlayStation 10 stage in the next five years.

10
Mechanism

Venture capital's incentive structure is broken: a $5B fund returning 1.01x earns more for its GP than a $500M fund returning 3x, and large funds outbid smaller ones on valuation to deploy capital, hurting entrepreneurs.

Maris explains that large fund incentives are misaligned because GPs earn more fees on giant funds even with mediocre returns, and these funds outbid smaller investors by offering enormous checks at inflated valuations that entrepreneurs find hard to refuse.

transcript

Bill Maris: So A $5 billion venture fund that returns 1.01x gets to say that they are in the 75th percentile and can raise their next fund, and no one at the Stanford endowment is going to get in trouble for writing that check. ... So now let's look at the GP dynamic. Well, if I have a $5 billion fund, I return 1.01x, I'm going to make more money than Bill with his $500 million fund that returns 3x. Okay, so that's also a strange incentive. ... Giant fund Y, we're friends, it's a different model, but giant fund Y says, well, we have this giant fund. We need to put 250 million in. And then an entrepreneur says, well, my company's valuation is 100. No, your valuation is now 4 billion. And we'll give you 250 million for a percent of your company. They're going to take that deal every day.

explains mechanism · 2extends · 2

11
Claim

Venture capital incentive structures are fundamentally broken: a $5B fund returning 1.01x makes its GPs more money than a $500M fund returning 3x, and LPs reward mediocrity because nobody at an endowment gets fired for writing a big check to a name-brand fund.

Maris argues that the VC industry has misaligned incentives where large funds with mediocre returns are rewarded more than smaller funds with excellent returns, and LPs have no incentive to demand better performance from brand-name mega-funds.

transcript

Bill Maris: A $5 billion venture fund that returns 1.01x gets to say that they are in the 75th percentile and can raise their next fund, and no one at the Stanford endowment is going to get in trouble for writing that check. They need to put 2 or 500 million into a fund multiple times. So I understand that dynamic. So now let's look at the GP dynamic. Well, if I have a $5 billion fund, I return 1.01x, I'm going to make more money than Bill with his $500 million fund that returns 3x. Okay, so that's also a strange incentive.

supports · 1

Highlight slides
Small VC Funds Outperform Large Funds✦ from: Small venture capital funds outperform large funds, as shown by the data.Top Decile DPI Performance by Fund Size✦ from: Small venture capital funds outperform large funds, as shown by the data.Small VC funds dramatically outperform large funds✦ from: Small VC funds outperform large funds; funds below $750M averaged 4.76x DPI returns vs. 2.42x for funds over $1B, and 95% of top-decile performers were sub-$750M.Why smaller funds win✦ from: Small VC funds outperform large funds; funds below $750M averaged 4.76x DPI returns vs. 2.42x for funds over $1B, and 95% of top-decile performers were sub-$750M.Google could crush OpenAI and Anthropic by slashing token prices 80%✦ from: If Google arbitrarily cuts token prices by 80%, it would destroy the business models of OpenAI and Anthropic because customers would switch to a cheaper, essentially identical product.Google's Pricing Leverage✦ from: Google could crush AI competitors like OpenAI and Anthropic by arbitrarily cutting token prices, given its massive war chest.Impact on OpenAI & Anthropic✦ from: Google could crush AI competitors like OpenAI and Anthropic by arbitrarily cutting token prices, given its massive war chest.Why Google would do it✦ from: If Google arbitrarily cuts token prices by 80%, it would destroy the business models of OpenAI and Anthropic because customers would switch to a cheaper, essentially identical product.AI Today = Zork (1980s Text-Adventure)✦ from: AI today is in an 'Atari command line stage' analogous to 1980s text-adventure games like Zork — brittle, no memory, session resets — and will reach a 'PlayStation 10 stage' within five years through platforms (physics engines, controllers, GPUs), not just bigger models.AI in 5 Years = PlayStation 10✦ from: AI today is in an 'Atari command line stage' analogous to 1980s text-adventure games like Zork — brittle, no memory, session resets — and will reach a 'PlayStation 10 stage' within five years through platforms (physics engines, controllers, GPUs), not just bigger models.Key Infrastructure (What to Build)✦ from: AI today is in an 'Atari command line stage' analogous to 1980s text-adventure games like Zork — brittle, no memory, session resets — and will reach a 'PlayStation 10 stage' within five years through platforms (physics engines, controllers, GPUs), not just bigger models.
Related episodes
Video · a16z Podcast

Ben Horowitz on the Global Race for Tech, Power, and Influence

→
Video · Lenny's Podcast

The rise of taste, human authenticity and judgment in an AI world | Adam Mosseri (Head of IG)

→
Video · a16z Podcast

The $1 Trillion Firm That Refuses The Private Equity Label | a16z

→