Bill Maris: How Google Could Crush AI Competitors, Why Small Funds Win, and AI's Atari Stage
In this episode of All-In with Chamath, Jason, Sacks & Friedberg, Bill Maris shares insights from his experience leading Google Ventures and founding Section 32. Maris presents data showing that smaller venture capital funds consistently outperform larger ones, explaining how fund structure and incentive misalignment drive this pattern. He discusses AI's current limitations and predicts a fundamental shift from scaling language models to investing in infrastructure and physics engines.
The conversation covers deep tech investment opportunities in life sciences and computational biology, along with concerns about regulatory challenges and brain drain affecting U.S. innovation. Maris and the hosts examine ethical issues in venture capital, including how prolonged private company status disadvantages retail investors. Throughout the episode, Maris draws on his entrepreneurial journey and contrarian approach to investing, offering perspective on identifying transformative technology shifts and building successful venture funds.
This is a preview of the Shortform summary of the Jun 9, 2026 episode of the All-In with Chamath, Jason, Sacks & Friedberg
Sign up for Shortform to access the whole episode summary along with additional materials like counterarguments and context.
1-Page Summary
Why Smaller Venture Capital Funds Outperform Larger Funds
Smaller Venture Funds Yield Better Returns Than Larger Ones
Data from top decile DPI performance reveals that funds smaller than $750 million average returns of 4.76x, while those over $1 billion return just 2.42x. Nearly 95% of top decile performers were funds below $750 million. Bill Maris notes that his investments surpassed Google Ventures' benchmarks of around 4.1x returns between 2009 and 2018, while Section 32's six funds—averaging $400 million each—have achieved top decile performance with investments in companies like CrowdStrike, Cohere, and Coinbase.
Fund Structure Constraints Create Unrealistic Return Thresholds
Larger fund structures introduce daunting hurdles for meaningful returns. A $500 million fund needs $15 billion in exits for a 3x return—challenging but achievable. A $7 billion fund, however, would need $210 billion for the same multiple, often exceeding total annual venture-backed exit value. This forces large funds to compete for fewer massive deals, increasing risk and making consistent top-tier returns extremely difficult.
Superior Focus and Founder Engagement Set Smaller Funds Apart
Smaller funds benefit from selective investments and hiring, leading to better evaluation and cultural fit. Maris, who has managed both small and multi-billion dollar funds, points out that large funds become distracting, making it difficult to give founders needed attention. Smaller funds maintain concentrated ownership around 10% per deal, aligning interests between fund managers and entrepreneurs, while large funds often end up with diluted stakes and less meaningful engagement.
Perverse Incentives Benefit Fund Managers Despite Performance
Incentive structures in venture capital can reward scale over performance. A $5 billion fund returning just 1.01x can claim 75th percentile status, enabling managers to raise subsequent funds despite mediocre results. This creates fractured alignment where general partners benefit from scaling fund size over generating high returns. Deep-pocketed funds inflate valuations competing to deploy huge sums, pressuring entrepreneurs to accept outsized checks that may not serve their long-term interests.
AI Shift: From Large Models to Infrastructure and Physics Engines
The Gaming Industry's Transformation Parallels AI's Coming Evolution
Maris reflects on playing text-based games like Zork in the 1980s, where limited command-line interfaces made interactions brittle and inflexible. Today's gaming industry has transformed to photorealistic environments with persistent worlds and advanced physics simulation. This leap came not from incremental text improvements but from breakthroughs in controllers, physics engines, and GPUs. Maris argues that AI will undergo a similar transformation within five years, but the timeline will be dramatically compressed due to accelerating technology.
AI Systems Are Like Atari-Era Computing
Maris likens today's AI systems to Atari-era computing, predicting AI will soon leap from its current state to something comparable to PlayStation 10 in just five years. This advancement will come not from increasing model size or parameter count, but from investments in infrastructure: platforms, computing machinery, physics engines, and hardware. These foundational elements will move AI from its brittle, text-based phase to a far more dynamic and realistic stage.
Investing Strategically in AI-Enabling Technologies Yields Better Returns
Drawing from gaming history, Maris argues the industry's revolution came from infrastructure developments like new controllers and powerful GPUs, not from making command-line systems better. He applies the same logic to AI, prioritizing investment in tooling, middleware, and infrastructure that will enable next-generation AI experiences. These components will be the bottleneck and catalysts for future AI progress, making them more valuable than scaling foundational large language models.
Deep Tech Investment: Life Sciences, Computational Biology, AI-Driven Healthcare
Healthcare and Human Biology Are Large Markets Accelerated by Computational Approaches
Life sciences represent the largest total addressable market across all sectors, justifying significant venture capital allocation. Maris discusses how longevity, once fringe science, is now a core investment theme. Section 32 invests in longevity and age-related disease sectors, including companies like New Limit, founded by Blake Byers and Brian Armstrong. Advances in computational biology accelerate research and drug screening at lower costs than traditional wet lab experiments, though therapeutic ventures requiring human clinical trials remain specialist areas due to complexity and regulatory hurdles.
Deep Tech Sectors Now More Accessible Through AI and Simulation
Historically, deep tech ventures were long-cycle, capital-intensive, and high-risk endeavors. SpaceX's success demonstrated that deep tech business models can succeed, inspiring new entrepreneurs. Now, AI-driven physics engines, simulations, and advanced modeling tools enable faster product iteration, lower prototyping costs, and help de-risk technical approaches in hardware-intensive fields, making these sectors more accessible for startups.
Computational Biology and In-Silico Simulation Are Top Investment Areas
Computational biology and computer-based simulation have emerged as top investment areas in life sciences. Maris emphasizes that if a realistic simulation of a human cell in silico is achieved, drug discovery will accelerate dramatically. However, current computational biology still lacks the fidelity to replace experimental validation. Because of regulatory burdens like FDA approval, platform investment in computational biology is seen as likely to yield higher risk-adjusted returns than direct investment in therapeutic companies.
Regulatory and Government Support Shape Life Sciences Investment
Maris points out that recent funding cuts and anti-science sentiment in the U.S. have diminished the country's capacity for basic research, driving talent elsewhere. David Friedberg highlights that countries like China are recruiting top scientists from Europe and India—talent that once fed U.S. research institutions. Tighter H1B visa regulations and immigration barriers further exacerbate this brain drain, posing substantial risk for U.S.-based innovation and leadership in deep tech and life sciences.
Ethical Concerns in Venture Capital
Modern venture capital is marked by late-stage private companies remaining private longer than ever, creating deep ethical and financial consequences for retail investors and the broader public.
Prolonged Private Status Benefits Early Investors but Risks Retail Investors
Unicorn companies increasingly seek exceptions to S&P 500 inclusion rules, forcing passive funds, ETFs, and 401k holders to acquire stakes in late-stage, overpriced private companies. Maris and Jason Calacanis warn that only about 1% of investors—typically the earliest ones—have access to high-growth periods, while 99% buy in when valuations peak. As these bloated private positions enter indices, they inflate retirement accounts, making the general public "bag holders" who shoulder downside after others have extracted maximum gain.
Public Benefit Corporations Deceive Investors
Maris criticizes venture-backed firms that adopt public benefit language while largely prioritizing returns for insiders. Despite branding themselves as agents of societal progress, these firms structure financials to maximize insider profit, only bringing inflated prices to public markets once appreciation opportunities have passed. True commitment to societal benefit would show through earlier IPOs and equitable access to upside.
Incentives Enable Predatory Pricing
Competitive dynamics enable dominant, well-capitalized incumbents to use predatory pricing to threaten smaller players. Maris poses a hypothetical where Google could cut AI token prices by 80%, destroying OpenAI's economic viability. Friedberg and Calacanis note such subsidized pricing mirrors Uber's strategy: using investor capital as a weapon to capture market share and crush competition, rendering long-term unit economics irrelevant.
Uncertainty in Public Market Receptiveness Poses Downside Risk
Maris stresses that for late-stage firms like SpaceX or Anthropic, the business case will have to be proven in public markets, which is avoided during long fundraising stretches. Venture firms attempt to manage this risk through extended lockups and staged offerings, signaling anxiety over what public markets might actually pay. This gap between top decile and median venture returns masks losses industry-wide, threatening retail investors and employees holding equity at artificially high private valuations.
Career Insights and Future Vision
Recognizing Transformative Tech Shifts Offers Advantage for Early Founders and Investors
Maris recalls seeing a server in a Wall Street office closet housing the company's email and websites. Realizing he could multiply this concept and host for many clients, he foresaw the rise of centralized data centers and web hosting. Inspired, Maris left Wall Street immediately, using credit cards to move to Vermont and found a web hosting business in 1997. He emphasizes that identifying transformative trends often seems "insane," likening it to glimpsing the future through a keyhole. For Maris and other entrepreneurs, trusting their future-oriented vision—even amid skepticism—proved essential.
Entrepreneurial Resilience Distinguishes Successful Founders
Running his business from a small Vermont apartment, Maris faced extreme conditions. Servers emitted so much heat that opened windows caused rooms to freeze—glasses of water iced over by noon. When a thunderstorm caused the roof to leak onto servers, Maris bought tar and a mop and ascended onto the roof during lightning and rain to patch the leak himself. Despite tarring himself into a corner, he saved his servers—his shoes are still stuck on that Vermont rooftop. This willingness to do whatever it takes distinguishes founders who ultimately succeed.
Data Science and Machine Learning Enhance Venture Investment
At Google Ventures, Maris and his team used machine learning to construct optimal investment portfolios, running millions of simulations to determine fund composition and sizing. From 2009 to 2018, Google Ventures portfolios achieved approximately 4.1x returns, placing them in the top decile. Maris credits this direct application of computer science as the major driver of their performance, concluding, "Don't bet against computer science."
Contrarian Career Choices Create Wealth and Impact Opportunities
After helping found Google Ventures, Maris chose to start his own venture fund, resisting advice to raise the largest possible pool for management fees. He believed smaller funds produce greater returns due to their ability to focus on founders and avoid the distractions of managing vast sums. Maris founded Section 32 and launched six funds—each averaging $400 million—investing in companies like CrowdStrike, Cohere, and Coinbase. All six funds have performed in the top decile, affirming that small, focused funds driven by contrarian discipline and empirical analysis outperform industry norms.
1-Page Summary
Additional Materials
Clarifications
- DPI stands for "Distributions to Paid-In" capital and measures the cash returned to investors relative to the amount they invested. It reflects realized returns, showing how much money has actually been paid out, unlike other metrics that include unrealized gains. DPI is significant because it indicates the liquidity and success of a fund in returning capital to its limited partners. High DPI values suggest strong fund performance and effective exit strategies.
- "Top decile performers" refers to funds that rank within the highest 10% based on their returns compared to all other funds. This classification highlights exceptional performance, distinguishing these funds from the majority. Being in the top decile often attracts more investors and better deal flow due to proven success. It also implies these funds have outperformed 90% of their peers in the same category or time period.
- Larger funds have more capital to deploy, so achieving the same return multiple means generating much higher absolute profits. Since exit opportunities are limited, raising a huge fund forces managers to seek extremely large deals or company valuations. This scarcity of mega-exits increases competition and risk, making consistent high returns harder. Smaller funds can target smaller, more frequent wins that cumulatively yield strong multiples without needing outsized exits.
- "Concentrated ownership around 10% per deal" means a venture fund holds a significant equity stake in each startup it invests in, typically about 10%. This level of ownership allows the fund to have meaningful influence and potential for substantial returns if the company succeeds. It also aligns the interests of the fund managers and founders, fostering closer collaboration and support. Smaller funds can maintain such stakes more easily than larger funds, which often must dilute ownership across many investments.
- Venture capital funds typically charge a management fee (around 2%) on total assets and take a carried interest (usually 20%) on profits. Larger funds generate higher fees regardless of performance, incentivizing growth in fund size over returns. This can lead to managers raising bigger funds even if returns diminish, as fees remain lucrative. Consequently, fund managers may prioritize scaling assets under management rather than maximizing investor gains.
- Inflated valuations occur when startups are valued higher than their actual financial performance justifies, often due to competitive funding environments. This can lead to unrealistic expectations for future growth and exit prices. When valuations correct downward, later investors and employees holding equity may face significant losses. Inflated valuations also pressure startups to prioritize rapid scaling over sustainable business models.
- The analogy compares current AI to early, simple video game consoles like Atari, which had limited graphics and capabilities. "PlayStation 10" symbolizes a future AI state with vastly improved realism, complexity, and interactivity, similar to how modern gaming consoles far surpass early ones. This highlights that AI's major leap will come from better infrastructure and hardware, not just bigger models. The comparison emphasizes rapid, transformative progress rather than gradual improvement.
- In AI development, "tooling" refers to software tools that help build, test, and deploy AI models efficiently. "Middleware" is the software layer that connects different AI components and systems, enabling them to communicate and work together smoothly. "Infrastructure" includes the physical hardware and cloud resources, like servers and GPUs, that provide the computing power needed for AI tasks. Together, these elements create the foundation that supports advanced AI applications beyond just the models themselves.
- Total addressable market (TAM) is the total revenue opportunity available for a product or service if it achieved 100% market share. Life sciences rank highest because they encompass vast sectors like pharmaceuticals, biotechnology, medical devices, and healthcare services, all addressing fundamental human health needs. These sectors have large, growing demand driven by aging populations, chronic diseases, and technological advances. The scale and essential nature of healthcare create enormous economic potential compared to other industries.
- Computational biology uses algorithms and data analysis to understand biological systems, often analyzing genetic or molecular data. In-silico simulation refers to computer-based models that mimic biological processes to predict outcomes without physical experiments. Traditional wet lab experiments involve hands-on testing with biological materials like cells or chemicals in a laboratory setting. Wet labs provide empirical validation, while computational methods offer faster, cost-effective hypothesis generation and screening.
- The FDA approval process requires extensive clinical trials to prove a drug's safety and effectiveness, which is time-consuming and costly. This regulatory scrutiny increases uncertainty and delays potential returns for investors in life sciences. Many drug candidates fail during trials, raising the risk of total loss on investment. Consequently, investors often prefer platform technologies or computational tools that face fewer regulatory hurdles.
- Public benefit corporations (PBCs) are legally required to balance profit with social or environmental goals. However, some PBCs prioritize financial returns for insiders over their stated public benefits. This can mislead investors who expect genuine societal impact alongside profits. The legal structure allows them to claim social responsibility without sacrificing shareholder value.
- Predatory pricing occurs when a dominant company temporarily lowers prices below cost to drive competitors out of the market. This strategy can eliminate smaller rivals who cannot sustain losses, reducing competition long-term. Once competitors exit, the dominant firm may raise prices to recoup losses and increase profits. In tech markets, this can stifle innovation and limit consumer choice.
- Lockups are contractual periods after an IPO during which insiders and early investors cannot sell their shares, preventing sudden stock price drops. Staged offerings involve releasing shares to the public in multiple phases over time, rather than all at once, to manage market impact and valuation. Both mechanisms help stabilize stock prices and build investor confidence during a company's transition to public markets. They also allow companies to raise capital gradually while demonstrating performance milestones.
- "Bag holders" are investors left holding assets that have significantly dropped in value after others have sold at higher prices. In late-stage private companies, early investors often exit or profit before public market investors can buy shares. Retail investors typically enter at peak valuations, facing higher risk of losses if the company's value declines post-IPO. This dynamic can cause retail investors to bear the financial downside while earlier investors reap most gains.
- Machine learning analyzes large datasets to identify patterns and predict outcomes, helping investors assess potential returns and risks. It can simulate millions of portfolio combinations to optimize asset allocation and diversification. This approach reduces human bias and improves decision-making efficiency in selecting startups. Ultimately, it enables venture funds to construct portfolios with higher probabilities of strong performance.
- Smaller venture funds can be more agile, allowing managers to build closer relationships with founders and provide tailored support. They often focus on niche markets or early-stage startups where they can add more value and identify unique opportunities. Smaller funds face less pressure to deploy large amounts of capital quickly, reducing the risk of overpaying for investments. Additionally, concentrated ownership stakes in smaller funds align incentives more closely between investors and entrepreneurs.
- CrowdStrike is a cybersecurity company known for its cloud-native endpoint protection platform. Cohere specializes in natural language processing and AI, providing language models for businesses. Coinbase is a leading cryptocurrency exchange platform facilitating digital asset trading. Their success exemplifies how smaller venture funds can achieve top returns by investing in high-growth, innovative tech companies.
Counterarguments
- The outperformance of smaller funds may be partly due to survivorship bias, as unsuccessful small funds are less visible or may not report returns, skewing the data in favor of those that succeed.
- Larger funds often have access to later-stage deals and can participate in follow-on rounds, which may offer lower multiples but greater absolute dollar returns and lower risk profiles, appealing to certain institutional investors.
- The ability of smaller funds to maintain high ownership percentages and close founder engagement may not scale for all sectors or geographies, especially in highly competitive or capital-intensive industries.
- Incentive misalignment is not unique to large funds; smaller funds can also face conflicts of interest, such as pressure to show quick returns or to raise subsequent funds.
- Large funds can provide critical capital for scaling companies that require significant resources, such as those in deep tech, infrastructure, or life sciences, where small funds may be unable to meet capital needs.
- The analogy between gaming industry evolution and AI development may oversimplify the technical and market challenges unique to AI, which may not follow the same trajectory or speed of transformation.
- Investing in AI infrastructure and tooling does not guarantee superior returns, as these areas can become crowded and subject to rapid commoditization, reducing long-term margins.
- While computational biology and in-silico simulation are promising, the regulatory and scientific hurdles remain significant, and breakthroughs may take longer than anticipated, limiting near-term returns.
- The assertion that U.S. innovation is threatened primarily by funding cuts and immigration policy overlooks other factors, such as global competition, changing research priorities, and the increasing role of private capital in funding basic research.
- The criticism of public benefit corporations may not account for genuine efforts by some firms to balance profit and purpose, and delayed IPOs can also reflect prudent business strategy rather than solely insider enrichment.
- Predatory pricing and market dominance are not exclusive to large, well-capitalized incumbents; smaller or newer entrants can also engage in aggressive pricing or unsustainable growth tactics.
- The use of data science and machine learning in venture investing is widespread, and while helpful, it does not eliminate the inherent unpredictability and risk of early-stage investing.
- Contrarian or smaller fund strategies may not be suitable for all investors, particularly large institutions with mandates to deploy significant capital or those seeking lower volatility and more predictable returns.
Get access to the context and additional materials
Why Smaller Venture Capital Funds ($500m-$750M) Outperform Larger Funds: Data Analysis of Returns and Structural Incentives
Smaller Venture Funds Yield Better Returns Than Larger Ones
Data from top decile DPI performance shows that funds smaller than $750 million average returns of 4.76x, while those over $1 billion return just 2.42x. Nearly 95% of the top decile performers over the measured period were funds below $750 million, with a distinct compression of returns appearing as fund size increases beyond this threshold. For example, Google Ventures achieved around 4.1x returns between 2009 and 2018 according to internal estimates. Bill Maris, reflecting on his own track record, notes that investments he led surpassed even Google Ventures benchmarks. Similarly, Section 32’s six funds—all averaging about $400 million—are cited as top decile performers, having made successful investments in companies such as CrowdStrike, Cohere, and Coinbase.
Fund Structures' Constraints Create Unrealistic Return Thresholds and Misaligned Incentives
Larger fund structures introduce daunting hurdles for meaningful returns. A $500 million fund, assuming an average 10% ownership per company, requires $5 billion in collective exit value just to return invested capital. To reach a 3x return target, such a fund would need $15 billion in exits—a challenging but achievable goal. Contrast this with a $7 billion fund, which would need to generate $210 billion in exit value for a 3x multiple. This figure often exceeds the total annual value of venture-backed M&A and IPO exits, making such targets largely unsustainable and exposing the fund to the risk that there may not be enough mega-deals available.
The constraints for large funds mean they must compete for fewer massive deals, increasing their risk profile and making consistent top-tier returns extremely difficult.
Superior Focus, Founder Engagement Set Smaller Funds Apart
Smaller funds benefit from their ability to make more selective investments and hire more selectively for teams, which leads to better evaluation and tighter cultural fit. Bill Maris, with experience managing both small and multi-billion dollar funds, points out that large funds become distracting due to their size and staff, making it difficult to give founders the attention they need.
In smaller funds, managers can maintain more concentrated ownership—typically around 10% per deal—which aligns the interests of fund managers and entrepreneurs. Large funds, required to deploy much bigger checks, often end up with diluted stakes and less meaningful engagement with each portfolio company.
Perverse Incentives Benefit Fund Managers Despite Performance, Undermining Alignment With Partners and Entrepreneurs
Incentive structures in venture capital can ...
Here’s what you’ll find in our full summary
Why Smaller Venture Capital Funds ($500m-$750M) Outperform Larger Funds: Data Analysis of Returns and Structural Incentives
Additional Materials
Clarifications
- DPI stands for "Distributions to Paid-In capital" and measures the cash or stock returned to investors relative to the amount they invested. It reflects realized returns, showing how much money has actually been paid out, unlike other metrics that include unrealized gains. DPI is a key indicator of a fund’s performance because it focuses on tangible returns rather than paper valuations. Higher DPI means investors have received more actual returns from the fund.
- "Top decile" refers to the top 10% of funds ranked by performance. It signifies the highest-performing segment within a group, indicating superior returns compared to 90% of peers. This metric helps investors identify funds that consistently deliver exceptional results. Being in the top decile is a strong indicator of fund quality and skillful management.
- Ownership percentage in venture capital refers to the portion of a company’s equity that an investor holds after investing. It represents the investor’s share of the company’s value and voting rights. This percentage determines how much influence the investor has and how much profit they receive if the company is sold or goes public. Ownership is diluted when the company issues more shares in future funding rounds.
- Exit value refers to the total amount of money a venture capital fund receives when it sells its stake in a portfolio company, typically through an acquisition or an initial public offering (IPO). This value determines the fund’s returns because it represents the realized gains from investments. Higher exit values mean the fund can return more capital to its investors, increasing the multiple on invested capital. Exit value is crucial since it directly affects whether a fund meets or exceeds its target return thresholds.
- A "3x return target" means the fund aims to triple the money invested by its limited partners. This target is important because it reflects a benchmark for successful venture capital performance, compensating for risk and time. Achieving a 3x return ensures the fund generates significant profits beyond just returning the original capital. It also influences how funds are structured and how investments are selected.
- Limited partners (LPs) are investors who provide the capital to a venture fund but do not manage its daily operations. General partners (GPs) are the fund managers who make investment decisions and run the fund. GPs earn management fees and a share of profits (carried interest), while LPs receive returns based on the fund’s performance. LPs have limited liability, meaning they risk only their invested capital, whereas GPs have unlimited liability for the fund’s obligations.
- Venture capital funds pool money from investors to invest in startups, typically structured with a fixed lifespan and target return. Fund size dictates how many and how large investments must be, influencing risk tolerance and portfolio diversification. Smaller funds can focus on early-stage, high-potential startups with concentrated ownership, while larger funds often invest in later-stage companies requiring bigger capital but yielding diluted stakes. This structural difference shapes investment strategy, founder engagement, and potential returns.
- Larger funds must invest more capital across many companies, so their total return target scales up accordingly. Because each investment typically yields a similar ownership percentage, the fund’s overall exit value must be much higher to achieve the same multiple. This creates a nonlinear challenge, as the market has limited opportunities for extremely large exits. Thus, bigger funds face greater pressure to find rare, massive successes to meet return goals.
- "Diluted stakes" means that a fund owns a smaller percentage of a company after multiple funding rounds or when large investments are made. This reduces the influence and potential financial gain for the fund manager and entrepreneurs. Smaller ownership can limit a fund manager’s ability to impact company decisions and align interests with founders. For entrepreneurs, diluted stakes may mean less control and smaller personal returns from the company’s success.
- Inflated valuations occur when startups are priced higher than their fundamental business value, often due to intense competition among investors. Large funds with substantial capital must deploy big checks quickly, pushing prices up to secure deals. This can lead to unrealistic expectations for growth and profitability. Over time, inflated valuations increase risk for both investors and entrepreneurs if the company fails to meet these heightened benchmarks.
- Mega-deals are very large investment exits, often exceeding hundreds of millions or billions of dollars. They disproportionately drive returns for large venture funds because these funds need massive exits to meet their high return targets. Howe ...
Counterarguments
- Larger funds may provide critical capital for later-stage companies, enabling startups to scale globally and compete with established incumbents.
- The data cited may be subject to survivorship bias, as successful smaller funds are more likely to be highlighted, while unsuccessful ones are less visible.
- Larger funds can offer portfolio companies access to broader networks, resources, and follow-on funding, which can be valuable for long-term growth.
- Some sectors, such as deep tech or capital-intensive industries, require larger investments that only big funds can provide.
- The venture capital landscape is cyclical, and periods of high returns for smaller funds may not persist across different market environments.
- Larger funds can diversify across more companies and sectors, potentially reducing risk through broader exposure.
- The alignment of interests between fund managers and entrepreneurs is not solely determined by fund size; governance structures and fund philosophy also play significant roles.
- Some ...
Get access to the context and additional materials
Ai Shift: From Large Models to Infrastructure and Physics Engines
The Gaming Industry's Shift From Text-Based To Photorealistic Environments Parallels Ai's Five-Year Evolution
Bill Maris reflects on playing games like Zork and Planetfall in the 1980s, describing their limited, turn-based, command-line interfaces, which often made interactions brittle and inflexible—for example, typing "grab the lamp" instead of "lantern" led to confusion and halted gameplay. These early gaming systems restricted the potential for any real AI or immersive environment because interactions were so narrowly defined and processor limitations were stark.
Maris observes that the gaming industry today is radically transformed: games are now photorealistic, with persistent environments, advanced physics simulation, and intuitive controllers. The leap from primitive command-line play to these rich, modern experiences did not come from incremental improvements to text input or storyline but from breakthroughs in gaming controllers, physics engines, and graphical processing units (GPUs). He emphasizes that the step change was infrastructure-driven, not a slow march of incremental story improvements. Maris argues that a similar transformation is about to occur in AI within the next five years, but the timeline will be dramatically compressed due to the accelerating pace of technology.
Ai Systems Are Like Atari-Era Computing; Improvements in Five Years Through Infrastructure, Not Model Size
Maris likens today’s AI systems to Atari-era computing, where the limitations facing those early systems mirror current AI constraints—both in usability and realism. He predicts that AI will soon experience a leap comparable to the gaming industry’s trajectory from Atari to PlayStation 10 within just five years. This future advancement will come not from increasing the size of language models, training data, or parameter count, but from investments in infrastructure: platforms, computing machinery, physics engines, and hardware. These foundational elements, not simply larger models, will move AI from its current brittle, text-based phase to a far more dynamic and realistic stage comparable to the most advanced gaming environments.
Investing Strategically In Ai-enabling Technologies and Platforms Yields Better Returns Than in Foundational Large Langua ...
Here’s what you’ll find in our full summary
Ai Shift: From Large Models to Infrastructure and Physics Engines
Additional Materials
Clarifications
- Zork and Planetfall are early text-based adventure games from the late 1970s and early 1980s. They relied entirely on typed commands to interact with the game world, without graphics or sound. These games are significant because they represent the origins of interactive storytelling and player-driven exploration in gaming. Their limitations highlight how far gaming technology has evolved from simple text input to immersive, graphical experiences.
- A command-line interface (CLI) requires users to type exact text commands to interact with a system. It lacks flexibility because the system only recognizes specific words or phrases, so slight variations cause errors. This rigidity makes interactions fragile, as users must guess the correct commands precisely. Modern interfaces use graphics and natural language to allow more intuitive and forgiving interactions.
- Photorealistic environments in gaming are virtual worlds designed to look as realistic as real life, with detailed textures, lighting, and shadows. Persistent environments mean the game world continues to exist and evolve even when the player is not actively interacting with it. This persistence allows for ongoing changes, such as weather shifts or player-driven events, that affect the game state over time. Together, these features create immersive, believable experiences that feel alive and continuous.
- Physics engines are software components that simulate real-world physical behaviors like gravity, collisions, and object movement in virtual environments. They enable games to create realistic interactions, such as a ball bouncing or characters responding naturally to forces. In AI, physics engines help create dynamic, believable simulations for training and testing models in environments that mimic real-world physics. This realism allows AI systems to learn and operate in more complex, lifelike scenarios beyond static or purely text-based inputs.
- Graphical Processing Units (GPUs) are specialized hardware designed to handle complex calculations for rendering images and videos quickly. Their architecture allows them to perform many operations in parallel, making them ideal for both graphics and large-scale data processing tasks. In AI, GPUs accelerate the training and inference of machine learning models by efficiently managing the massive computations involved. This speed and efficiency enable more complex and realistic simulations in gaming and faster, more powerful AI systems.
- The "Atari-era computing" analogy refers to the early days of video games when hardware and software were very limited, resulting in simple graphics and gameplay. Current AI systems are similarly constrained by their infrastructure, limiting their capabilities and realism. Just as gaming advanced dramatically with better hardware and engines beyond Atari, AI progress depends on improved computing platforms and tools, not just bigger models. This comparison highlights that foundational technology upgrades, not incremental tweaks, drive major leaps in performance and experience.
- In AI models, "parameter count" refers to the number of adjustable elements, like weights, that the model uses to learn from data. These parameters determine how the model processes input and generates output. Larger parameter counts generally allow models to capture more complex patterns but require more computing power. Increasing parameters alone does not guarantee better performance without supporting infrastructure.
- Foundational large language models are massive AI systems trained on vast text data to understand and generate human language. AI-enabling technologies and platforms include the hardware, software tools, and infrastructure that support and enhance AI applications. These technologies improve AI performance, scalability, and integration beyond just the model's capabilities. Investing in infrastructure focuses on building the environment where AI can operate effectively, not just the AI model itself.
- In AI infrastructure, "tooling" refers to software tools that help developers build, test, and deploy AI models efficiently. "Middleware" is software that connects different AI components or systems, enabling them to communicate and work together smoothly. "Simulation tools" create virtual environments to test AI behavior and performance under controlled conditions. Together, th ...
Counterarguments
- While infrastructure improvements have been crucial in gaming, advances in content creation, storytelling, and game design have also played significant roles in enhancing user experience and immersion.
- The analogy between gaming industry evolution and AI development may oversimplify the unique challenges and requirements of AI, which involve not just rendering or simulation but also reasoning, language understanding, and generalization.
- Recent progress in AI capabilities, such as large language models, has been primarily driven by scaling model size and data, suggesting that further improvements in model architecture and training methods could still yield significant gains.
- Infrastructure and hardware advancements alone may not address fundamental limitations in current AI algorithms or models, such as reasoning, common sense, or explainability.
- Investment in foundational models and algorithms can lead to breakthroughs that enable new applications and use cases, which in turn can drive demand for better infrastructure.
- The timeline for AI transformation may be optimistic, as breakthroughs ...
Get access to the context and additional materials
Deep Tech Investment: Life Sciences, Computational Biology, Ai-driven Healthcare
Healthcare and Human Biology Are Large Markets Accelerated by Computational and Ai Approaches
The life sciences, particularly human biology and healthcare, represent the largest total addressable market (TAM) across all sectors, justifying significant venture capital allocation and attention. Bill Maris discusses how longevity, once considered fringe science, is now a core investment theme as the market recognizes its vast potential. Section 32, for instance, invests in the longevity and age-related disease sector, including companies like New Limit, founded by Blake Byers and Brian Armstrong, signaling a shift toward innovative health solutions that target the core problems of aging.
Advances in computational biology also accelerate research, hypothesis testing, and drug screening, allowing startups and investors to move faster and with lower costs compared to traditional wet lab experiments. However, therapeutic ventures that require human clinical trials remain specialist investment areas due to the complexity and regulatory hurdles involved. As Maris notes, even after discovering a promising compound, only about 5% of the work is done—substantial safety testing and regulatory compliance are still required.
Deep Tech Sectors, Once Defined by Long Investment Cycles, Capital Intensity, and High Risk, now More Accessible for Startups Through Ai and Physics Simulation Capabilities
Historically, deep tech ventures were seen as long-cycle, capital-intensive, and high-risk endeavors. The success of Elon Musk and SpaceX has demonstrated that deep tech business models can succeed, inspiring new entrepreneurs to tackle significant societal challenges. Now, AI-driven physics engines, simulations, and advanced modeling tools enable faster product iteration, lower prototyping costs, and help de-risk technical approaches in hardware-intensive fields, making these sectors more accessible for startups.
AI and simulation technologies also accelerate engineering development, reduce capital requirements for proof-of-concept, and improve market validation and revenue opportunities for deep tech ventures. They allow for more agile experimentation, validation of technical assumptions, and a clearer path to commercialization compared to the legacy high-barrier approaches.
Computational Biology and In-silico Simulation Are Top Investment Areas in Life Sciences, With Cell Simulation Accelerating Therapeutic Discovery
Computational biology and in silico (computer-based) simulation have emerged as top investment areas in the life sciences. Investors now see immense potential in platforms and tools that accelerate therapeutic discovery, as digital simulation can conduct hypothesis testing and initial screening much more rapidly compared to traditional laboratory methods. Bill Maris emphasizes that if a realistic simulation of a human cell in silico is achieved, drug discovery and development will accelerate dramatically. While progress is rapid, Maris acknowledges that current computational biology lacks the fidelity to replace experimental validation; significant work remains before simulation-dominated drug discovery can fully supplant traditional approaches.
B ...
Here’s what you’ll find in our full summary
Deep Tech Investment: Life Sciences, Computational Biology, Ai-driven Healthcare
Additional Materials
Counterarguments
- While life sciences and healthcare represent large markets, other sectors such as energy, finance, or technology infrastructure can also have significant total addressable markets and attract substantial venture capital.
- The shift of longevity and age-related disease research from fringe to mainstream investment themes is still met with skepticism by some experts due to the complexity of aging and the lack of proven interventions that significantly extend healthy human lifespan.
- Computational biology accelerates early-stage research, but the translation of computational findings into clinically effective therapies remains challenging and often requires extensive experimental validation.
- The high failure rate and long timelines of therapeutic ventures are not solely due to regulatory hurdles; biological complexity and unpredictable human responses also play major roles.
- The success of companies like SpaceX is not easily replicable, as they benefited from unique leadership, timing, and access to capital, which may not be available to most deep tech startups.
- AI-driven simulations and modeling tools can reduce some costs and risks, but they may introduce new risks related to model accuracy, data quality, and overreliance on in silico results.
- The promise of in silico cell simulation is significant, but current models are far from capturing the full complexity of living systems, and breakthroughs may take much longer than anticipated.
- Platform investments in computational biology may face their own challenges, such as market saturation, competition, and the need for c ...
Actionables
- you can track and compare the global movement of scientific talent by following international research job postings and LinkedIn profiles of scientists, helping you spot which countries are gaining or losing expertise and consider how this might affect innovation in your field or region
- For example, set up alerts for research positions in the US, China, and Europe, and note trends in where top scientists are relocating, which can inform your perspective on where future breakthroughs may emerge.
- a practical way to understand the impact of regulatory hurdles on innovation is to map out the approval process for a recent medical device or drug, noting each step and the time required, so you can better appreciate why some investments or startups focus on platforms rather than direct therapeutics
- For instance, create a simple timeline or checklist for a new diabetes drug, highlighting preclinical, clinical, and regulatory milestones, and compare this to the development path of a computational biology tool.
- you can experiment ...
Get access to the context and additional materials
Ethical Concerns in Venture Capital: Prolonged Private Status of Late-Stage Companies Forcing Retail Investors Into Overpriced Positions and Misaligned Incentives Among GPs, Entrepreneurs, and LPs
The modern venture capital landscape is marked by late-stage private companies remaining private longer than ever, which creates deep ethical and financial consequences for retail investors and the broader public. This strategic delay in going public not only benefits insiders and early investors but also distorts markets and exposes ordinary individuals’ retirement accounts to significant risk and overvaluation.
Prolonged Private Status Benefits Early Investors but Risks Retail Investors
Unicorn companies increasingly seek exceptions to S&P 500 inclusion rules, circumventing regulatory protections that were originally designed to shield retail investors from volatile and opaque assets. Bill Maris and Jason Calacanis warn that these exceptions break fundamental rules, forcing passive funds, ETFs, and ultimately 401k holders to acquire stakes in late-stage, overpriced private companies without early-stage access.
Only about 1% of investors, typically the earliest ones, have access to high-growth periods; 99% are left buying in when valuations are at their peak, introducing severe risk for undiversified retirement portfolios. As these bloated private positions enter indices, they inflate the value of 401k and retirement accounts, making the general public “bag holders” who shoulder the downside after others have extracted maximum gain.
Public Benefit Corporations With Sustainability Narratives Deceive Investors and Misalign Governance
Venture-backed firms often adopt language around public benefit and stakeholder capitalism, claiming to act in humanity’s best interest while largely prioritizing returns for insiders. Bill Maris criticizes this tendency, observing that companies wrap themselves in “public benefit language” yet reserve most value creation for founders and a select investor class. Despite branding themselves as agents of societal progress, these firms structure their financials to maximize insider profit, only bringing inflated prices to public markets once appreciation opportunities have largely passed.
True commitment to societal benefit would show through earlier IPOs, equitable access to upside, and governance structures that prioritize stakeholder welfare—not just LP and founder gain. Instead, sustainability and public benefit claims often mask the extraction of maximum insider value before the risk is transferred to public investors.
Incentives in Market Structure for Larger Competitors to Undermine Smaller Ones Through Predatory Pricing
The competitive dynamics of venture-driven industries enable dominant, well-capitalized incumbents to use predatory pricing to threaten smaller players. Bill Maris poses the hypothetical where a tech giant like Google could arbitrarily cut AI token prices by 80%, destroying the economic viability of OpenAI’s models or any cost-sensitive rivals. David Friedberg and Jason Calacanis note that such subsidized pricing mirrors rideshare strategies like Uber’s: use investor capital and capital as a weapon—offering unsustainable prices to capture market share and crush competition.
This price compression by incumbents renders long-term unit economics largely irrelevant, undermining the via ...
Here’s what you’ll find in our full summary
Ethical Concerns in Venture Capital: Prolonged Private Status of Late-Stage Companies Forcing Retail Investors Into Overpriced Positions and Misaligned Incentives Among GPs, Entrepreneurs, and LPs
Additional Materials
Clarifications
- Late-stage private companies are firms that have progressed beyond early startup phases but have not yet gone public through an initial public offering (IPO). They typically have established products, significant revenue, and large valuations. Remaining private longer allows these companies to raise capital without public market scrutiny. This can delay transparency and price discovery for outside investors.
- Unicorn companies are privately held startups valued at over $1 billion. The term highlights their rarity and high growth potential. They often attract significant venture capital investment before going public. Their valuations can be based on future growth expectations rather than current profits.
- The S&P 500 inclusion rules determine which companies qualify to be part of the S&P 500 stock index, based on factors like market capitalization, liquidity, and public float. These rules ensure the index represents large, stable, and transparent companies, protecting investors from excessive volatility and opaque valuations. Regulatory oversight tied to these rules helps maintain market integrity and investor confidence. Exceptions to these rules can expose investors to higher risks from less transparent or overvalued companies.
- Passive funds are investment funds that track a market index rather than actively selecting stocks. ETFs (Exchange-Traded Funds) are a type of passive fund traded on stock exchanges like individual stocks. 401k holders are individuals investing retirement savings through employer-sponsored plans, often using passive funds or ETFs. These vehicles pool money from many investors to buy a broad range of assets, providing diversified exposure with lower fees.
- Early-stage access means investing in a company when it is just starting and its valuation is relatively low, allowing investors to benefit from significant growth. Peak valuations occur when a company is mature and its price is at or near the highest point, often before going public. Investors who buy at peak valuations face higher risk because the price may drop after the company eventually goes public or fails to meet expectations. Early investors gain outsized returns, while late investors often pay more and have less upside potential.
- Indices are collections of stocks or assets that represent a segment of the market, like the S&P 500. They serve as benchmarks for investment funds and ETFs, guiding what assets these funds buy. Traditionally, indices include publicly traded companies, but some now incorporate shares of private companies through special rules or exceptions. This inclusion exposes retail investors to private company risks indirectly via their index-based investments.
- Venture-backed firms are companies that receive funding from venture capitalists, who invest in early-stage or growth-stage businesses with high growth potential. These firms use the capital to develop products, expand operations, and scale quickly. Venture capitalists typically gain equity stakes and influence over company decisions in exchange for their investment. The goal is to eventually exit through an IPO or acquisition, generating returns for investors.
- Public benefit corporations (PBCs) are legally recognized companies that must balance profit-making with creating a positive impact on society or the environment. Stakeholder capitalism is a business approach where companies consider the interests of all stakeholders—employees, customers, communities, and shareholders—rather than focusing solely on maximizing shareholder profits. Both concepts aim to promote ethical and sustainable business practices beyond traditional profit motives. However, legal and practical enforcement of these goals can vary widely.
- Some companies adopt "public benefit language" to signal commitment to social or environmental goals, often to attract impact-focused investors or improve public image. However, their governance structures typically prioritize maximizing financial returns for founders and early investors over broader stakeholder interests. Profit distribution usually favors insiders through preferential shares or control rights, limiting benefits to employees, customers, or communities. True public benefit corporations legally embed stakeholder considerations into decision-making, but many firms use the language without these binding commitments.
- Predatory pricing is a strategy where a company deliberately lowers prices below cost to drive competitors out of the market. Once rivals are weakened or eliminated, the company raises prices to recoup losses and gain monopoly power. Uber used this tactic by heavily subsidizing rides to attract customers and undercut traditional taxi services, making it hard for competitors to survive. This approach relies on deep investor funding to sustain losses temporarily.
- Unit economics refers to the direct revenues and costs associated with a single unit of a product or service, showing whether that unit is profitable. Price compression occurs when dominant competitors lower prices aggressively, reducing the revenue per unit for all players. This makes it harder for startups to cover their costs and achieve sustainable growth. Without healthy unit economics, startups struggle to attract investment and survive long-term.
- Mega-IPOs are extremely large initial public offerings where a private company sells shares to the public for the first time, ofte ...
Counterarguments
- The prolonged private status of companies can allow businesses to mature and stabilize before facing the scrutiny and volatility of public markets, potentially resulting in healthier, more sustainable public companies.
- Regulatory changes and index inclusion rules are subject to oversight and debate, and exceptions may be granted to reflect evolving market realities or to ensure indices remain representative of the broader economy.
- Passive funds and ETFs are designed to track indices, and their exposure to late-stage private companies is a byproduct of index methodology rather than direct venture capital decisions.
- Retail investors still have access to diversified investment vehicles, and fiduciary standards require fund managers to act in the best interests of their clients, potentially mitigating some risks.
- The use of public benefit language and stakeholder capitalism is not inherently deceptive; some companies may genuinely pursue both profit and social good, and there are regulatory frameworks (such as B Corp certification) that provide some accountability.
- Delaying IPOs can protect companies from short-term market pressures and allow them to focus on long-term growth and innovation.
- Predatory pricing is subject to antitrust laws and regulatory scrutiny, and not all comp ...
Get access to the context and additional materials
Career Insights and Future Vision: Identifying Trends, Embracing Risks, and Leveraging Data Science, as Shown by Bill Maris's Google Ventures Journey
Recognizing Transformative Tech Shifts via Emerging Infrastructure Offers Advantage For Early Founders and Investors
Office Closet Web Servers Led To Centralized Data Centers and Web Hosting Opportunities
Bill Maris recalls looking in a Wall Street office closet and seeing a server, which housed the company’s email and websites. Realizing that if one company could keep their digital assets in a closet, he could multiply that concept and host many for various clients, Maris foresaw the rise of centralized data centers and web hosting.
Abandoning Wall Street For Internet Infrastructure Enabled Founding a Successful Bootstrapped Company
Inspired by this insight, Maris left his Wall Street job immediately, despite the comfort and security it offered. Using his credit cards, he moved to Vermont and founded a web hosting and data center business out of his apartment in 1997, laying the groundwork for his entrepreneurial career.
Perceiving Trends Often Seems Insane, Requiring Founders and Investors to Trust Their Vision Despite Opposing Pressures
Maris emphasizes that identifying transformative trends can often seem “insane,” likening the experience to glimpsing the future through a keyhole. He illustrates this with an anecdote about a man live-streaming a presidential inauguration on a laptop before camera phones became ubiquitous, demonstrating how trendsetters often appear out of step with their times. For Maris and other entrepreneurs, trusting their future-oriented vision—even amid skepticism—proved essential.
Entrepreneurial Resilience and Willingness to Face Challenges Distinguish Successful Founders From Those Intimidated by Obstacles
Operating Apartment Data Center With Cooling, Leak, and Risk Challenges
Running his web hosting business from a small Vermont apartment, Maris faced extreme conditions. The servers emitted so much heat that he would open windows, causing the room to become freezing cold—glasses of water iced over by noon and he slept with a rug for warmth.
Risky Decision to Repair Leaking Roof In Thunderstorm to Protect Business
When a thunderstorm caused the apartment roof to leak onto his servers, Maris’s landlord dismissed the issue. Not deterred, Maris bought tar and a mop, then ascended onto the roof during lightning and rain to patch the leak himself, prioritizing the business above his own safety.
Perseverance In Incorrectly Tarring a Roof Reflects Successful Founders
Despite tarring himself into a corner on the roof, Maris managed to save his servers, humorously noting his shoes are still stuck on that Vermont rooftop. This willingness to do whatever it takes and embrace risk—even when faced with adversity—is, according to Maris, what distinguishes founders who ultimately succeed.
Data Science and Machine Learning Enhance Venture Investment Portfolio Construction Over Subjective Decisions
Google Ventures Uses Historical Investment Data and Machine Learning to Optimize Portfolio Composition and Fund Sizing Strategies
At Google Ventures, Maris and his team gathered as much venture data as possible and used what was then called “machine learning” (avoiding the term “AI” as Google deemed it too speculative) to construct optimal investment portfolios. By running millions of simulations and back-testing strategies, they determined how best to compose their funds and size them for maximum return.
Google Avoided Ai In Venture Decisions, Reframed As "Machine Learning"
Due to cultural reluctance within Google, Maris recounts, “AI” had to be rebranded as “machine learning” to avoid scaring people, even as the technology powered their data-driven investment process.
Empirical Results: Data-Driven Portfolios at Google Ventures Outperformed Other Venture Firms (2009-2018), Validating Computer Science in Venture Capital Decisions
From 2009 to 2018, Google Ventures portfolios constructed using machine learning achieved approximately 4.1x returns, placing them in the top decile among venture funds and outpacing most contemporaries. Maris credits this direct application of computer science and analytics as the major driver of their performance.
Computer Science and Analytical Rigor Enhance Outcomes Across Industries, Forming a Reliable Investment Thesis
Applying Computer Science to Venture Capital Portfolio Construction Enhances Outcomes
Maris asserts that applying computer science to the right problem at the right time produces the right results, a principle that held true at Google Ventures.
Betting Against Computational Approaches Is a Losing Strategy
Having witnessed many industries, Maris ...
Here’s what you’ll find in our full summary
Career Insights and Future Vision: Identifying Trends, Embracing Risks, and Leveraging Data Science, as Shown by Bill Maris's Google Ventures Journey
Additional Materials
Counterarguments
- The success of Google Ventures and Section 32 may be influenced by factors beyond data-driven approaches, such as access to exclusive deal flow, brand reputation, and network effects, which are not solely attributable to machine learning or computer science.
- Smaller, concentrated funds can also carry higher risk due to less diversification, and their outperformance may not be universally replicable across all market conditions or for all managers.
- The narrative of entrepreneurial resilience and risk-taking, such as repairing a roof during a thunderstorm, may romanticize potentially unsafe or unsustainable behaviors that are not advisable for all founders.
- Machine learning and data-driven strategies in venture capital are dependent on the quality and relevance of available data, which can be limited or biased, especially in early-stage investing where historical data is sparse.
- The rebranding of “AI” as “machine learning” to avoid cultural resistance highlights that organizational ...
Actionables
- you can run a personal resilience audit by listing recent situations where you faced setbacks or discomfort, then noting how you responded and brainstorming one bolder action you could have taken to protect your goals or values, helping you build a habit of perseverance and decisive action in the face of adversity.
- a practical way to experiment with data-driven decision-making is to track a simple recurring choice in your life (like meal planning, budgeting, or exercise routines), collect basic data for a month, and use a free spreadsheet to analyze patterns and adjust your approach based on what the numbers show, rather than relying on gut feeling.
- you can test the impact of focusing ...
Get access to the context and additional materials
1-Page Summary- Why Smaller Venture Capital Funds ($500m-$750M) Outperform Larger Funds: Data Analysis of Returns and Structural Incentives
- Ai Shift: From Large Models to Infrastructure and Physics Engines
- Deep Tech Investment: Life Sciences, Computational Biology, Ai-driven Healthcare
- Ethical Concerns in Venture Capital: Prolonged Private Status of Late-Stage Companies Forcing Retail Investors Into Overpriced Positions and Misaligned Incentives Among GPs, Entrepreneurs, and LPs
- Career Insights and Future Vision: Identifying Trends, Embracing Risks, and Leveraging Data Science, as Shown by Bill Maris's Google Ventures Journey
Create Summaries for anything on the web
Download the Shortform Chrome extension for your browser
