AI Revolution: Unprecedented Acceleration, Market Infancy, and Shifting Economics - Episode Hero Image

AI Revolution: Unprecedented Acceleration, Market Infancy, and Shifting Economics

The a16z Show · · Listen to Original Episode →
Original Title:

TL;DR

  • AI is accelerating at an unprecedented rate, surpassing previous technology waves and reshaping markets faster than incumbents can react, indicating the market is still in its early stages.
  • The proliferation of AI is enabled by existing internet infrastructure, allowing for rapid global adoption and monetization, unlike previous technologies that required extensive physical build-out.
  • Usage-based pricing, or "tokens by the drink," is becoming standard for AI, offering startups flexibility and enabling cloud providers to monetize advanced intelligence at scale.
  • The cost of AI is collapsing faster than Moore's Law, driven by hyper-deflation in per-unit costs, which is expected to spur further demand and economic optimization.
  • The AI industry will likely mirror the computer industry's structure, with a few "god models" at the top and a cascade of smaller, specialized models proliferating across devices.
  • Open-source AI models are rapidly improving and democratizing knowledge, accelerating expertise spread and potentially challenging the dominance of proprietary, closed-source systems.
  • Regulatory fragmentation, particularly at the state level in the US, poses a risk to AI development by creating complexity and potential liability for developers, hindering innovation.
  • Venture portfolios are designed to bet on multiple, conflicting AI strategies simultaneously, acknowledging the inherent uncertainty and the potential for diverse outcomes in this transformative field.

Deep Dive

Marc Andreessen views Artificial Intelligence as the most significant technological revolution of his lifetime, surpassing even the internet in its potential impact. He contends that despite rapid adoption and impressive revenue growth in AI companies, the market is still in its nascent stages, with current products likely to evolve dramatically over the next decade. This accelerated pace of development, driven by rapidly falling model costs and swift capability gains, is rapidly reshaping industries and competitive landscapes, outpacing the ability of incumbents to adapt.

The AI revolution is characterized by a fundamental shift in how intelligence is commoditized and delivered. Unlike previous technological waves that required extensive physical infrastructure build-out, AI can be "downloaded" and accessed globally via existing internet infrastructure, enabling unprecedented adoption rates. This has led to a proliferation of AI applications, with consumers and businesses alike increasingly integrating AI into their daily lives and operations. The economic model for AI is also transforming, with a move towards usage-based and value-based pricing, which Andreessen sees as more creative and potentially lucrative than traditional SaaS models. This pricing flexibility, coupled with the rapid deflation of per-unit AI costs, suggests substantial future revenue growth for AI-centric companies.

The implications of this AI wave are far-reaching. The rapid advancement and democratization of AI capabilities, exemplified by open-source models achieving parity with proprietary ones, suggest that innovation will continue to accelerate. This dynamic creates opportunities for both startups and incumbents, though the competitive landscape is fluid. Andreessen also highlights the geopolitical dimension, noting that the AI race is primarily between the US and China, underscoring the importance of domestic innovation and policy frameworks that foster, rather than hinder, technological progress. The recent surge in AI adoption, despite public anxieties, demonstrates a divergence between stated fears and actual behavior, suggesting that AI, like previous transformative technologies, will ultimately become an indispensable tool.

Action Items

  • Analyze AI adoption trends: Track usage data for 3-5 AI applications to understand real-world behavior versus stated opinions.
  • Evaluate AI pricing models: For 2-3 internal projects, compare usage-based pricing against value-based pricing to optimize revenue.
  • Investigate AI chip alternatives: Research 3-5 emerging AI chip startups to assess their potential impact on hardware costs.
  • Draft open-source AI contribution guidelines: Define standards for internal teams engaging with open-source AI models, considering potential downstream liabilities.
  • Measure AI's impact on productivity: For 3-5 business processes, quantify AI's contribution to efficiency gains and cost reduction.

Related Episodes

Key Quotes

"AI is moving faster than any technology wave before it and the rules are being written in real time for decades new platforms followed a familiar arc build infrastructure attract developers capture the value ai is breaking that pattern models are improving weekly costs are collapsing and entire markets are being rebuilt before incumbents can react what looks stable today may not exist a year from now."

Marc Andreessen highlights the unprecedented speed of AI development, contrasting it with the more predictable evolution of previous technologies like the internet. This rapid pace, characterized by weekly model improvements and collapsing costs, suggests a fundamental shift in how markets and established companies operate, implying that current market structures are highly susceptible to disruption.

"The reason this is so big i mean maybe obvious to folks at this point but i'll just go through it quickly so if you kind of trace all the way back to the 1930s there was actually a debate among the people who were actually invented the computer and it was the sort of debate between whether they kind of understood the theory of computation before they actually built the things and they had this big debate over whether the computer should be basically built in the image of what at the time were called adding machines or calculating machines where you think of sort of essentially cash registers ibm is actually the successor company to the national cash register company of america and that was of course the path that the industry took which was building these kind of hyper literal mathematical machines you know that could execute mathematical operations billions of times per second but of course had no ability to kind of deal with human beings the way humans like to be dealt with and so you know couldn't understand human speech human language and so forth and that's the computer industry that got built over the last 80 years and that's the computer industry and this built all the wealth and financial returns in the computer industry over the last 80 years you know across all the generations of computers from mainframes through to smartphones but they knew at the time they knew in the 30s actually they understood the basic structure of the human brain they understood that they had a theory sort of human cognition and actually they had the theory of neural networks so they had this theory that there's actually the first neural network paper economic paper was published in 1943 which was over 80 years ago which is extremely amazing"

Marc Andreessen explains that the current AI revolution is significant because it represents a divergence from the historical path of computing. He contrasts the early debate between building computers as literal calculating machines versus those designed on the model of the human brain (neural networks). Andreessen emphasizes that the latter, though a "path not taken" for decades, is now being realized, suggesting a fundamental shift in computational capability.

"if you run a survey or a poll of what for example american voters think about ai it's just like they're all in a total panic it's like oh my god this is terrible this is awful it's going to kill all the jobs it's going to ruin everything if you watch the revealed preferences they're all using ai ai is moving faster than any technology wave before it and the rules are being written in real time"

Marc Andreessen points out a significant disconnect between public perception and actual behavior regarding AI. He notes that while surveys might show widespread panic about AI's potential negative impacts, people's actions (revealed preferences) demonstrate widespread adoption and use of AI tools. Andreessen suggests that this divergence indicates that AI's rapid integration is outpacing public discourse and the establishment of clear societal rules.

"you know and and then the consumer the reason i go through that is the consumer ai products could basically deploy to all of those people basically as quickly as they want to adopt right um and so it sort of the internet's the carrier wave for ai to be able to proliferate at a kind of white speed uh in into the broad base of the global population and and that's a let's just say that's a potential rate of proliferation of a new technology that's just far faster than has ever been possible before like well you know like you couldn't download electricity right you couldn't download you know you couldn't download into plumbing um you know you couldn't download television but you can download ai"

Marc Andreessen explains that the existing global internet infrastructure acts as a powerful "carrier wave" for AI adoption. He argues that unlike previous foundational technologies like electricity or plumbing, AI can be distributed digitally and rapidly to billions of users worldwide. This digital distribution, facilitated by the widespread availability of the internet and smartphones, allows for an unprecedented rate of technological proliferation.

"the core business model right is is actually quite quite interesting the core business model is is basically tokens by the drink right and so it's a sort of tokens of intelligence uh you know per dollar and oh and then by the way this is the other fun thing is if you look at what's happening with uh the price of ai the price of ai is falling much faster than moore's law and when that right to go through that in great detail but basically like all of the inputs into ai on a per unit basis the costs are collapsing"

Marc Andreessen describes the core business model for AI as "tokens by the drink," implying a pay-as-you-go system for computational intelligence. He further emphasizes that the cost of AI is rapidly decreasing, outpacing even Moore's Law, as the per-unit costs of its underlying inputs are collapsing. This trend suggests a future of increasingly accessible and affordable AI.

"the smartest models will always be at the top but the volume of models will actually be the smaller models that proliferate out and right that's what happened with microchips uh it's what happened with computers which became microchips and then it's what happened with operating systems and with with everything else that we built in software"

Marc Andreessen posits that the future AI industry will likely mirror the structure of the computer industry, with a pyramid of models. He suggests that while a few highly capable "god models" will exist at the top, the vast majority of AI applications will utilize smaller, more specialized models. Andreessen draws parallels to the proliferation of microchips, computers, and operating systems, indicating a trend towards widespread, accessible AI deployment.

Resources

External Resources

Books

  • "Rise of the Machines" - Mentioned as a historical reference for the debate surrounding computer development in the 1930s.

Articles & Papers

  • "The first neural network paper" (1943) - Referenced as an early academic exploration of neural networks.

Videos & Documentaries

  • Interview with McCulloch on YouTube - Mentioned as a historical artifact of early AI discussions.
  • YouTube - Mentioned as a platform where historical interviews are available.

People

  • McCulloch - Co-author of the first neural network paper and featured in a YouTube interview.
  • Pitts - Co-author of the first neural network paper.
  • Elon - Mentioned in relation to the possibility of routine trips to Mars.
  • Marc Andreessen - Featured in the AMA, discussing AI timelines, US vs. China, and the price of AI.
  • Ben - Mentioned as a partner with whom discussions and disagreements occur.
  • Mario Draghi - Former Prime Minister of Italy, author of a report on European competitiveness.
  • Elias Kiver - Funded to start a new foundation model company after leaving OpenAI.
  • Mira Murati - Funded to start a new foundation model company after leaving OpenAI.
  • Fei Fei Li - Funded to start a world model foundation company out of Stanford.

Organizations & Institutions

  • OpenAI - Mentioned as a provider of AI products and a source of talent for new ventures.
  • Anthropic - Mentioned as a provider of AI products and a new incumbent in the AI space.
  • Google - Mentioned as a provider of AI products (Gemini) and a major player in the cloud war.
  • Microsoft - Mentioned as a provider of AI products and a major player in the cloud war.
  • AWS (Amazon Web Services) - Mentioned in relation to GPU usage and cloud services.
  • IBM - Mentioned as a successor company to the National Cash Register Company and a historical path in computer development.
  • National Cash Register Company of America - Predecessor to IBM, associated with the development of adding machines.
  • Stanford - Mentioned as the institution from which Fei Fei Li departed to start a company.
  • AMD - Mentioned as a competitor to Nvidia in the AI chip market.
  • Huawei - Mentioned as a main chip company in China.
  • Alibaba - Mentioned as the producer of the AI model "Quan".
  • Tencent - Mentioned as a primary company doing work in AI in China.
  • Baidu - Mentioned as a primary company doing work in AI in China.
  • ByteDance - Mentioned as a primary company doing work in AI in China.
  • DeepSeek - Mentioned as a Chinese AI model and a significant release in the open-source AI space.
  • Moonshot - The company that produces the "Kimi" AI model.
  • European Union (EU) - Mentioned for passing the "AI Act," which is described as hindering AI development.
  • Apple - Mentioned as a company not launching leading-edge AI capabilities in Europe due to the EU AI Act.
  • Meta - Mentioned as a company not launching leading-edge AI capabilities in Europe due to the EU AI Act.
  • University research lab - Contrasted with hedge funds as a source of AI models.
  • Hedge fund - Mentioned as the origin of the DeepSeek AI model.
  • CCCP (Chinese Communist Party) - Mentioned in relation to China's governance model and employment concerns.
  • Johnson Administration - Mentioned in relation to a 1960s AI panic and a committee formed to address technological change.
  • A16z - The venture capital firm hosting the discussion.

Tools & Software

  • ChatGPT - Mentioned as an AI product available for public use and for its role in everyday life.
  • Grok - Mentioned as an AI product available for public use.
  • Gemini - Mentioned as an AI product available for public use.
  • Sora - Mentioned as a state-of-the-art video generation tool.
  • VEO - Mentioned as a state-of-the-art video generation tool.
  • Suno - Mentioned as a state-of-the-art music generation tool.
  • Ido - Mentioned as a state-of-the-art music generation tool.
  • GPU (Graphical Processing Unit) - Discussed as the hardware currently used for AI, its historical context, and future competition.
  • CPU (Central Processing Unit) - Discussed in contrast to GPUs in the context of personal computer architecture.

Articles & Papers

  • Kimi (Chinese AI model) - Mentioned as a reasoning model with capabilities comparable to GPT-5, released as open source.

Websites & Online Resources

  • a16z.com - Mentioned for disclosures.
  • a16z.substack.com - Mentioned for subscribing to content.

Other Resources

  • AI (Artificial Intelligence) - The primary subject of the discussion, covering its revolution, timelines, economics, and societal impact.
  • Neural Networks - Discussed as a theoretical path for computer development that was not initially taken.
  • Cybernetics - Mentioned as a field that evolved from the study of neural networks.
  • Moore's Law - Referenced in the context of the falling price of AI inputs.
  • Tokens by the drink - Described as the core business model for AI infrastructure, referring to per-unit pricing of intelligence.
  • Usage-based pricing - Discussed as a pricing model for AI services, particularly beneficial for startups.
  • Seat-based pricing - Contrasted with usage-based pricing for AI services.
  • SAS pricing models - Mentioned as a potential replication for AI application pricing.
  • Open source - Discussed in relation to AI models, particularly from China, and its benefits for knowledge proliferation.
  • Closed source - Discussed as an alternative to open source for AI models.
  • Foundation models - Mentioned as a category of AI startups being funded.
  • World models - Mentioned as a type of foundation model company.
  • Robotics - Mentioned as a future area of competition with AI, where China currently has an advantage.
  • Printing press - Referenced as a historical example of a technology causing societal panic.
  • Committee for the Triple Revolution - A group from the 1960s that issued a manifesto similar to current AI concerns.
  • False consciousness - A Marxist explanation for differing opinions and behaviors.
  • Revealed preferences - Observed behavior used to understand people's true actions and thoughts.
  • Surveys, focus groups, polls - Methods used to ask people for their opinions.
  • VR (Virtual Reality) - Mentioned as a potential way to experience Mars without physically traveling there.
  • AI Act (EU) - Legislation in Europe described as detrimental to AI development.
  • GDPR - Mentioned in the context of European regulation that may be unwound.
  • Draghi Report - A report on European competitiveness that highlighted issues with over-regulation.
  • SB 1047 - A California bill modeled after the EU AI Act, which was vetoed.
  • Algorithmic discrimination - A concern addressed by AI regulations.
  • Downstream liability - A concept in SB 1047 that would assign liability to open-source developers.
  • Trillion dollar questions - A term used to describe significant, impactful questions in the AI space.
  • The internet - Discussed as a foundational technology that enabled the rapid proliferation of AI.
  • Mobile broadband - Mentioned as a key component of the internet's infrastructure that enabled AI adoption.
  • Cloud war - The competition between cloud service providers like AWS, Azure, and Google Cloud.
  • AI chips - Discussed in the context of supply, demand, and future availability.
  • Gpus - Mentioned as the current standard for AI chips, with discussion on their limitations and future alternatives.
  • US vs. China - The geopolitical and economic competition between the two countries in the AI space.
  • Geopolitics - Discussed in relation to the US-China AI competition.
  • Policy and regulation - Discussed in the context of federal and state-level AI laws.
  • Federalism - The system of government where states have their own laws.
  • Interstate commerce - The concept of commerce that crosses state lines, relevant to AI regulation.
  • National importance - The scope of AI as a technology affecting the entire nation.
  • Value-based pricing - A pricing strategy focused on the value delivered to the customer.
  • Cost-plus pricing - A pricing strategy based on the cost of production.
  • Marginal productivity - The increase in output resulting from adding one more unit of input, relevant to AI pricing.
  • Human-AI symbiotic relationship - The collaboration between humans and AI, impacting pricing models.
  • High prices - Discussed as potentially beneficial for customers by enabling product improvement.
  • VC (Venture Capital) - Mentioned in the context of investment decisions and the "tragedy of the commons" in policy engagement.
  • LPs (Limited Partners) - Investors in venture capital funds.
  • Founders - Entrepreneurs who start companies.
  • Mars - Mentioned as a potential future destination for human travel.
  • Cryogenics - Discussed in relation to the possibility of preserving individuals for future revival.
  • San Francisco Bay Area - Implied location of the discussion participants.
  • California - Mentioned as a location where AI development is significant and where the speaker is reluctant to leave.
  • Midwesterners - A group characterized by humility or the ability to fake it.
  • Wall Street Journal - Mentioned as a source of news that can remind of investment mistakes.
  • CNBC - Mentioned as a source of news that can remind of investment mistakes.
  • The internet - Mentioned as a platform for communication and information dissemination.
  • The printing press - Referenced as a historical example of a technology that caused societal upheaval.
  • Marxism - Mentioned in relation to historical fears of automation and wealth concentration.
  • Protestantism - Mentioned in relation to the societal changes brought about by the printing press.
  • The 1960s - A period mentioned for AI-related panic.
  • The 1980s - A period mentioned for AI boom-bust cycles.
  • The 2000s - A period mentioned for panic around outsourcing.
  • The 2010s - A period mentioned for panic around robots.
  • The 1930s - A period mentioned for early computer development debates.
  • The 1940s - A period mentioned for early neural network research.
  • The 1960s - A period mentioned for early AI research.
  • The 1970s

---
Handpicked links, AI-assisted summaries. Human judgment, machine efficiency.
This content is a personally curated review and synopsis derived from the original podcast episode.