How OpenAI Builds for 800 Million Weekly Users: Model Specialization and Fine-Tuning
How OpenAI Builds for 800 Million Weekly Users: Model Specialization and Fine-Tuning
Podcast53 min 24 sec
Listen to Episode
Note: AI-generated summary based on third-party content. Not financial advice. Read more.
Quick Insights

Investors should be cautious with Opendoor (OPEN), as its high-risk, low-margin business model is a cautionary example of technology applied to a fundamentally challenging industry. The most significant long-term opportunity in AI lies not just in model creators, but in companies possessing unique and valuable datasets that can be leveraged for a competitive advantage. A key emerging theme is the rise of specialized AI models, creating opportunities in companies that serve niche industries like finance or medicine. Furthermore, the most immediate market may be in practical enterprise automation tools that help businesses automate procedural, rules-based work. This shift towards specialized, deterministic AI for business is a more tangible investment thesis than the pursuit of a single, all-powerful AGI.

Detailed Analysis

OpenAI (Private Company)

  • Business Model: OpenAI operates a unique dual model, being both a horizontal platform with its API and a vertical, consumer-facing company with its flagship app, ChatGPT.
    • The company's mission is to create AGI and "distribute the benefits as broadly as possible," which justifies having both a first-party app and a third-party developer platform.
  • Explosive Growth: ChatGPT has achieved 800 million weekly active users, described as "historic" growth, representing roughly 10% of the global population.
    • The user reach from the API is also massive and, at times, has been even larger than ChatGPT's direct reach.
  • Strategic Shift: The company has moved away from the initial industry belief that there would be "one model to rule them all."
    • They now embrace a future with a "proliferation of specialized models" for different use cases, which is viewed as a healthier dynamic for the ecosystem and for OpenAI itself.
  • Competitive Moat: OpenAI's models are described as an "anti-disintermediation technology."
    • It is very difficult for developers to abstract away the underlying model (e.g., GPT-4, GPT-5). Users and developers become familiar with a specific model's personality and capabilities, leading to high "stickiness" and making it difficult to switch to a competitor.
  • Data Strategy: OpenAI is creating powerful data flywheels through its fine-tuning products.
    • The company recognizes that businesses are sitting on "giant treasure troves of data."
    • Its Reinforcement Fine-Tuning API allows companies to use their proprietary data to create state-of-the-art models for specific tasks.
    • OpenAI is piloting programs that offer discounted inference and free training to customers willing to share their valuable data, further enhancing its own models.
  • Open Source Approach: OpenAI has released open-source models and sees the risk of cannibalizing its paid API business as "very low."
    • The strategy is that a "rising tide" in the AI ecosystem ultimately benefits the market leader, OpenAI.
    • They believe the complexity and cost of running inference for their largest, most powerful models at scale creates a natural barrier, even if the model weights were open-sourced.

Takeaways

  • Dominant Market Position: OpenAI's massive user base and sticky platform create a powerful competitive advantage. The dual API and consumer app strategy allows it to capture value across the entire AI ecosystem.
  • Data is the Next Frontier: The focus on fine-tuning with proprietary data is a key evolution. Investors should look for companies (both public and private) that possess unique, valuable datasets that can be leveraged with AI, as this is becoming a primary source of competitive advantage.
  • The "Platform Paradox" is a Strength: By enabling thousands of startups (even potential competitors) to build on its API, OpenAI solidifies its position as the foundational layer of the AI economy, similar to how cloud providers operate. This ecosystem growth is a bullish long-term signal.

Opendoor (OPEN)

  • Business Model Context: The guest, Sherwin Wu, previously worked at Opendoor for six years on the machine learning models that priced homes.
  • Core Challenge: Opendoor's business is fundamentally a "deep technology problem" centered on accurately pricing assets. A single wrong prediction can cost millions.
  • Financial Profile: It is described as a "way lower margin business than OpenAI." The model is highly sensitive to factors like holding costs, market variability, and long-tail risks associated with holding expensive inventory.

Takeaways

  • High-Risk, Low-Margin Model: The discussion highlights the inherent difficulties in Opendoor's iBuying business. Unlike a high-margin software business, it operates on thin spreads ("basis points") and is exposed to significant real-world financial risk with every transaction.
  • Technology vs. Business Model: While Opendoor has sophisticated technology at its core, the underlying business model is fundamentally challenging and capital-intensive. This serves as a cautionary example for investors evaluating tech companies operating in physically-grounded, low-margin industries.

AI Industry Investment Themes

  • The Proliferation of Models: The initial thesis of a single, winner-take-all AGI model is now considered incorrect. The market is evolving towards a diverse ecosystem of many specialized models.
    • Insight: This suggests there is room for multiple winners in the AI model space. Investment opportunities may exist not just in the largest foundational model companies but also in those creating highly specialized or efficient models for niche industries (e.g., coding, medical, finance).
  • Context Engineering over Prompt Engineering: The focus for developers is shifting from simply writing better prompts to "context engineering."
    • Insight: The key is now about providing the AI with the right data and tools at the right time (e.g., through advanced Retrieval-Augmented Generation, or RAG). This creates opportunities for companies building the infrastructure and tools that manage this complex data interaction layer.
  • The Rise of Deterministic Agents: There is a massive market for AI agents that can automate procedural, rules-based work, not just creative or knowledge-based tasks.
    • Insight: OpenAI's "Agent Builder" focuses on creating workflows that follow Standard Operating Procedures (SOPs). This is crucial for enterprise adoption in regulated industries or for functions like customer support, sales, and marketing. Investors should look for AI companies that are building practical, deterministic tools for enterprise automation, as this is likely a larger and more immediate market than fully autonomous agents.
  • Usage-Based Pricing is the Future: The shift to usage-based billing for AI services is described as a "one-way ratchet" and a "huge shift in the industry."
    • Insight: This model aligns cost directly with value and is becoming the standard for AI infrastructure. Companies that can effectively manage, predict, and optimize usage-based costs will have a competitive advantage. This also creates opportunities for companies building billing and financial operations (FinOps) tools specifically for the AI stack.
Ask about this postAnswers are grounded in this post's content.
Episode Description
In this episode, a16z GP Martin Casado sits down with Sherwin Wu, Head of Engineering for the OpenAI Platform, to break down how OpenAI organizes its platform across models, pricing, and infrastructure, and how it is shifting from a single general-purpose model to a portfolio of specialized systems, custom fine-tuning options, and node-based agent workflows. They get into why developers tend to stick with a trusted model family, what builds that trust, and why the industry moved past the idea of one model that can do everything. Sherwin also explains the evolution from prompt engineering to context design and how companies use OpenAI’s fine-tuning and RFT APIs to shape model behavior with their own data. Highlights from the conversation include:  • How OpenAI balances a horizontal API platform with vertical products like ChatGPT • The evolution from Codex to the Composer model • Why usage-based pricing works and where outcome-based pricing breaks • What the Harmonic Labs and Rockset acquisitions added to OpenAI’s agent work • Why the new agent builder is deterministic, node based, and not free roaming   Resources:  Follow Sherwin on X: https://x.com/sherwinwu  Follow Martin on X: https://x.com/martin_casado   Stay Updated: If you enjoyed this episode, be sure to like, subscribe, and share with your friends! Find a16z on X: https://x.com/a16z Find a16z on LinkedIn: https://www.linkedin.com/company/a16z Listen to the a16z Podcast on Spotify: https://open.spotify.com/show/5bC65RDvs3oxnLyqqvkUYX Listen to the a16z Podcast on Apple Podcasts: https://podcasts.apple.com/us/podcast/a16z-podcast/id842818711 Follow our host: https://x.com/eriktorenberg Please note that the content here is for informational purposes only; should NOT be taken as legal, business, tax, or investment advice or be used to evaluate any investment or security; and is not directed at any investors or potential investors in any a16z fund. a16z and its affiliates may maintain investments in the companies discussed. For more details please see http://a16z.com/disclosures Stay Updated: Find a16z on X Find a16z on LinkedIn Listen to the a16z Podcast on Spotify Listen to the a16z Podcast on Apple Podcasts Follow our host: https://twitter.com/eriktorenberg   Please note that the content here is for informational purposes only; should NOT be taken as legal, business, tax, or investment advice or be used to evaluate any investment or security; and is not directed at any investors or potential investors in any a16z fund. a16z and its affiliates may maintain investments in the companies discussed. For more details please see a16z.com/disclosures. Hosted by Simplecast, an AdsWizz company. See pcm.adswizz.com for information about our collection and use of personal data for advertising.
About a16z Podcast
a16z Podcast

a16z Podcast

By Andreessen Horowitz

The a16z Podcast discusses tech and culture trends, news, and the future – especially as ‘software eats the world’. It features industry experts, business leaders, and other interesting thinkers and voices from around the world. This podcast is produced by Andreessen Horowitz (aka “a16z”), a Silicon Valley-based venture capital firm. Multiple episodes are released every week; visit a16z.com for more details and to sign up for our newsletters and other content as well!