The video features a live discussion focused on comparing OpenAI and Anthropic, particularly regarding their revenue growth and enterprise adoption. The host welcomes a global audience and introduces Ara Karazian, the chief economist at Ramp, a company that tracks AI spending across thousands of businesses. Ara explains that Ramp leverages unique data from corporate spending to analyze AI adoption trends, including which models companies use and how their AI budgets evolve. This data provides rare insights into the private sector’s AI investments, helping businesses understand market dynamics and adoption patterns.
Ara highlights that while AI spending is growing rapidly—up to 14 times year-over-year for some companies—it still represents a small fraction of overall corporate budgets. The primary use case driving this spending is coding, especially among tech, finance, and professional services sectors. Most business AI spend is on APIs powering backend products or coding agents, with subscriptions playing a smaller role. Ara notes that companies are increasingly experimenting with multiple AI models and platforms to optimize costs and performance, often shifting to cheaper models when budgets are strained.
The discussion also covers the competitive landscape between OpenAI and Anthropic. Ramp’s data shows that Anthropic has recently surpassed OpenAI in new enterprise customer adoption, driven by its focus on technical users and coding agents, as well as expanding into non-technical use cases with products like Cloud Cowork. Despite some negative press, such as the Department of Defense labeling Anthropic a security threat, the company’s growth has accelerated. The conversation touches on Google’s AI efforts, noting that while Google’s Gemini models have significant adoption through free Workspace integration, their paid enterprise adoption lags behind due to less developer-friendly tooling.
Open source AI models and their future viability are also discussed. Ara argues that while open source models will always lag behind frontier models in performance, they have a role in serving less demanding use cases and providing cost-effective alternatives. However, the economics of developing and maintaining open source frontier models are challenging, as the high R&D and infrastructure costs make it difficult to compete with well-funded labs. The conversation emphasizes that AI adoption is dynamic, with companies continuously experimenting across multiple providers and use cases, and that the market is far from reaching a cost-benefit plateau.
In closing, Ara and the host reflect on the evolving AI market, noting that adoption rates are expected to reach 90% among firms within a few years. Future research will focus on measuring the intensity and effectiveness of AI use, cost trends, and productivity impacts. The host encourages viewers to follow Ara’s work and stay tuned for upcoming content, including an interview with Google CEO Sundar Pichai. The session ends with thanks to the audience and a promise of more live streams exploring AI industry insights.
