A recent discussion on the “a16z Podcast” offered a striking and somewhat provocative analogy regarding the trajectory of artificial intelligence development. Ben Scharfstein, who serves as the head of product for enterprise applications at Scale AI, articulated what he believes to be one of his most controversial opinions: that the core business model of leading AI laboratories such as OpenAI or Meta AI bears far greater similarity to the structure of Hollywood content studios than to traditional software companies. In his view, these organizations should not be primarily conceptualized as firms pursuing the steady, incremental growth model typical of enterprise software, but rather as creative and risk-driven institutions more akin to film studios that make massive upfront investments in hopes of achieving cultural and financial breakthroughs.
In elaborating on this point, Scharfstein explained that foundational AI labs operate in a manner comparable to renowned movie producers. He described the process as one in which companies allocate vast sums of capital to the creation of “blockbuster” projects, in this case the large language models, which—like cinematic hits—must generate substantial returns within a relatively narrow timeframe before becoming outdated. To illustrate, he likened their approach directly to Marvel Studios, a powerhouse in modern entertainment whose billion-dollar spectacles, such as the famous superhero films, are engineered both as immediate attention-grabbers and as franchise-building vehicles. These films, although immensely popular upon release, often have a finite period in which they dominate cultural discourse before waning in relevance. Similarly, AI models may be enormously impactful at launch but begin to lose novelty once newer technologies appear.
Scharfstein noted that this pattern does not necessarily indicate weakness, but rather demonstrates the cyclical rhythm of innovation that revolves around major releases followed by sequels, extensions, or adaptations. Just as Marvel can extend the value of a successful film into spin-offs, sequels, or cross-media adaptations such as video games, AI companies can repurpose leading models into iterative products, updated frameworks, or domain-specific applications. Although such adaptations may begin to resemble software development in the traditional sense, the essence of the model remains more comparable to a creative studio betting heavily on cultural and technological relevance.
Scale AI itself occupies a distinctive place within this ecosystem. The company, which has become widely recognized for its expertise in data annotation services supporting top technology firms like Google and Meta, also builds custom enterprise platforms powered by generative AI. According to the company’s publicly available information, Scale AI deploys leading models such as OpenAI’s GPT-4, Cohere’s Command, and Meta’s Llama 2 to deliver solutions tailored for corporate clients. Scale AI recently attracted a $14.3 billion strategic investment from Meta, underscoring its central role as an enabler of frontier AI systems. Notably, OpenAI has reported revenues climbing to an annualized $10 billion less than three years after launching ChatGPT, demonstrating how impactful and lucrative a single “blockbuster” AI release can be.
The company’s operational reality, however, also reflects the volatility and high stakes implicit in the model Scharfstein described. Business Insider obtained insight into Scale AI’s extensive collaborations with nearly every leading LLM producer, including Meta, OpenAI, and Google’s DeepMind, for AI training purposes. These partnerships reveal the interconnectedness of major actors in the field but also expose the dependence on continuous, large-scale experimentation. This rapid pace has not been without consequences: just last month, Scale AI implemented significant restructuring measures, reducing its workforce of 1,400 employees by approximately 14%. Around 200 staff members in the generative AI division were affected. Internal communications from interim CEO Jason Droege acknowledged that the company had expanded its generative AI operations “too quickly” during the past year, in effect straining resources and neglecting other divisions, including public-sector projects.
Taken together, Scharfstein’s argument highlights an evolving understanding of how artificial intelligence companies should be conceptualized. Rather than being pigeonholed into the mold of conventional software firms, they may be better compared to cultural producers driven by blockbuster releases—ventures that must sustain both public fascination and revenue streams through continuous reinvention. His Marvel comparison effectively communicates the precarious balance between innovation and obsolescence, illustrating the immense risks and potential rewards characterizing the nascent AI industry.
Sourse: https://www.businessinsider.com/scaleai-exec-says-ai-labs-movie-studios-marvel-ben-scharfstein-2025-8
