From research to revenue: A sliding scale for AI startups

The artificial intelligence industry is currently struggling to distinguish between labs building commercial products and those focused purely on research. It is difficult to determine which AI companies are genuinely attempting to generate revenue. To address this, a new five-level scale measures the commercial ambition of companies building a foundation model—defined as a large-scale AI model trained on vast amounts of data. Under this framework, the distinction between Level 1 (pure research) and Level 5 (significant revenue generation) becomes clear. Players like OpenAI, Anthropic, and Gemini operate at Level 5, meaning they are already generating significant revenue. The scale is designed to measure a company’s commercial ambition, helping observers track where these organizations aim to land.

For newer startups, commercial focus varies widely. A look at how several labs position themselves on this scale reveals highly divergent paths:

• Safe Superintelligence: Operating as a Level 1 startup, the company has raised $3 billion to focus on research rather than commercial products. However, Ilya Sutskever, the founder of Safe Superintelligence, has noted that the lab could pivot “if timelines turned out to be long, which they might,” or because there is value in powerful AI impacting the world.
• humans&: Positioned as a Level 3 lab, the startup has been vague regarding its plans for monetizable products, though it has hinted at building workplace tools.
• World Labs: Operating as a Level 4 lab, the spatial AI company founded by Fei-Fei Li announced funding of $230 million in 2024. It has since released a world-generating model and a commercial product.

The scale also highlights the instability that can occur when a lab’s commercial trajectory is uncertain. Thinking Machines Lab, which raised a $2 billion seed round, is currently facing internal instability. Nearly half of the executives on the founding team have departed the company, including former CTO and co-founder Barret Zoph. These departures, which also involve other executives associated with founder Mira Murati, occurred amid concerns about the direction of the company. This internal friction has led observers to analyze whether the startup is operating as a Level 4, Level 2, or Level 3 lab.

Why it matters

The AI industry is experiencing confusion regarding the commercial ambitions of new labs, leading to drama and uncertainty about whether these companies are building products or just conducting research.