The AI industry's latest identity crisis just got a framework. As foundation model startups raise billions without clear revenue strategies, TechCrunch introduced a five-level scale measuring commercial ambition - not success. The timing couldn't be sharper. With Humans& raising $480 million while staying vague on products, Thinking Machines Lab bleeding executives, and Safe Superintelligence turning down Meta's acquisition offer, the question isn't who's making money. It's who's even trying.
The AI gold rush has created a peculiar problem - it's getting impossible to tell which labs actually want to make money. Veterans from OpenAI, Google, and Meta are launching foundation model startups with billion-dollar war chests and zero revenue pressure. Investors are so eager to fund anything AI-adjacent that business plans have become optional.
TechCrunch just proposed a solution: a five-level commercialization scale measuring ambition rather than actual profits. Level 5 companies like OpenAI and Anthropic mint millions daily. Level 1 labs treat "true wealth" as self-actualization. The middle tiers - where most new startups land - reveal the industry's existential confusion about whether AI research should prioritize science or shareholders.
The scale arrives as several high-profile labs navigate this tension in real time. Humans&, which raised $480 million this week, earned a Level 3 rating for having "many promising product ideas" without committing to specifics. The startup floated vague plans for AI workplace tools replacing Slack and Google Docs, but observers remain puzzled about execution details.
"It's my job to know what this stuff means, and I'm still pretty confused," TechCrunch's Russell Brandom wrote, capturing the industry's broader bewilderment.
Thinking Machines Lab's trajectory tells a messier story. Former OpenAI CTO Mira Murati's $2 billion seed round suggested a Level 4 operation with detailed commercialization plans. Then CTO Barret Zoph and five other executives departed within two weeks, citing concerns about company direction. Nearly half the founding team is now gone just one year in.
The exodus signals a gap between aspirations and reality - what looked like Level 4 ambition may have been Level 2 or 3 preparation. Internal turmoil aside, TechCrunch stopped short of downgrading the lab's rating, noting there "isn't quite enough evidence to justify" it yet.
World Labs represents the scale's success case. When AI pioneer Fei-Fei Li raised $230 million in 2024 for spatial AI research, observers pegged it as Level 2 - academic ambitions with loose commercial contours. Li, who established the ImageNet challenge that launched modern deep learning, could coast on awards and Stanford professorships indefinitely.
But World Labs shipped both a world-generating model and commercial product Marble within 18 months, capturing demand from gaming and special effects industries before major labs could compete. The startup now operates at Level 4, possibly approaching Level 5 as revenue materializes.
Safe Superintelligence occupies the opposite extreme. Former OpenAI chief scientist Ilya Sutskever raised $3 billion while actively rejecting commercial pressures. Meta attempted to acquire SSI but Sutskever refused, keeping the lab focused purely on superintelligent AI research without product cycles or revenue targets.
The Level 1 classification fits SSI's stated mission, but Sutskever hinted at flexibility during a recent Dwarkesh Patel interview. SSI might pivot commercially "if timelines turned out to be long" or if "there is a lot of value in the best and most powerful AI being out there impacting the world." Translation: if research progresses too slowly or succeeds too spectacularly, expect sudden monetization.
The scale's deeper insight concerns AI funding dynamics. There's so much capital chasing foundation models that entrepreneurs can choose their commercialization level without investor pushback. Research-focused founders can operate at Level 2 indefinitely while business-minded operators jump straight to Level 4. Both approaches attract billions in venture backing.
This creates strategic confusion. OpenAI spent years at Level 1 as a nonprofit before leaping to Level 5 practically overnight, sparking ongoing controversy about its restructuring. Meta's early AI research operated at Level 2 when leadership wanted Level 4 urgency, contributing to internal friction.
The framework also highlights talent migration patterns. Industry veterans launching new labs often bring Level 5 experience but Level 2 or 3 ambitions, creating friction when team members expect aggressive commercialization. Thinking Machines Lab's executive departures may reflect exactly this disconnect.
For investors, the scale offers diagnostic value. A Level 1 or 2 lab isn't necessarily a bad bet - Google funded DeepMind for years without demanding immediate revenue. But mismatched expectations between founders and backers can torpedo promising startups faster than technical failures.
The timing matters because foundation model economics are shifting. Training costs continue rising while API pricing faces downward pressure from competition. Labs that haven't mapped commercialization strategies may find themselves squeezed as capital gets more selective. Even OpenAI, firmly at Level 5, faces questions about path to profitability given infrastructure expenses.
What's unclear is whether this abundance of research-focused labs benefits the industry long-term. Multiple well-funded teams pursuing pure science could accelerate breakthroughs. Or it could create an overhang of non-commercial technology that never reaches users, wasting resources that might have funded practical applications.
The scale doesn't judge which level is "correct" - just makes motivations transparent. A Level 1 lab run by Ilya Sutskever might deliver more value than a dozen Level 4 startups chasing enterprise sales. But knowing where labs stand helps everyone set appropriate expectations.
The commercialization scale crystallizes AI's current paradox - too much money chasing too few business models. As foundation model startups navigate between pure research and profit-seeking, transparency about intentions matters more than revenue milestones. Whether Humans& clarifies its product roadmap, Thinking Machines Lab stabilizes its team, or Safe Superintelligence maintains its research purity, the industry needs clearer signals about who's building businesses versus advancing science. With capital still abundant but patience finite, expect more labs to pick their level and commit.
