AI Labs: Failing to Monetize? The New Profitability Test

The artificial intelligence landscape is currently experiencing a unique moment for companies building their own foundation models. A new generation of industry veterans, having honed their skills at major tech companies, are now venturing out independently. Alongside them are legendary researchers with immense experience, though their commercial aspirations may be less defined. While some of these new labs undoubtedly have the potential to become OpenAI-sized behemoths, others may focus on impactful research without prioritizing immediate commercialization. This creates a fascinating, and increasingly complex, dynamic within the AI industry.

The Monetization Mystery

The proliferation of AI labs has led to a growing difficulty in discerning which entities are genuinely focused on generating revenue. It’s becoming harder to tell who is actually trying to make money. To address this ambiguity, I propose a five-level sliding scale to assess the ambition of any company developing a foundation model. This scale isn’t about measuring success, but rather the level of commitment to profitability.

• Level 5: We are already generating millions of dollars in revenue daily.
• Level 4: We have a detailed, multi-stage plan to become exceptionally wealthy.
• Level 3: We have several promising product ideas that will be unveiled in due course.
• Level 2: We have the basic outlines of a potential plan.
• Level 1: True fulfillment comes from self-acceptance.

Established players like OpenAI, Anthropic, and Gemini comfortably reside at Level 5. The scale becomes more intriguing when examining the newer generation of labs, brimming with ambitious ideas but possessing plans that are often harder to decipher. Crucially, these labs generally have the freedom to choose their desired level. The current influx of investment in AI means that business plans are rarely scrutinized, even for research-focused projects. A less aggressive pursuit of wealth, perhaps remaining at Level 2, might even lead to a more fulfilling existence than striving for Level 5.

The Confusion and Its Roots

The core issue lies in the lack of clarity regarding an AI lab’s position on this scale. Much of the recent industry debate stems from this confusion. The anxiety surrounding OpenAI’s transition from a non-profit organization arose because the lab spent years operating at Level 1 before abruptly jumping to Level 5. Conversely, Meta’s initial AI research could be argued to have been firmly at Level 2, despite the company’s ultimate ambition of reaching Level 4.

A Look at Contemporary AI Labs

Let's examine four prominent AI labs and assess their current standing on the profitability scale.

Humans&

Humans& garnered significant attention this week and served as a key inspiration for developing this scale. The founders present a compelling vision for the next generation of AI models, shifting the focus from simply scaling laws to prioritizing communication and coordination tools. However, despite the positive press, Humans& has remained hesitant to articulate how this vision will translate into monetizable products. They intend to build products, but are unwilling to commit to specifics. Their stated goal is to create an AI-powered workplace tool, potentially replacing platforms like Slack, Jira, and Google Docs, while fundamentally redefining how these tools operate. It’s a complex concept, and even experts are still trying to fully grasp its implications. Based on this, I would place Humans& at Level 3.

Thinking Machines Lab (TML)

Rating TML is particularly challenging! The appointment of Mira Murati, former CTO and project lead for ChatGPT, and a subsequent $2 billion seed round strongly suggests a well-defined roadmap. Murati doesn’t strike me as someone who enters ventures without a clear plan. Therefore, entering 2026, I would have confidently placed TML at Level 4.

However, recent events have cast doubt on this assessment. The departure of CTO and co-founder Barret Zoph, coupled with the simultaneous exit of at least five other employees, has raised concerns about the company’s direction. Many cited anxieties regarding the viability of the initial plan. Within just one year, nearly half of TML’s founding executive team has departed. This suggests that the initial plan, once considered robust, may have proven less solid than anticipated. In terms of the scale, they may have aspired to Level 4 but realized they were closer to Level 2 or 3. While insufficient evidence currently exists to warrant a downgrade, the situation is rapidly evolving.

World Labs

Fei-Fei Li is a highly respected figure in AI research, renowned for establishing the ImageNet challenge, which spurred the development of contemporary deep learning techniques. She currently holds a Sequoia-endowed chair at Stanford and co-directs two AI labs. Her accomplishments are numerous, and she could easily dedicate the remainder of her career to receiving accolades. Her book is also highly recommended!

When Li announced a $230 million funding round for spatial AI company World Labs in 2024, it initially appeared to be a Level 2 or lower endeavor. However, the AI landscape changes rapidly. Over the past year, World Labs has released both a full world-generating model and a commercialized product built upon it. Simultaneously, demand for world-modeling capabilities has emerged from the video game and special effects industries – and no other major lab has developed a comparable solution. This positions World Labs as a Level 4 company, potentially poised to reach Level 5 soon.

Safe Superintelligence (SSI)

Founded by former OpenAI chief scientist Ilya Sutskever, Safe Superintelligence (SSI) exemplifies a classic Level 1 startup. Sutskever has actively shielded SSI from commercial pressures, even declining an acquisition attempt from Meta. There are no defined product cycles, and aside from the ongoing development of a superintelligent foundation model, there doesn’t appear to be a concrete product. Despite this, he secured $3 billion in funding! Sutskever has consistently prioritized the scientific aspects of AI over business considerations, and all indications suggest that SSI remains fundamentally a scientific project.

However, the AI world is dynamic, and it would be premature to dismiss SSI’s potential commercial viability entirely. During a recent appearance on the Dwarkesh podcast, Sutskever outlined two scenarios that could prompt a shift in SSI’s strategy: “if timelines turned out to be long, which they might,” or “because there is a lot of value in the best and most powerful AI being out there impacting the world.” In essence, if the research progresses exceptionally well or encounters significant challenges, SSI could rapidly ascend several levels on the scale.

The Future of AI Monetization

The AI landscape is still in its nascent stages. The ability to translate groundbreaking research into sustainable, profitable businesses remains a significant hurdle for many labs. The five-level scale provides a useful framework for understanding the varying ambitions and priorities within the industry. As the field matures, we can expect to see greater clarity regarding the monetization strategies of these labs, and a more accurate assessment of their positions on the profitability spectrum. The coming years will be crucial in determining which AI labs will thrive, and which will remain focused on the pursuit of knowledge for its own sake. The pressure to monetize will undoubtedly increase, but the most successful companies will likely be those that strike a balance between scientific innovation and commercial viability. The future of AI isn't just about building powerful models; it's about building sustainable businesses around them.

Key Takeaway: The AI industry is currently grappling with a "profitability test," where the ambition to monetize is as important as actual revenue generation. Understanding where each lab falls on this spectrum is crucial for navigating the evolving landscape.

Further Reading: Stay updated on the latest AI developments with resources from GearTech and other leading technology publications.