Microsoft and OpenAI have set a financial benchmark for defining artificial general intelligence (AGI), a goal typically linked to technological capability by AI researchers.
Under the terms of their partnership, AGI will be considered achieved when OpenAI’s AI systems generate $100 billion in cumulative profits. This decision, revealed through internal documents according to The Information, represents a departure from traditional definitions, which focus on replicating human-level intelligence across diverse tasks.
The profit-focused approach has practical implications for Microsoft’s access to OpenAI’s technologies. The clause ensures that Microsoft retains rights to use OpenAI’s models and infrastructure until AGI is reached, creating a mutually beneficial but complex relationship between the two companies.
A Financial Metric for AGI
AGI, often described as a system capable of performing tasks at or beyond human capability in multiple domains, has long been a subject of theoretical debate. OpenAI’s interpretation ties it explicitly to financial outcomes, addressing investor expectations and providing a tangible milestone.
The agreement benefits Microsoft by prolonging its exclusive access to OpenAI’s technology, embedded in products like Azure and Microsoft 365 Copilot.
“OpenAI’s mission is to ensure that artificial general intelligence (AGI)—by which we mean highly autonomous systems that outperform humans at most economically valuable work—benefits all of humanity. We will attempt to directly build safe and beneficial AGI, but will also consider our mission fulfilled if our work aids others to achieve this outcome,” OpenAI states in Charter. However, the $100 billion threshold reflects a shift towards quantifiable, investor-aligned objectives.
Financial Pressures and Strategic Adjustments
Despite its estimated valuation of $157 billion, OpenAI faces mounting financial challenges. The company projects a $5 billion loss in 2024 and cumulative losses of $44 billion by 2028. Compute expenses for training large AI models like GPT-4 and its upcoming successor account for much of this strain, with annual costs expected to climb to $9.5 billion by 2026.
Partnerships with semiconductor manufacturers TSMC and Broadcom aim to reduce these costs through the development of custom AI chips, slated for release in 2026.
To generate immediate revenue, OpenAI launched ChatGPT Pro in December 2024. Priced at $200 per month, the subscription targets professionals and enterprises, offering enhanced model access, reliability, and developer tools like the Canvas Interface. This initiative underscores OpenAI’s strategy to balance long-term innovation with short-term profitability.
Microsoft’s Independent Strategy
Microsoft, a key investor in OpenAI with $14 billion committed since 2019, has started reducing its reliance on the company’s technology. Proprietary large language models (LLMs) are now integrated into Microsoft 365 Copilot, signaling Microsoft’s intention to develop in-house solutions. This shift reflects broader concerns about resource allocation and rising costs within the partnership.
Disputes over computing power further strain the collaboration. OpenAI depends heavily on Microsoft’s Azure infrastructure for AI training and deployment. OpenAI employees have raised concerns about resource limitations potentially hindering innovation. Microsoft, meanwhile, seeks to avoid overextending its commitments amid growing competition from rivals like Google and Anthropic.
Revisiting the AGI Clause
A key aspect of the partnership is the AGI clause, which mandates to transfer control of AGI to OpenAI’s nonprofit board upon achievement. Initially designed to prevent monopolization, the clause is now under review. Revising it could allow Microsoft continued access to OpenAI’s future innovations, aligning their interests more effectively.
CEO Sam Altman emphasized the importance of investor alignment at The New York Times DealBook Summit, stating, “If someone invests in a competitor, they don’t get access to our roadmap.” This underscores OpenAI’s efforts to balance strategic partnerships with financial autonomy.
Rising Competition in Generative AI
OpenAI’s financial definition of AGI comes amid intense competition in the generative AI space. Google’s Gemini model series, now offering its own thinking model with Gemini 2.0 Flash Thinking, and Anthropic’s Claude series are challenging OpenAI’s dominance. Meanwhile, Elon Musk’s venture xAI is pushing its Grok Ai aggressively, with plans to expand its Colossus supercomputer tenfold.
OpenAI has been responding to the increasing pressure with a multitude of releases during its 12 Days of OpenAI this December, which included the $200/month ChatGPT Pro subscription based on the new o1 model and unveiling the upcoming o3 reasoning model with unprecedented performance.
How tight the AI race has become particularly between Google and OpenAI becomes clear after the surprising release of Google’s Veo 2 AI video generation tool which offers 4K output and overall higher quality. OpenAi launched its long awaited Sora video generator a few days earlier, which after Google’s Veo 2 release already appears somewhat outdated.
