The OpenAI Board Crisis: Governance Lessons for AI Investors

On 17 November 2023, the board of OpenAI fired Sam Altman as CEO, citing a loss of confidence in his candour. Five days later, he was reinstated, the board had been reconstituted, and the episode had generated more commentary about AI governance than the previous five years of AI development combined. For investors in AI companies and AI infrastructure, the OpenAI crisis was not just a corporate drama — it was a case study in the governance risks that are unique to organisations developing transformative AI systems.

The structural peculiarity of OpenAI — a capped-profit company controlled by a non-profit board, with a mission to develop AI for the benefit of humanity — is not typical of the AI companies that most investors will encounter. But the underlying governance challenge is universal: how do you structure oversight of an organisation whose technology is advancing faster than the frameworks designed to govern it? The OpenAI board had the legal authority to fire the CEO. What it lacked was the operational understanding, the stakeholder management capability, and the institutional legitimacy to exercise that authority without triggering a crisis that nearly destroyed the company.

The lessons for investors are practical. First, governance structure matters as much as technology capability when evaluating AI companies. The voting rights, board composition, and mission alignment of an AI company are not footnotes in the due diligence process — they are central to the risk assessment. Second, key-person risk in AI companies is extreme. The departure of a founding CEO or chief scientist can destroy value in ways that are difficult to model but easy to observe. Third, the relationship between AI safety commitments and commercial imperatives is a genuine tension that boards and investors need to understand and manage.

At Lensbridge, the OpenAI crisis reinforced our preference for infrastructure assets over application-layer companies in the AI space. Infrastructure assets — data centres, networks, power systems — do not have governance structures that can unravel in five days. Their value is embedded in physical assets and long-term contracts, not in the judgment of a single individual. That stability is, in the current environment, a feature rather than a limitation.