Summary Bullets:
• OpenAI’s latest fundraising round puts it at a $150 billion valuation.
• The startup recently released the OpenAI o1 platform, designed to solve complex tasks in science, coding, and math.
The world’s most high-profile technology startup has been in the news again, this time because of a new funding round that has increased its valuation to $150 billion. Microsoft, SoftBank, and Nvidia joined the fundraising alongside Thrive Capital and Khosla Ventures. MGX, the Abu Dhabi (the UAE) investment firm that has also been in the news lately thanks to its investment in data centers (please see: Microsoft Joins BlackRock in $100-billion Investment Drive to Fund Data Center Infrastructure Projects, September 20, 2024) also contributed.
Alongside this news, OpenAI also announced plans to move to a full-blown for-profit governance structure to attract more investment. OpenAI switched its governance structure from a non-profit to a “capped profit” vehicle in 2019, after Microsoft, which has so far sunk $13 billion into the startup, became its main investor. The startup is therefore currently split into two parts, with a non-profit arm at the helm and a for-profit subsidiary to raise capital and hire talent, but it is still subject to the direction of the non-profit. The for-profit’s equity structure has caps that limit the maximum financial returns to investors and employees. Because of this unusual structure, investors remain wary of putting money into the startup, whose insatiable hunger for computing infrastructure to train its generative AI (genAI) models makes it reliant on investors with deep pockets. The recent changes are therefore aimed at putting financial benefactors at ease.
However, all these changes have led to increasing turmoil at the helm of OpenAI, and after CTO Mira Murati unexpectedly abandoned the company last month, a fresh onslaught of senior departures took place. Murati was the brains behind the launch of ChatGPT as a standalone product driven by the underlying large language model (LLM) technology. She was also instrumental in salvaging the company during a dizzying week of leadership changes in November 2023 (please see: Microsoft Hires Former OpenAI CEO to Head New AI Research Unit, November 20, 2024). Her sudden departure was received with consternation and was followed by the resignations of CRO Bob McGrew and VP of Research Barret Zoph. Only two out of eleven founders remain active in the company now, suggesting the wounds made manifest during last year’s crisis are yet to heal.
The startup has been mired in controversy regarding safety (or lack thereof) in developing its AI systems. It launched its safety and security committee to make recommendations on critical safety and security decisions in May 2024, but even this move was greeted with skepticism, resulting in high-profile CEO Sam Altman being forced to announce he was quitting the committee to ensure its independence. A former senior employee at OpenAI stated the company is chasing “shiny products” and neglecting safety. Even its most recent release, OpenAI o1, was impacted by rumors about safety. These new series of models are designed to “reason” through complex tasks and solve harder problems than previous AI systems in science, coding, and math. The technology is closer to a more advanced type of LLM aimed at mimicking more deliberate and analytical thinking.
OpenAI needs to be careful not to lose sight of safety and other ethical considerations in the frantic attempt to retain its first-mover advantage against competitors such as Google, Amazon, Anthropic, and Meta. The transition to a full for-profit structure, a requisite to continue receiving much-needed cash, adds another layer of complexity to its competitive strategy in what is already a ferocious landscape, genAI technologies. Rivals have kept busy in the meantime, with Meta launching its first multimodal platform, Llama 3.2, which can use a combination of images and text prompts, last month. And we can expect Google to come up with its own AI releases before the end of 2024. Unless of course, OpenAI surprises everyone and decides to unveil GPT-5.