Stability AI: Challenges and Outlook for the One-Time King of Text-to-Image Generation

Management turmoil could undermine one of the first generative AI success stories.

Editor's note: Bloomberg now reports that Stability AI has explored selling the company, and investors are urging CEO Emad Mostaque to resign. This aligns with Synthedia’s forecast that Stability AI will likely need a white knight acquirer or large investor to survive.

Becoming an early winner of a generative AI market segment offers significant advantages but does not ensure lasting success. Stability AI, the creator of Stable Diffusion, arrived in the market after DALL-E and Midjourney. However, as a high-quality open-source solution, it quickly became the most widely used text-to-image AI foundation model by developers and others looking to offer image generation.

Within three months, the company claimed to have more than 10 million users of the Stable Diffusion model across all channels and 1.5 million for the Stability AI Dream Studio. DALL-E reported 1.5 million, and Midjourney had 2-3 million users in the same period.

Stable Diffusion’s broad adoption was even cited as key evidence in the famous memo written by a Google research engineer that began, “We have no moat, and neither does OpenAI.” Stable Diffusion supposedly showed that open-source would beat proprietary models in the generative AI model competition. Stability AI may not have been the first to market, but it appeared to be the first to scale.

However, despite this momentum, Stability AI struggled to raise funding this past spring at its desired valuation and has had several acrimonious executive departures. Open-source is still trailing in the large language model (LLM) segment, showing it is not necessarily an inevitable winner in every generative AI category. What does this mean for the Stability AI?