Emad Mostaque leaves Stability AI amid company crisis
The founder of Stability AI has stepped down as CEO and resigned from the board. His exit follows layoffs, financial strain and the departure of researchers from a company that helped popularize generative image AI.
Emad Mostaque has stepped down as CEO of Stability AI, the company behind Stable Diffusion, one of the most influential image-generation models in the current wave of artificial intelligence. He is also leaving the company’s board of directors.
The departure comes on March 22, after months of financial difficulties, layoffs and the exit of several researchers and executives. The board has appointed chief operating officer Shan Shan Wong as interim CEO while it searches for a permanent successor.
Mostaque has said he wants to focus on decentralized artificial intelligence. In a message posted on X, he argued that centralized AI cannot be fought with more centralized AI. The idea sums up a fundamental disagreement: while major labs are controlled by Microsoft, Google, Amazon or Meta, the founder argues that AI models and infrastructure should be distributed among more players.
The company that brought image generation to millions of users
Stability AI rose to prominence in 2022 with Stable Diffusion, a system capable of creating images from a written description. Its importance was not just down to the quality of its results: the company released the model’s weights — the numerical files that capture what it learned during training — allowing developers and companies to run and adapt it on their own hardware.
That approach helped drive the rapid expansion of creative tools, editing applications and services based on visual generation. It also put Stability AI at the center of debates over copyright, training data and the commercial use of machine-generated images.
The company raised $101 million in a round announced in October 2022, at a valuation close to $1 billion. At the time, investor enthusiasm for generative AI made Stability AI one of the most visible names outside the major technology companies.
Rapid growth also drove up costs
The company’s position has deteriorated since then. Training competitive models requires vast amounts of computing capacity, expensive infrastructure that is typically controlled by cloud providers such as Amazon Web Services, Google Cloud and Microsoft Azure. Releasing open models, moreover, does not by itself guarantee enough recurring revenue to sustain that investment.
Stability AI cut staff in October 2023. Other high-profile departures followed, including Ed Newton-Rex, who led the company’s audio division. Newton-Rex publicly explained that he was leaving because he disagreed with using copyrighted works to train generative systems without permission.
The company has also faced lawsuits in the United States over the images used to train Stable Diffusion. Those lawsuits have not yet resolved the central question: to what extent does training models on works available online require authorization or compensation for their creators?
The challenge of finding a business model
Mostaque’s resignation does not change Stable Diffusion’s technical importance or the broad community that has grown around it. But it does leave a crucial business question unanswered: how to finance the development of open models when the most powerful competitors have the budgets, chips and infrastructure of the biggest platforms behind them.
Shan Shan Wong’s tenure will be a test for Stability AI. The company needs to retain talent, secure funding and show that it can turn its technology into sustainable products and services without abandoning the open spirit that made it relevant. For the industry, its situation shows that releasing models can accelerate innovation, but it does not automatically solve who pays the cost of building and maintaining them.
