OpenAI’s annualized recurring revenue topped $20 billion last year and its data centre capacity expanded to 1.9 gigawatts, SiliconANGLE reported on Jan. 19.

Revenue tripled from $6 billion in 2023. Data centre capacity also rose tenfold over the period. OpenAI stressed that it is adopting a strategy of investing in stages to respond to demand in line with its growth pace.

OpenAI is also focusing on hardware optimisation to improve cost efficiency. It cut inference costs to below $1 per 1 million tokens. It is operating by using advanced hardware for model training and low-cost infrastructure for large-scale tasks. But some point to cost cutting as still an urgent task.

The Information forecasts OpenAI will post an $8 billion loss in 2025, up $1.5 billion from the initial estimate. Losses are seen more than doubling to $17 billion in 2026.

OpenAI has moved to develop its own chips and build data centres to cut costs. Last year, it signed a $10 billion deal with Broadcom to jointly design AI accelerators. It is also building a customised data centre, called Stargate, in cooperation with SoftBank. It is also seeking new revenue models. Chief Financial Officer Sarah Friar projected that licensing, IP-based contracts and performance-based pricing models will emerge in the AI market. OpenAI recently announced plans to insert paid ads into ChatGPT responses and is testing them for some users in the United States.

OpenAI is also developing AI agents and workflow automation tools. The tools help users connect multiple applications to automate tasks.