Sam Altman, CEO of OpenAI, has just shared a critical insight into the future of AI infrastructure. He believes that as models become increasingly intelligent, global compute capacity will fail to keep pace with demand in the near future.
Developments
According to Altman, enterprise customers are demanding solid guarantees regarding bandwidth and compute resources. To alleviate this pressure and support long-term planning, OpenAI is starting to offer token discount packages for customers who commit to continuous usage for 1 to 3 years. This is a rare move in the industry, turning AI compute into a reserved asset much like real estate or electricity.
Why It Matters
This announcement indirectly confirms that OpenAI is facing demand that far outstrips supply, while signaling the upcoming release of larger models (potentially GPT-5) that will require massive resources. For AI startups in Vietnam, this serves as a reminder to focus on optimizing model inference efficiency rather than relying solely on raw hardware power, which is becoming increasingly expensive and difficult to access.
A Win-Win Strategy
Altman describes this as a win-win move: customers secure cost and resource predictability, while OpenAI gains the cash flow and forecasting data needed to build massive future data centers.