OpenAI publishes another article: Better models open up new "work areas".

Following the release of GPT-6 Astra, Sarah Friar, Chief Business Officer of OpenAI, wrote an article explaining the company's business logic: stronger model capabilities, lower computing costs, and a continuously expanding user base create a compounding effect, driving AI to leap from an auxiliary tool to a core productivity driver.

In a recent article, OpenAI Chief Business Officer Sarah Friar stated that GPT-6 Astra is the most intelligent and aligned model to date, achieving industry-leading performance in areas such as computer operation, software engineering, cybersecurity, scientific research, and professional work. She pointed out that more powerful models are making previously impossible tasks possible due to time, cost, or professional barriers, thereby opening up entirely new market opportunities.

In terms of commercial impact, Friar emphasized that OpenAI currently boasts over 1 billion weekly active users and 2.5 million enterprise customers. This dual-track approach, encompassing both consumer and enterprise clients, ensures that every model upgrade directly reaches existing users, creating diversified revenue streams. Simultaneously, the company's self-developed computing power strategy—covering a full stack of data centers, chips, software, models, and products—is continuously reducing the delivery costs of intelligent services, providing an economic foundation for large-scale expansion.

User stickiness continues to increase, and consumption and enterprise sectors reinforce each other.

Friar cited internal OpenAI data showing that after six months of registration, individual ChatGPT subscribers saw their daily message volume increase by approximately 50% compared to the first month, and the number of different task types they tried roughly doubled. This trend indicates that users' reliance on AI tools increases significantly with longer usage time.

In terms of business model, OpenAI helps users discover AI application scenarios through ad-supported free access, and then monetizes user value through subscription and pay-as-you-go models. Friar points out that there is a clear mutually reinforcing effect between the consumer and enterprise ends: individual users familiar with ChatGPT will bring their usage habits into the workplace, while enterprise deployment, in turn, increases users' expectations and reliance on AI, thus influencing their usage behavior in their personal lives. She predicts that as OpenAI's intelligent agent products evolve, the traditionally relatively independent consumer and enterprise markets will continue to merge.

With a leap in model capabilities, AI is beginning to undertake complex scientific research tasks.

In his article, Friar revealed internal productivity data from OpenAI: the research team currently has the workload of 3.1 agent workdays for every human workday. While researchers are accelerating code contributions and running more experiments, they are delegating increasingly complex tasks to agents.

On the scientific breakthrough front, Friar announced that an internal OpenAI model has provided a solution to the Navier-Stokes Millennium Prize Problem. This mathematical problem, unsolved for nearly 90 years, is considered one of the most profound open problems in mathematics. Friar characterized this as a significant milestone in AI's ability to participate in mathematical research.

In addition, the article cites a case from Boston Children’s Hospital: with the help of AI-assisted research, specialists found more than 40 diagnostic answers in previously undiagnosed rare disease cases.

Jalapeño, a self-developed chip, will go into production this year, significantly reducing computing costs.

Regarding the economics of computing power, Friar disclosed two key advancements.

First, GPT-5.6 Sol helps optimize production service software, reducing end-to-end service costs by 20% and improving token generation efficiency by over 15%. Second, OpenAI's first self-developed inference chip, Jalapeño, demonstrated in InferenceX testing that, compared to existing commercial systems, peak token throughput per watt was 1.5 to 1.9 times higher, while end-to-end latency was reduced by 1.7 to 3.6 times. The company plans to begin deploying Jalapeño by the end of the year, while continuing to use it in conjunction with accelerators from NVIDIA, AMD, and other partners.

Friar stated that a better model can reduce the number of attempts required to complete a task, while better hardware and software can reduce the time and cost of each attempt. The synergistic improvement of both will enable OpenAI to take on more workloads from its existing computing power.

The Logic of Compound Interest: The Flywheel Effect of Research, Products, and Computing Power

At the end of the article, Friar elaborates on OpenAI's core business flywheel: better models open up new viable jobs, more efficient computing power makes these jobs economically viable on a larger scale, and the revenue from increasing user adoption feeds back into investment in next-generation research and infrastructure.

She also emphasized the importance of capital discipline, stating that the company will use "the scale of demand it can serve, the speed at which capital investment is transformed into productivity, and whether the return matches the committed capital" as the criteria for evaluating each investment. Friar believes that the combined effect of these advantages gives OpenAI the confidence to maintain its leading position in the continuous generational evolution of AI.

Risk Warning and DisclaimerInvesting involves risk; please exercise caution. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial situation, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific circumstances. Any investment decisions made based on this information are at your own risk.