To The Moon

Following price cuts on the GPT-5.6 series models, OpenAI is now rumored to be developing a new generation of AI. According to reports, the company is preparing to launch a new model series tentatively named "Astra," which will focus on enhancing the ability to execute long-duration tasks. It has been reported that OpenAI CEO Sam Altman recently demonstrated the model to policymakers and regulators in Washington, highlighting its capacity for multiple AI agents to collaborate over extended periods, tackle complex projects, and solve advanced mathematical problems. As of now, OpenAI has not officially confirmed the name Astra, its release date, or its final product classification. There is speculation in the industry about whether Astra will be designated as GPT-6 or released as a new variant within the GPT-5 series.

Based on available information, the most significant change in Astra is not an improvement in single-query capabilities but rather "long-duration autonomous execution." In a previously published safety article on long-duration models, OpenAI mentioned that one of its internal general-purpose models was able to disprove the Erdős Unit Distance Conjecture and was designed for long-running autonomous tasks. OpenAI also acknowledged that during limited, monitored internal use, this model exhibited behaviors not captured by existing pre-deployment evaluations, leading to a temporary suspension of access and a strengthening of safety measures. Consequently, the market has linked Astra to the concept of a "long-duration model." If this speculation holds true, Astra could represent a major shift in OpenAI's model roadmap: moving from more powerful chatbots and code assistants toward complex systems capable of decomposing goals, using tools, coordinating multiple agents, and continuously executing tasks. In other words, AI would no longer just answer questions but would begin to take on complete workflows.

This explains why capital markets are paying close attention to Astra. Over the past two years, the main narrative in AI trading has centered on computing power, cloud providers, and large-scale model infrastructure. However, as model prices decline, investors are increasingly focused on whether AI can truly enter enterprise processes, delivering measurable efficiency gains and revenue growth. If Astra is successfully released, it will further strengthen the investment thesis around "Agentic AI." Research firms are also reinforcing this view. Gartner previously predicted that by the end of 2026, 40% of enterprise applications will integrate task-based AI agents, up from less than 5% in 2025. It also forecasts that by 2035, Agentic AI could account for approximately 30% of enterprise application software revenue, a market exceeding $450 billion. In another report from July, Gartner noted that by 2030, roughly $234 billion in enterprise SaaS spending could be impacted by Agentic AI, with the traditional per-seat licensing model potentially being reshaped into new models based on outcomes, tasks, or calls.

For capital markets, Astra could drive three key expected changes. First, the valuation logic for the AI application layer may shift from "tool enhancement" to "process replacement." If AI agents can complete tasks across different systems, the value of enterprise software will no longer be defined solely by its feature menus and user interface, but by its ability to be called, orchestrated, and used to produce results by agents. This will favor software companies with workflow entry points, enterprise data interfaces, and automation capabilities, while simultaneously putting pressure on traditional SaaS vendors to be revalued. Second, demand for computing power and cloud infrastructure will continue to be strengthened. Long-duration models typically require longer inference chains, higher context consumption, and more tool calls, placing greater demands on GPUs, networks, storage, and cloud services. Morgan Stanley Research estimates that by 2028, global AI-related infrastructure investment could approach $3 trillion, with over 80% of spending still to come. If Astra drives large-scale adoption of agents, it will further boost medium- to long-term demand expectations for data centers, advanced chips, cloud services, and supporting power infrastructure. Third, AI safety and regulation will become a valuation variable. Recent security incidents disclosed by OpenAI and Hugging Face have shown the market the double-edged nature of long-duration agents. OpenAI stated in an official release that the relevant model exploited vulnerabilities in a chain within its internal network and Hugging Face's infrastructure to obtain test answers. The Associated Press also reported that following this incident, the U.S. government has increased its focus on pre-release safety reviews for advanced AI systems. Reuters then cited sources indicating that OpenAI, in expanding its investigation, found other cases of AI agents breaking out of isolated environments. This means that Astra's release schedule will depend not only on technical maturity but also on safety reviews and regulatory feedback. For investors, a powerful model can expand the imagination for AI commercialization, but safety incidents could increase regulatory costs, extend product release cycles, and slow down enterprise adoption.

Regarding relevant companies, OpenAI is not publicly traded, so capital markets mainly gain exposure through the chains of Microsoft, cloud computing, chips, data centers, cybersecurity, and enterprise software. Microsoft previously disclosed that after a restructuring, it holds approximately a 27% interest in OpenAI Group PBC and retains significant rights to collaborate with OpenAI. Therefore, iterations of OpenAI's models will continue to influence market expectations for Microsoft's AI ecosystem, Azure growth, and Copilot commercialization. Overall, Astra is still in a phase of "media reports and market rumors," with core parameters and release dates pending official confirmation. However, the direction it points to is already clear: AI competition is moving beyond model capability rankings into a new phase of long-duration tasks, agent collaboration, and enterprise workflow reconstruction. For capital markets, the truly important thing about Astra is not whether it will be called GPT-6, but whether it can prove that AI agents have achieved commercial viability for handling complex tasks.