On September 2, Sam Altman, CEO of OpenAI, clearly stated that the company will develop humanoid robots and other forms of robot bodies in a full-stack self-research and development approach, while also developing specialized robots for data centers. The relevant business has been integrated into the OpenAI Robotics team, led by Aditya Ramesh. This plan aims to assist skilled workers in building infrastructure in the short term, and in the long run, it seeks to popularize all-around personal assistants, deepening the collaborative design between hardware and machine learning.

At the same time, OpenAI's layout in the cross-industry AI hardware ecosystem is accelerating. The first hardware terminal designed by Jonathan Ive, former chief designer of Apple (internal code name "Gumdrop"), is expected to be launched in 2027 with a price exceeding $300. This device is positioned as a screenless portable AI computer, shaped like a donut and the size of an ice hockey puck, equipped with movable parts to enhance the human-like interactive experience, aiming to create a "physical embodiment" of ChatGPT. The hardware team currently has about five products under development, with Foxconn already involved in prototype production.

Amid rapid business expansion and preparations for the 2027 IPO, OpenAI still faces two major challenges. On one hand, Apple filed lawsuits and applied for preliminary injunctions in July and August of the same year, accusing OpenAI of stealing trade secrets of unreleased products through poaching; OpenAI denied these allegations.

On the other hand, despite OpenAI's annualized revenue surpassing $20 billion in 2025 and second-quarter revenue reaching $6.7 billion in 2026 (an 18% increase quarter-on-quarter), growth has shown signs of slowing down, with operating losses continuing to expand. In comparison, the competitor Anthropic achieved quarterly revenue exceeding $11.5 billion during the same period and realized operational profitability, exerting positive market pressure on OpenAI. OpenAI's recent in-depth expansion into integrated software and hardware marks the acceleration of general artificial intelligence (AGI) toward embodied and terminal real-world applications.