OpenAI Launches GPT-Realtime-2 and Faces $18B Chip Deal Hurdle – May 2026 Updates

Key points

• •In early May 2026, OpenAI released several real-time voice AI technologies including GPT-Realtime-2, three new real-time voice models, and enhanced voice intelligence features in its API. These innovations improve developers' ability to build responsive and intelligent voice-based applications with live translation and voice reasoning capabilities akin to GPT-5 performance.[mezha.net][TechCrunch][qz.com][the-decoder.com]
• •OpenAI faced a critical financial hurdle in its $18 billion custom AI chip agreement with Broadcom as of May 2026, raising uncertainties about the future scalability and hardware infrastructure central to OpenAI’s AI model deployment. This setback has led to investor concerns and potential delays in OpenAI’s plans to expand its in-house AI hardware capabilities.[marketscreener.com]
• •On May 7, 2026, OpenAI upgraded ChatGPT's default conversational model to GPT-5.5 Instant, enhancing response speed and AI reasoning for millions of users worldwide. This upgrade signifies an ongoing evolution of OpenAI's flagship conversational AI, keeping it competitive in rapid-response AI services.[qz.com]
• •OpenAI's GPT-5.3-Codex, released on May 6, 2026, represents a landmark in AI-assisted coding, leveraging AI self-improvement methodologies that contributed to its own development. This innovation underlines OpenAI’s deepening capabilities in automated code generation and software development.[Mashable]
• •OpenAI is bolstering its executive leadership by hiring The Trade Desk’s Chief Strategy Officer, with the latter company having generated $689 million in revenue, signaling strategic growth and expanded business development efforts at OpenAI.[Let's Data Science]
• •In cybersecurity, OpenAI launched a new advanced AI model to rival Anthropic’s Mythos in early May 2026 and has been expanding access to this model among cyber defense professionals. This strategy marks OpenAI’s significant push into AI-driven cyber defense markets.[Politico]
• •OpenAI introduced a new 'Trusted Contact' safety feature on May 7, 2026, designed to help identify and support users showing signs of potential self-harm, reinforcing its commitment to ethical and responsible AI use.[TechCrunch]

Relevant articles