15 May 2026 PM: Cerebras IPO Soars 108% as OpenAI Takes Aim at Apple

An afternoon of big numbers and bigger battles: OpenAI reportedly lines up lawyers against Apple, Cerebras makes history on Nasdaq with its stock doubling in a single session, and a $650 million startup arrives with a plan to build AI that improves itself forever.

OpenAI is reportedly preparing legal action against Apple, in what would be one of the highest-profile AI partnership disputes the industry has seen. According to TechCrunch, OpenAI grew so frustrated with Apple’s handling of the ChatGPT integration that it is now actively exploring its legal options. The company reportedly felt the deal failed to deliver the subscriber growth and prominent placement it had expected when Apple embedded ChatGPT into Siri and Apple Intelligence in 2024.

If formal legal action follows, it would set a significant precedent for how AI integration agreements are structured and enforced. Apple has not commented, and OpenAI has not confirmed it is pursuing litigation. The dispute fits a wider pattern: OpenAI has clashed with multiple former partners, including its ongoing legal fight with Elon Musk. For businesses currently negotiating AI integration deals, this is a sharp reminder that even the biggest partnerships can unravel when distribution expectations are not nailed down precisely. If you are thinking about how to structure a clear AI policy for your business, the OpenAI-Apple situation is instructive.

Cerebras Systems completed one of the most dramatic stock market debuts in years on Thursday, raising $5.5 billion and watching its share price more than double within hours of opening to public trading. The AI chip company priced shares at $185 on Wednesday evening, well above its original range of $115 to $125, after increasing the offering to 30 million shares. When markets opened on Thursday, shares jumped to $385, a rise of 108 per cent, before settling around $330 by mid-session. At the IPO price alone, Cerebras entered the market with a fully-diluted valuation of $56.4 billion.

Cerebras builds chips designed specifically for AI inference, the ongoing compute processing required every time a language model responds to a prompt. The company counts OpenAI, Amazon Web Services, and Saudi Arabia’s Mohamed bin Zayed University of Artificial Intelligence among its customers. IPO plans had previously stalled in 2024 following a regulatory review tied to investment from Abu Dhabi-based Group 42. Those plans revived after the company reported $510 million in revenues for 2025, a 76 per cent year-on-year increase, alongside a swing to $237.8 million in net profit. This is the first major tech IPO of 2026, and the scale of investor demand suggests appetite for AI infrastructure remains very hot.

If you manage code or technical workflows on the move, OpenAI has just made that significantly easier: Codex is now available directly through the ChatGPT mobile app. OpenAI confirmed this week that its Codex coding agent can now be monitored, steered, and approved from any device including smartphones. Codex is an AI coding agent that takes plain-language instructions and carries out programming tasks with some degree of autonomy, rather than simply suggesting the next line of code. Users can assign a task from their phone, check progress, and approve the output without sitting at a desk.

This is a practical update for anyone managing software projects across multiple locations, or for freelancers and small business operators who rely on AI-assisted development. Understanding what AI agents can actually do today is useful context: Codex handles defined tasks well but benefits from human review at key decision points. OpenAI’s move to bring it to mobile signals that AI coding workflows are becoming part of everyday working life, not only desktop development environments.

A new $650 million AI startup called Recursive Superintelligence emerged from stealth this week with one of the most ambitious stated goals in the industry: building an AI that can improve itself indefinitely without human input. The company is led by Richard Socher, who founded the AI search startup You.com, alongside Peter Norvig, a former director of research at Google, and Tim Rocktaeschel, who led self-improvement research at Google DeepMind. Backing comes from Greycroft and GV. The core idea involves creating a system that identifies its own weaknesses and redesigns itself to address them, cycling through that loop autonomously rather than waiting for humans to run a new training process.

Socher told TechCrunch the team expects to ship products within quarters, not years, though no specific date was given. Recursive self-improvement has been a theoretical goal of AI research for decades. Whether this startup can operationalise it at commercial scale remains an open question, but the calibre of the research team and the scale of the funding make it one to follow closely.

More than 50 employees have reportedly left SpaceXAI since the company was formed by merging Elon Musk’s xAI with parts of SpaceX in February. The departures, reported by TechCrunch, span multiple teams. Sources point to burnout, leadership disruption following the merger, and liquidity events that left some employees with less financial incentive to stay. SpaceXAI has not commented. The company was presented as a consolidation of Musk’s AI assets alongside SpaceX’s compute infrastructure. Whether the losses represent manageable transition turbulence or a deeper structural problem will become clearer over the coming months as product development pace either holds or slips.

Here is everything else worth knowing from today’s AI news.

• What the jury will actually decide in the Elon Musk vs Sam Altman case. The fraud trial over OpenAI’s transition to a for-profit structure is now in front of a jury. TechCrunch breaks down exactly what is being decided. Source
• Cisco cuts nearly 4,000 jobs to spend more on AI, while reporting record revenue. The redundancies explicitly redirect budget toward AI investment even as the company claims its strongest revenue quarter yet. TechCrunch
• Wirestock raises $23M to supply multimodal AI training data. The platform connecting creators with AI companies buying licensed datasets raised fresh funding as demand for high-quality training material intensifies. TechCrunch
• Ontario auditors find AI medical note takers routinely get basic facts wrong. An audit of AI-powered clinical documentation tools found recurring inaccuracies in patient notes, raising questions about the pace of deployment without sufficient validation. The Register
• Clawdmeter turns your Claude Code usage stats into a desktop dashboard. A new open-source tool gives developers a real-time visual readout of Claude Code API consumption, useful for anyone tracking usage caps or managing costs. TechCrunch
• Khosla Ventures bets $10M on Synthetic, an autonomous AI bookkeeping service for startups. Founder Ian Crosby previously co-founded Bench before it collapsed. He is now building an AI-native replacement for the entire bookkeeping category. TechCrunch

Watch whether OpenAI files formal legal papers against Apple in the coming weeks. If it does, the lawsuit will force the industry to define exactly what “prominent integration” and “distribution expectations” mean inside an AI partnership agreement, and that clarity will matter for any business currently negotiating similar deals with platform companies.

This is a daily news update for informational purposes only. AI products and policies change rapidly. Verify details directly with providers before making decisions. Nothing here is financial or legal advice.

AI Daily is Cristoniq’s afternoon update on developments in artificial intelligence, published every weekday afternoon.