AI News Roundup — July 12, 2026
Apple sued OpenAI in federal court alleging trade secret theft and a coordinated recruitment campaign, citing more than 400 former Apple employees now working at OpenAI — filed just weeks ahead of OpenAI's reported confidential IPO preparations at a $730 billion valuation. Google Search became fully powered by Gemini 3.5 Flash-generated summaries, replacing the traditional ranked-links format entirely, while SK Hynix completed the largest ADR debut in Wall Street history on the strength of its dominant position in AI memory.
Apple Sues OpenAI for Trade Secret Theft — Cites Over 400 Former Employees Now at OpenAI
Apple filed a lawsuit against OpenAI in the US District Court for the Northern District of California on July 11, alleging trade secret theft and breach of contract, citing more than 400 former Apple employees who now work at OpenAI. Apple alleges the misconduct was directed by OpenAI's senior leadership, including Chief Hardware Officer Tang Tan — a 24-year Apple veteran who most recently served as VP of product design for iPhone and Apple Watch. The complaint accuses Tan of using Apple's confidential project code names during recruiting, asking job candidates to bring Apple hardware components to interviews, coaching departing employees on evading Apple's security procedures, and probing for details about unannounced products. Separately, the suit names Chang Liu — an eight-year Apple senior systems electrical engineer — for allegedly failing to return an Apple laptop after joining OpenAI and using it to download confidential technical documents, then sharing that information with other Apple employees applying to OpenAI. Apple's filing states directly: "This is the tip of the iceberg... OpenAI's nascent hardware business now rests on the shakiest of foundations, rotten to its core by its illegal reliance on misappropriated trade secrets." The lawsuit lands amid reports that OpenAI is developing its first hardware product — possibly a smartphone using AI agents instead of apps — that would directly compete with the iPhone. OpenAI's response on X: "We have no interest in other companies' trade secrets."
TechCrunch · CNBC · Build Fast With AI
OpenAI Preps Confidential IPO Filing at $730 Billion Valuation — Timing Collides With Apple Lawsuit
OpenAI is reportedly preparing a confidential IPO filing with Goldman Sachs and Morgan Stanley, targeting a public debut as soon as September 2026 at a private-market valuation around $730 billion — which would make it the largest technology IPO in history by a wide margin. The company hired Dean Ball on July 6 to lead a new Strategic Futures team reporting directly to Chief Strategy Officer Jason Kwon, part of months of senior finance and policy hiring ahead of a listing. The competitive backdrop is uncomfortable: Fortune reported Anthropic has overtaken OpenAI on revenue, running at approximately $47 billion annualized against OpenAI's projected $25-33 billion for 2026, with Claude Code alone growing from $1 billion to $2.5 billion in annualized revenue between end-2025 and February 2026. A trade secret lawsuit from the world's most litigious hardware company is exactly the kind of legal overhang that concerns IPO bankers weeks before a confidential filing. Between the Apple lawsuit, Anthropic's revenue lead, and a $730 billion valuation demanding flawless execution, September 2026 is shaping up as the most heavily scrutinized tech listing since the dot-com era.
Gemini 3.5 Pro Finally Has a Date: July 17 — 2M Token Context, Aggressive Pricing to Undercut Sol
Google DeepMind's Gemini 3.5 Pro is now set for general availability on July 17, per a leaked launch plan circulating since July 10. The model was reportedly rebuilt on an entirely new pretraining run rather than adapted from Gemini 2.5 Pro, ships a 2-million-token context window (double anything currently on the frontier), and gates its Deep Think extended reasoning mode behind the $250/month Ultra subscription. Expected API pricing: approximately $1.25 input and $10 output per million tokens — undercutting GPT-5.6 Sol by 4x on input cost while doubling everyone's context window. The delay means Gemini 3.5 Pro is now six weeks late, launching into the most competitive single week in AI history — five days after GPT-5.6, nine days after Grok 4.5 — while Google has been absorbing painful talent losses (Noam Shazeer to OpenAI, John Jumper to Anthropic in June). The pricing strategy suggests Google isn't trying to win the benchmark headline but the workload migration: at $1.25 input with 2 million tokens of context, entire categories of RAG architecture become simpler since developers can put the entire corpus directly in the prompt.
Google Search Fully Transitions to Gemini 3.5 Flash-Generated Summaries — The Ranked Links Era Ends
As of July 10, Google Search results are entirely powered by Gemini 3.5 Flash-generated summaries, replacing the traditional ten-blue-links format across the board, with source links embedded inside the AI-generated page rather than listed separately. This is the full rollout of the format Google has been testing in stages since AI Overviews launched in 2024 — meaning the default experience of the world's most-used website is now a generated document, not a ranked index. The implication for the web's traffic economy is structural: for 25 years, publishers produced content, Google ranked it, and clicks flowed downstream. That contract is now gone — visibility means being cited inside the AI answer, not ranked beneath it. Publishers have watched referral traffic erode for two years, but a fully generated results page converts gradual erosion into a structural cliff. Answer-engine optimization and generative-engine optimization — content that is entity-rich, directly quotable, and structured for AI extraction — is now the entire discoverability game for anyone who publishes online.
SK Hynix Completes Largest ADR Debut in Wall Street History
SK Hynix began trading on Nasdaq under ticker SKHY on July 10, with its $28-29 billion ADR offering priced at $149-166 per share — surpassing Alibaba's 2014 debut of $21.8 billion as the largest ADR listing in history. The fundamentals justify the appetite: SK Hynix holds roughly 60% of the global high-bandwidth memory (HBM) market and posted Q1 2026 revenue of $35.55 billion at a 72% operating margin — an extraordinary figure for a memory business that spent decades as one of tech's most brutally cyclical commodity segments. Every frontier AI training cluster and high-end inference deployment needs HBM stacked next to the accelerator, and Nvidia's entire data-center roadmap is gated on HBM supply. SK Hynix beat both Samsung and Micron to each successive HBM generation and locked in Nvidia qualification wins, turning memory into a sole-source-like business with genuine pricing power. The Nasdaq listing gives American institutional capital direct access to that position at the exact peak of AI memory demand.