Transformer Co-Inventor Noam Shazeer Defects from Google to OpenAI
In what may be the most consequential talent move of 2026, Noam Shazeer — co-author of the 2017 paper "Attention Is All You Need" that introduced the Transformer architecture underpinning virtually every modern AI model — has left Google DeepMind to join OpenAI as Lead for Architecture Research.
The move is especially striking given that Google spent $2.7 billion in 2024 to acquire Character.AI assets and bring Shazeer back after he departed in 2021. Less than two years later, one of AI's most important researchers has walked out the door — this time to Google's fiercest rival.
"I'm excited to share that I'll be joining OpenAI and look forward to working with the exceptional team there," Shazeer announced on X. At OpenAI, he will oversee the physical neural network architectures underlying all of the company's models — a role that could reshape the competitive landscape as OpenAI prepares GPT-5.6 for a late-June launch and eyes a September IPO.
The departure also comes at a delicate time for Google, which is simultaneously teasing Gemini 3.5 Pro for a June release window. Losing the person who literally co-invented the architecture behind these models sends a powerful signal about where the center of gravity in AI research is shifting.
Fable 5 Standoff Enters Day 11 — Free Trial Expires, Model Still Offline
Anthropic's flagship Fable 5 and Mythos 5 models remain completely inaccessible worldwide, now 11 days after a US Department of Commerce export control directive forced the company to suspend access on June 12.
The situation took a new turn over the weekend: the White House reportedly issued an ultimatum demanding Anthropic eliminate all potential jailbreaks before any relaunch — a requirement cybersecurity experts have called "technically impossible" for current frontier models. CEO Dario Amodei has pushed back, arguing the flagged vulnerability is minor and not unique to Fable 5.
Adding insult to injury for subscribers, Fable 5's free trial window expired on June 22, meaning users who signed up for early access now need paid usage credits for a model they can't even use. Polymarket traders put the odds of restoration before July 1 at roughly 58–67%.
Meanwhile, Anthropic opened its Seoul headquarters — its third Asia-Pacific office — and signed a safety-focused MOU with South Korea's Ministry of Science and ICT, even as its most powerful models remain dark.
SpaceX Cements $60 Billion Cursor Acquisition — Largest VC-Backed Startup Deal Ever
SpaceX is pushing forward with its $60 billion all-stock acquisition of Cursor, the AI coding assistant that has become one of the most popular developer tools in the industry. The deal, announced just days after SpaceX's record-breaking Nasdaq IPO, is expected to close in Q3 2026.
Cursor, founded in 2022, has scaled to approximately $2.6 billion in annualized revenue with rapid enterprise adoption. The acquisition would make it the largest purchase of a VC-backed startup in history (excluding Musk's earlier xAI self-deal) and signals SpaceX's ambition to compete directly with OpenAI's Codex and Anthropic's Claude Code in the AI-powered development tools market.
Fortune noted that SpaceX's surging stock effectively paid for the entire acquisition in just a few hours of trading — a stark illustration of how post-IPO momentum can be weaponized for rapid consolidation.
China Unveils $295 Billion Sovereign AI Infrastructure Plan
In a move with enormous geopolitical implications, China has unveiled a five-year, 2 trillion yuan ($295 billion) plan to build a nationwide network of interconnected AI data centers. When power grid integration is included, the total projected investment could reach $740 billion.
The plan, led by the National Development and Reform Commission, mandates that at least 80% of technology — including AI chips — must come from domestic suppliers like Huawei, effectively locking out Nvidia and AMD. State firms including China Mobile and China Telecom will operate the bulk of the infrastructure.
The timing is noteworthy: Beijing announced the plan while the US was simultaneously tightening export controls on frontier AI models. The dueling moves underscore how AI has become the central front in US-China technological competition, with both nations increasingly pursuing self-sufficiency over interdependence.
ChatGPT Market Share Drops Below 50% for the First Time
For the first time since its launch, ChatGPT's global market share has fallen below 50%, dropping to 46.4% according to the latest usage data. Google's Gemini now commands 27.7% of the market, while Anthropic's Claude has climbed to 10.3%.
The shift reflects what analysts describe as a move from default selection to active comparison — users are no longer simply defaulting to ChatGPT but actively evaluating alternatives. The trend is amplified by June 2026 being the most concentrated AI model launch month in history, with major releases from Anthropic (Fable 5), Google (Gemini 3.5 Flash), xAI (Grok 4.3), DeepSeek (V4), and China's Zhipu AI (GLM-5.2).
Speaking of competition: Zhipu AI's open-weight GLM-5.2 scored 62.1 on SWE-bench Pro versus GPT-5.5's 58.6, while costing roughly one-sixth as much — another sign that the model layer's competitive moat is now measured in weeks, not quarters.