Xi Jinping Unveils Global AI Cooperation Organization at Shanghai Summit
In a landmark move that underscores the intensifying geopolitical competition over artificial intelligence, Chinese President Xi Jinping delivered the opening keynote at the 2026 World AI Conference in Shanghai on July 17, calling for AI development to be "a symphony of international cooperation, not a solo performance by a single country."
The headline announcement: 29 countries — including Russia, Brazil, Indonesia, Pakistan, and South Africa — signed an agreement to establish the World AI Cooperation Organization (WAICO), a new governance body that Beijing wants headquartered in Shanghai. The move positions China as the architect of a parallel AI governance framework, distinct from Western-led initiatives.
Xi also pledged 5,000 AI training opportunities for developing nations and outlined cooperation agreements with ASEAN, the League of Arab States, and the African Union. The conference runs through July 20, with over 1,100 companies and 1,400 guests in attendance.
The timing is pointed: the announcement comes as US export controls continue to constrain China's access to advanced chips, yet Chinese AI labs are producing increasingly competitive models.
Sources: NPR, CNBC, Al Jazeera
Moonshot AI's Kimi K3 Pushes Chinese AI Into Frontier Territory
Speaking of competitive Chinese models: Moonshot AI launched Kimi K3 on July 16, a sprawling Mixture-of-Experts system with 2.8 trillion total parameters and a 1-million-token context window. It's the largest open-weight model ever released, and its benchmarks tell a story the industry is paying close attention to.
Kimi K3 uses 896 experts but activates only 16 per token — roughly 1.8% of the total pool — keeping inference costs manageable despite the enormous parameter count. The model's signature innovation, Kimi Delta Attention (KDA), a hybrid linear attention architecture, delivers 6.3x faster decoding at million-token context lengths compared to standard transformer attention.
Performance-wise, K3 outperforms all rivals except Anthropic's Claude Fable 5 and OpenAI's GPT-5.6 Sol on overall capability, putting it firmly in frontier territory. It's priced at $3 per million input tokens and $15 per million output tokens — competitive with Anthropic's Sonnet tier. Open weights are promised by July 27.
The model is available in two variants: K3 Max for chat and agent tasks, and K3 Swarm Max for large-scale parallel processing.
Sources: Fortune, MarkTechPost, TechCrunch
Anthropic Begins IPO Roadshow, Targeting $965 Billion Valuation
The AI industry's most anticipated public offering took a concrete step forward this week. Anthropic, the company behind Claude, has begun scheduling investor meetings through Goldman Sachs, Morgan Stanley, and JPMorgan Chase, advancing toward an IPO that could come as early as October 2026.
The pre-roadshow meetings follow a confidential S-1 filing submitted to the SEC on June 1 and a massive $65 billion Series H round that valued the company at $965 billion. If the listing holds at that valuation, Anthropic would debut as one of the most valuable companies ever to go public.
The IPO push comes at a moment of strength: Anthropic has reportedly overtaken OpenAI in self-reported revenue, is on track to hit $47 billion in annual revenue, and projects profitability by 2029 — a year ahead of OpenAI's timeline. The company is also in talks with Samsung to build a custom AI inference chip using Samsung's 2nm process, following OpenAI's unveiling of its own Jalapeño inference chip in June.
Sources: CNBC, Bloomberg, TechCrunch
Google Delays Gemini 3.5 Pro as Alphabet Shares Slide
Not everything is going smoothly in the AI race. Google has delayed the broader release of Gemini 3.5 Pro after internal testing revealed the model fell short of expectations in coding performance and complex, long-horizon reasoning tasks.
The delay — originally slated for a June release — sent Alphabet shares down more than 4% when reports surfaced. It's a rare stumble for Google's AI efforts, which have been on an otherwise strong trajectory with the Gemini family.
In other Google AI news, the company rebranded NotebookLM to Gemini Notebook on July 16, adding a secure cloud computing environment that lets Pro users run code directly inside notebooks — a move that puts it in more direct competition with tools like Replit and Cursor.
Sources: AIapps, BuildFastWithAI
Quick Hits
Five Eyes warns AI will transform cyber warfare "in months, not years": The intelligence alliance issued a stark warning that frontier AI models will fundamentally transform both offensive and defensive cyber capabilities on a timeline measured in months. Microsoft is preparing Project Perception, an AI cybersecurity product using models from Microsoft, OpenAI, and Anthropic.
Fireworks AI raises $1.5B: The inference infrastructure startup closed a Series D round valuing it at $17.5 billion, signaling continued investor appetite for the infrastructure layer powering AI deployment.
Apple reclaims most valuable company: Apple surpassed Nvidia in market capitalization, reclaiming the title of the world's most valuable publicly traded company.