Anthropic Surge & AI IPO Wave: OpenAI Delay
Key Questions
What is the status of OpenAI's IPO plans?
OpenAI has filed for an IPO but is considering a delay until 2027 due to market volatility and ongoing losses. Noam Shazeer has joined the company ahead of any potential IPO.
What is Anthropic's current valuation?
Anthropic maintains a $965B valuation despite the broader market conditions affecting AI IPOs. It continues to push for AI pauses and faces a class action lawsuit over token allowances.
What new model did Anthropic release recently?
Anthropic released its public Fable 5 model, which includes capability restrictions. It also made Claude free for US K-12 teachers.
How are Chinese AI models impacting the US market?
Moonshot AI's Kimi K3, a 2.8T parameter open-source model, rivals Fable 5 and outperforms GPT-5.6 Sol on benchmarks while using domestic chips, triggering a 'code red' in the US. Qwen3.8 with 2.4T parameters was also announced as open-weight.
What infrastructure deals are being discussed in AI?
Meta is reportedly in talks for a $10B compute lease with Anthropic. DeepSeek raised $7.4B amid aggressive pricing competition, and Kimi K3 demand is straining GPU capacity.
What regulatory or global AI discussions are underway?
G7 leaders met with AI CEOs as Europeans seek checks on US dominance. OpenAI called for global oversight and predicts AI doing AI research by 2028, while EU orders on Google aim to boost AI assistant competition.
What concerns exist around open-weight models?
Open-weight releases like Kimi K3 and Qwen3.8 are reigniting US-China AI rivalry but some argue they accelerate innovation rather than harm US interests. Kimi K3 shows strong cybersecurity performance at lower cost.
What other notable AI developments are mentioned?
Frontier models are scaling to 10T parameters. Thinking Machines Lab founded by Mira Murati and John Schulman is gaining attention, and OpenAI's strategic head described AI as a public good akin to 'automated communism'.
OpenAI and Anthropic file for IPOs, but OpenAI considers delaying to 2027 amid market volatility and losses; Anthropic's $965B valuation remains. Noam Shazeer joins OpenAI pre-IPO. Anthropic releases public Fable 5 model with capability restrictions. G7 leaders meet with AI CEOs; Europeans seek checks on US dominance. OpenAI calls for global oversight, predicts AI doing AI research by 2028. Anthropic urges AI pause. Anthropic faces class action over token allowances. Google Gemini 3.5 Pro delayed due to coding shortfalls. Claude now free for US K-12 teachers. EU orders Google to open Android and Search to rivals, boosting AI assistant competition. DeepSeek raises $7.4B, signaling aggressive pricing war. Moonshot AI releases Kimi K3 (2.8T params open-source) rivaling Fable 5 and beating GPT-5.6 Sol on benchmarks, using domestic chips and triggering 'code red' in US. Meta reportedly in talks for $10B compute lease to Anthropic, highlighting insatiable demand for training infrastructure. Thinking Machines Lab (Mira Murati, John Schulman) gains attention. Frontier models scaling to 10T parameters; open-weight models from China accelerating competition. Kimi K3's open-source release reignites US-China AI rivalry. K3 can be used as a teacher model for distillation, enabling smaller models to benefit from its logits. OpenAI's strategic head frames AI as public good as 'automated communism', adding governance tension. Qwen3.8 (2.4T params) announced open-weight, second only to Fable 5. Kimi K3 demand hitting GPU capacity, forcing subscription pauses; Huawei chip throughput could be key. Article argues US should not fear open-weight models, as they accelerate innovation. Kimi K3 shows strong cybersecurity benchmark performance vs GPT-5.6-terra at lower cost.