Moonshot AI released Kimi K3.1 Pro, a post-trained Kimi K3 that it calls its most capable open-weight model in the 3T-parameter class, with weights due by Nov. 4. Moonshot reports 65.7% on Terminal-Bench 4.0, just behind Claude Opus 5.5 at 66.4% and ahead of GPT-6 Astra at 58.2%, and a claim-level hallucination rate of 8.3% on its internal hard test, down from 23.4% for K3. A Highspeed variant runs two to three times faster, the context window is 1M tokens, and all figures are Moonshot's own. (Moonshot AI)
Private Companies
OpenAIfundraise
OpenAI is close to raising up to $30B at a $1.25T pre-money valuation, CNBC reported, citing people familiar with the deal. UAE investors and Coatue are leading, and OpenAI is aiming to secure signatures by the end of next week, though the final size is still being set. One person said OpenAI is also negotiating a commercial deal for more compute capacity from the UAE. Its March round valued the company at $852B post-money. (CNBC)
DeepSeekfundraise
DeepSeek has raised about $15B, or roughly 100B yuan, Bloomberg reported, calling it one of the biggest funding rounds in China. CATL and Tencent were named among the potential investors, and the report gave no valuation. DeepSeek did not respond to a request for comment. The report came days after the US ordered federal civilian agencies to remove DeepSeek's AI technology from their systems over security concerns. (TechCrunch)
Helionother
Helion signed a letter of intent with the Commerce Department for up to $175M in CHIPS Act R&D funding to develop high-voltage power semiconductors and modules for its fusion generators. The funding includes more than $100M for chip and packaging R&D in Washington State and about $80M for work with GlobalFoundries in Vermont. CEO David Kirtley said halving the 5% to 10% of power these devices lose would double a plant's output. (Department of Commerce)
Vertical Semiconductor signed a letter of intent with Commerce for up to $73M to advance vertical gallium nitride power transistors and demonstrate a 2.5MW AI data center rack. It targets 30% lower energy losses and a 50% smaller footprint, with no baseline given. It is one of four letters of intent worth up to $389M, and Commerce will take a minority, non-voting equity stake in each. CEO Cynthia Liao said, "Power is the most significant bottleneck in the build-out of AI infrastructure." (Compound Semiconductor)
Public Markets
Lam fell after reporting September-quarter revenue of $8.22B and guiding December-quarter revenue to $8.60B. CEO Tim Archer said on the call that supply chain constraints for certain components are continuing to limit chipmakers. AMD fell 8.1% and the iShares Semiconductor ETF nearly 7%. (Yahoo Finance)
Meta fell after narrowing its 2026 capex outlook to $135B–$145B from $130B–$145B. Third-quarter capex was $34.91B and free cash flow was negative $2.12B, on revenue of $64.6B, up 26%. (Meta)
Emerging
Space data centers: A paper from Haotian Zheng, Zhanwei Wang and Kaibin Huang, Low-Latency MoE Inference in Space Data Centers: A Wireless-Power-Transfer Perspective, spreads the experts of a mixture-of-experts model across satellites, where busy experts run short of power while idle ones waste what their solar panels harvest. The authors move power between satellites over inter-satellite lasers, using a water-filling rule to decide which experts donate and which receive. In simulations with Mixtral-8x7B and Phi-3.5-MoE, inference latency falls by up to 29.25% against the same system without power transfer, and no hardware was measured. (arXiv)