Groq is seeking a Series B at an $8B valuation, about 2.5 times the $3.15B pre-money valuation it set in August, Bloomberg reported, citing investor documents. Executive chairman Alex Davis, who also runs Disruptive, has approached investors, and the documents project revenue growing eightfold in 2027 and $2B in EBITDA by 2028. Groq raised $350M at a $3.5B valuation in August, after Nvidia signed a non-exclusive licensing agreement for its technology in December 2025 and hired founder Jonathan Ross and other staff. Groq says it now serves developers from 13 data centers and plans to grow from 54MW to more than 200MW of capacity in 2027 (Bloomberg)
Private Companies
OpenAIfundraise
OpenAI closed about $11B of a funding round that values it at $1.25T, Bloomberg reported, citing people familiar with the matter. It first sought more than $30B at about $1.4T and now aims for about $20B, leaving roughly $9B to close in the coming weeks. The company raised $122B at an $852B post-money valuation in March. Bloomberg did not name the investors in this tranche and tied the lower target to the AI stock selloff and concerns about data center construction after the US midterms (Bloomberg)
Public Markets
NvidiaMkt Cap: ~$4.5T
Nvidia's board opened an independent investigation into allegations that CEO Jensen Huang was the target of a foreign intelligence effort, following reports in The New York Times and the Financial Times. A committee of independent directors, assisted by outside counsel, will oversee it. The board said it has no reason to believe national security or Nvidia secrets were compromised, and Nvidia said Huang will continue as CEO. NBC News reported Nvidia shares fell nearly 5% in early premarket trading Monday. Nvidia reports earnings Nov. 17 (CNBC)
Google$284.78▲ 0.8%Mkt Cap: $3.5T
Google launched two TPUs. The TPU 8tx is built for secure, confidential inference and delivers up to 2x lower latency than the TPU 8t, Google says. The TPU 8i Max puts two TPU 8i chips on one package for double the compute, SRAM and HBM bandwidth (Google)
Emerging
Four-bit KV cache: A paper from Dayou Du, Woon Yee Ng, Guangming Sheng, Nick Lau, Jianyi Cheng and Luo Mai, BitDecoding-FP4: Native FP4 KV Cache Decoding via Algorithm-System Co-Design, stores the KV cache, the memory of earlier tokens a model keeps while generating long text, in 4-bit FP4 instead of BF16. It cuts query logit error from 60% to 1% and stays within 0.5 points of BF16 accuracy with NVFP4 and within 1 point with MXFP4. In end-to-end serving on Nvidia B300 GPUs with Qwen3-4B and Qwen3-30B-A3B, peak throughput is 2.4x to 3.0x that of BF16 KV. The throughput figures are measured on hardware, and the code is public (arXiv)