Latest news, developments and reporting on AI Chips & Compute.
Qingmao Intelligence, founded by a Tsinghua team, has built a self-evolving AI infrastructure enabling domestic chips to produce trillions of tokens daily.
ByteDance Seed and Tsinghua AIR unveiled CUDA Agent, an agentic RL system that trains LLMs to write CUDA kernels faster than torch.compile, reaching a 96.8% speed advantage on KernelBench.
Jim Cramer said Monday that the AI boom has rewritten the rules for memory stocks, and their massive rallies may not be over.
CNBC reports China's AI push is driven by national resolve and low costs rather than matching U.S. capital, yet a wide chip gap persists despite progress in low-cost models.
Groq raised $350M led by Disruptive with Nvidia participation, valuing it at $3.5B. The startup pivots from chipmaking to a neocloud operating Nvidia systems.
Nvidia will invest $1.5B in SB Energy and backstop $105B for OpenAI's Ohio data center.
According to a Hugging Face blog post, a constraint-aware GPU allocator improved GPU utilization by up to 33 percentage points and priority-weighted output by up to 105% on identical hardware and workloads, compared to a FIFO scheduler.
Nvidia said it will provide credit and compute for an OpenAI data center in Ohio, with initial 4.25 gigawatts of capacity, as part of growing AI infrastructure financing.
TSMC July sales surged 44.7% on AI demand, beating its 40% annual growth target.
Moore Threads reported H1 2026 revenue of 1.736 billion yuan, up 147.42% year-on-year and surpassing its full-year 2025 revenue, as losses narrowed sharply.
Mistral secured data-center partners, CoreWeave said older Nvidia chips remain valuable; Cerebras shares fell despite strong revenue.
AI-focused hedge fund Situational Awareness invests $400M in chip startup Source Foundry, bringing total to $500M, despite recent portfolio losses.