Thinking Machines Lab and NVIDIA are joining forces in a multi-year partnership to deploy over one gigawatt of next-gen NVIDIA Vera Rubin AI systems as early as next year, with NVIDIA making a significant investment to expand frontier AI access for enterprises, researchers, and scientists worldwide.
Nvidia pays a $150 million deposit to seize 600MW of data center capacity in Abilene, Texas after Oracle and OpenAI scrap a planned Stargate expansion following failed financing negotiations and weather-related infrastructure setbacks, with Nvidia reportedly eyeing Meta as a tenant to keep the site exclusively running its own silicon.
OpenAI is abandoning plans to expand its Stargate data center with Oracle in Texas, choosing newer Nvidia GPU clusters elsewhere as the AI industry faces a growing crisis where data centers become obsolete faster than they can be built, leaving Oracle — which carries over $100 billion in debt and …
Apple's M5 Max MacBook Pro, with 128GB unified memory, is enabling developers to run massive 70-billion-parameter AI models locally for free as cloud AI inference costs spiral out of control, with some startups reporting costs tripling in just three months.
Apple's MLX framework now supports the M5 chip's Neural Accelerators, delivering a massive 4x speedup in AI response times and up to 27% faster token generation, powered by the M5's 153GB/s memory bandwidth, enabling MacBook Pro users to run powerful large language models locally with ease.
A deepening AI infrastructure crisis is forcing companies like Anthropic to actively degrade their own products to conserve compute, as surging demand from agentic tools outpaces supply — and with DRAM constraints expected to cap growth at 15GW until 2027, experts warn inference shortages will worsen through 2026 and 2027, …
Andrej Karpathy's new 'autoresearch' tool enables an AI agent to autonomously run roughly 100 LLM training experiments overnight on a single NVIDIA GPU, modifying code, evaluating results, and iterating — all guided by a simple plain-text file instead of manual coding.
MIT researchers unveil Attention Matching, a groundbreaking KV cache compression technique that slashes large language model memory usage by up to 50x in seconds, delivering near-identical accuracy without the slow, compute-heavy optimization methods that have long bottlenecked AI deployment.
Apple's 18-core M5 Max stuns in Geekbench 6 benchmarks, outscoring AMD's 96-core Threadripper Pro, though experts caution the test favors fewer, faster cores and doesn't scale beyond 32 threads — still, the chip's 29,644 multi-core score and 614 GB/s memory bandwidth signal a major leap forward for Apple silicon.
Red Hat is making major strides in open AI infrastructure, emerging as the third-highest global contributor to PyTorch while driving innovations in multi-cloud inference, distributed model serving, and hardware-agnostic AI deployment through projects like vLLM and llm-d.