Weekly AI Update OPENAI'S FIRST CHIP TAKES AIM AT NVIDIA'S MOST PROFITABLE TURF OpenAI unveiled Jalapeño, an inference chip built with $AVGO that it says beat $NVDA's Blackwell systems on performance per watt in nearly all tested scenarios, with deployment inside OpenAI's own compute infrastructure expected by year end. AMAZON TRIPLES ITS NVIDIA CHIP ORDER JUST FIVE MONTHS AFTER THE LAST ONE $AMZN is adding another 2 million $NVDA GPUs to AWS data centers across 2027 and 2028, including Blackwell Ultra, Rubin and Rubin Ultra, a deal worth tens of billions of dollars that comes only five months after a commitment of more than 1 million chips. The expanded partnership also pulls $NVDA networking, CPUs, open models and its robotics stack into AWS, even as $AMZN pushes its own competing Trainium and Graviton silicon, now at a $25 billion annualized run rate. APPLE GOES 2 NM WITH M6 AND BUILDS ITS BIGGEST CHIP EVER $AAPL debuted the M6 in a new Mac mini and the M5 Ultra in a new Mac Studio, with M6 becoming the company's first 2 nanometer chip, packing a 12-core CPU, 12-core GPU and a Dual 16-core Neural Engine that doubles peak AI compute versus the prior generation. The M5 Ultra is $AAPL's most powerful chip to date, using UltraFusion to link four dies for the first time, with up to a 36-core CPU, an 80-core GPU, 512GB of unified memory and 1.2TB/s of bandwidth, 50 percent more than M3 Ultra. Both are aimed squarely at running large AI models locally on device rather than in the cloud.