Category
Envision has opened its Galaxy Campus in Inner Mongolia, integrating renewable power generation, transmission, and battery storage with AI computing infrastructure in a single design. The 2GW-plus capacity and million-chip figure are design targets for full completion, leaving current deployment scale and renewable supply performance as the next points to verify.
From late July to early August 2026, OpenAI, Anthropic, and Meta successively disclosed deviations by AI models during testing. In all three cases, the cause was not a rogue model but a failure of network isolation in the evaluation environment—exposing a structural problem: the infrastructure needed to safely measure the offensive capabilities of frontier models is still immature.
A survey of 1,500 companies in the UK, France, and Germany found that 73.9% worry about the US government cutting off their access, and 54.5% say they would grind to a halt within one business day. Yet only 44% have regularly rehearsed backup plans, and the fact that even authentication infrastructure is American-made is blocking any escape.
AMD has announced an agreement to acquire Taalas, whose technology fixes model weights into silicon. The plan is to combine it with Instinct and Helios, but with regulatory approval still pending, supported models, performance, and connection methods have not yet been disclosed.
OpenAI is switching the default model for ChatGPT's Free and Go plans to GPT-5.6 Luna, and will lift text conversation limits starting next week. Limits on images, files, and tools remain in place.
Reports suggest ASUS, GIGABYTE, and MSI plan to raise motherboard distribution prices in Q3 2026. While PCB costs are expected to rise 50%, the actual price increase for finished products remains unannounced.
ZEALS's newly announced semi-humanoid "D1" deliberately abandons bipedal walking, adopting a dual-arm design mounted on a wheeled base to achieve a starting price of 5 million yen. By placing the collection and learning cycle of field data at the core of its business, the company is exploring a path to humanoid commercialization from Japan that differs from the routes taken by the US and China.
While Microsoft touts an AI run rate topping $37 billion annually, Bloomberg estimates that 65-70% of it traces back to OpenAI. That figure uses a different denominator than total company or Microsoft Cloud revenue—here's how to read dependency through the contracts and multi-model shift.
Apple has built up component inventory to $7.645 billion, while TSMC's days of inventory rose to 87 amid its N2 ramp-up. We separate the reported $1 billion figure for stranded Apple-bound chips from what both companies have officially disclosed.
DeepSeek matched OpenAI's 80% price cut within hours, but couldn't withstand a surge to 7 trillion tokens per week and has now notified users of a "significant increase" in API pricing. A computing ceiling of roughly 20,000 GPUs exposes a structure in which the winner of a price war can turn into its loser.
