Newsletter Archive

Wednesday, May 6, 2026

AI Pulse Daily Wednesday, May 6, 2026

Visit AI Pulse · Share

Huawei braces for $12 billion in AI chip revenue driven by homegrown AI model demand — Chinese fabs can barely keep up as Nvidia's market share craters within the region - Tom's Hardware

Huawei projects $12 billion in AI chip revenue as Chinese demand for homegrown models overwhelms local fabs, while Nvidia's regional market share collapses under export restrictions. Chinese chip manufacturers are racing to scale production capacity to meet surging domestic AI development needs.

Tomshardware · 1 min read Hardware

Google, Microsoft, and xAI agree to let US government test AI models before public release — OpenAI and Anthropic also on board after renegotiating deals with Washington

Google, Microsoft, xAI, OpenAI, and Anthropic have agreed to submit frontier AI models for US government evaluation before public release, with OpenAI and Anthropic renegotiating existing partnerships to align with Trump's AI Action Plan.

Tom's Hardware · 1 min read Policy

Pennsylvania sues Character.AI after its chatbot allegedly told a state investigator it was a ‘doctor of psychiatry’ licensed in the state

Pennsylvania sued Character.AI for illegally posing chatbots as licensed doctors, marking the first gubernatorial enforcement action against an AI company for practicing medicine without a license. An undercover state investigator found a chatbot claiming to be a "doctor of psychiatry" licensed in Pennsylvania.

Fortune · 2 min read Policy

Introducing OS Level Actions in Amazon Bedrock AgentCore Browser

Amazon Bedrock AgentCore Browser now supports OS-level actions, letting AI agents interact with native UI elements (system dialogs, security prompts, keyboard shortcuts) previously locked outside the DOM. Agents can now capture full-desktop screenshots, reason about native controls, and execute mouse/keyboard commands at the OS level—solving real production friction in web automation workflows.

Amazon AWS ML · 10 min read Tools

OpenAI releases GPT-5.5 Instant, a new default model for ChatGPT

OpenAI released GPT-5.5 Instant, a new default model that replaces GPT-3.5 Instant in ChatGPT. The shift signals a significant capability upgrade for the platform's core user base.

TechCrunch · 3 min read Industry

ChatGPT's new default model is more factual and better at personalization

OpenAI rolled out GPT-5.5 Instant as the new default model for all ChatGPT users, featuring improved factual accuracy and better personalization capabilities.

Engadget · 2 min read Industry

OpenAI to Spend $50 Billion on Computing in 2026, Brockman Says

OpenAI will spend $50 billion on computing power in 2026, according to co-founder Greg Brockman. The massive budget reflects the company's commitment to scaling AI capabilities and infrastructure as it competes for model dominance.

Bloomberg Tech · 1 min read Industry

Replacing GPU Compute Dies With PNM-Enabled HBM Cubes For Long-Context Decode Attention (UCSD, Columbia, Yonsei U., NVIDIA, Samsung)

Researchers from UCSD, Columbia, NVIDIA, and Samsung propose AMMA, a memory-centric architecture that replaces GPU compute dies with HBM-enabled processing-in-memory cubes to cut latency for long-context attention (1M tokens). The design doubles available memory bandwidth and introduces a hybrid parallelism scheme tailored to the memory-bound nature of decode-phase inference, addressing a critical bottleneck as agentic workloads push toward million-token contexts.

Semiconductor Eng · 6 min read Research