AI Industry Development & Policy Dynamics

According to Google's announcement, the company released Gemini 3.6 Flash and Gemini 3.5 Flash-Lite today, with improved efficiency and lower costs than the previous generation. Gemini 3.6 Flash uses 17% fewer output

According to Axios, Anthropic's spending on U.S. government lobbying in the first half of 2026 has already surpassed its total expenditure for the entire 2025 fiscal year.

According to Wall Street Journal, CoreWeave's chief executive recently stated the company is deploying AI infrastructure at scale and is on the right path in building AI-native cloud services. The CEO also noted that as AI cloud revenue grows, the accounting impact of depreciation will

Advanced Micro Devices Inc. (AMD) received a price target increase to $640 from $511 by GF Securities ahead of its Advancing AI 2026 conference beginning July 22. The brokerage reiterated its 'Buy' rating and raised full-year revenue estimates to $52.3 billion for 2026 and $82.2 billion for 2027, ci

According to The Information, Google is developing a server chip codenamed Frozen v2 designed specifically to run Gemini faster and cheaper by hardwiring part of the AI model's architecture directly into the silicon. Engineers project a six to ten times improvement in efficiency compared to

OpenAI and Anthropic increased their federal lobbying spending to record levels in the second quarter of 2026, according to federal lobbying disclosures. The two AI developers spent $3.17 million combined during the quarter, up 23% from the first three months of the year, as they sought to

According to federal lobbying disclosures, OpenAI and Anthropic combined to spend $3.17 million on federal lobbying during the second quarter of 2026, a 23% increase from the first quarter. Anthropicspent $1.97 million from April through June, up 26% from Q1, while OpenAI spent $1.2 million,

According to Nvidia on Tuesday, the company's new Vera CPU can deliver up to 1.8x higher performance on agentic AI workloads compared to traditional x86 CPUs. The chip's custom Olympus cores were designed to maximize single-thread performance for branch-heavy, latency-sensitive workloads typical

According to Nvidia, on July 21, Vera CPU achieved up to 2.2x faster orchestration speed for AI agentic workloads compared to competing CPUs, while supporting 1.6x more concurrent agent execution. Testing by DeepInfra showed the improvements in handling complex agent operations. DeepInfra

According to Nvidia announced, the Vera Rubin platform for gigascale AI factories is expanding globally with CoreWeave, Google Cloud, Microsoft Azure, and Oracle Cloud now deploying Vera Rubin NVL72 systems. CoreWeave testing shows Vera Rubin NVL72 achieves 10x higher token throughput per

According to multiple analysts from BluBird Capital, Wedbush, and D.A. Davidson, Moonshot AI launched the open-weight model Kimi K3 last Thursday with 2.8 trillion parameters, surpassing some Western models including OpenAI's GPT-5.6 Sol and Anthropic's Claude Fable 5 in certain benchmarks. Due to s

According to Nvidia, on July 21 the company unveiled Spectrum-6 Ethernet switching system designed for next-generation massive AI factories based on the Vera Rubin platform. The system delivers 102.4Tbps switching capacity, double that of its predecessor, and is engineered to connect hundreds of

According to Bitcoin.com News, SpaceX (NASDAQ: SPCX) closed below $120 on July 20 after seven consecutive losing days, falling more than 11% below its $135 IPO price and roughly 47% from its post-listing peak of $225.64. The decline has erased over $1 trillion in market value. Economist Peter

According to Erik Voorhees, on July 21, Venice API enabled keyless access to on-chain data across Base, Solana, and Robinhood Chain. Agents can now call model inference through the API without requiring additional authentication

Today (July 21), Ethereum co-founder Vitalik Buterin proposed creating a new high-level programming language that compiles to formal proof systems like Lean and HOL, optimizing readability of definitions and theorems rather than proof processes themselves. According to PANews, Buterin stated the lan

According to Guru Club, China's first compliant embodied AI data platform, jointly developed by Shanghai Technology Exchange, Kexing Spacetime Technology, and Unitree Robotics, officially began operations on July 21. The platform is positioned as the country's first embodied AI data infrastructure d

Nvidia released detailed specifications and benchmarks for its Vera data center CPU on Tuesday, entering competition with Advanced Micro Devices and Intel in the AI server processor market. The chip giant delivered Vera processors to clients including OpenAI, Anthropic, and SpaceX in June,