AI Industry Development & Policy Dynamics

According to Meta CEO Mark Zuckerberg, the company is considering renting out AI computing capacity to external customers, as external offers have been high enough that Meta believes rental may provide more value than internal use. Zuckerberg emphasized that Meta has not experienced AI capacity

According to Beating, Kimi Code launched K2.7 Code HighSpeed mode today, exiting beta as a permanent feature. The mode runs on the same underlying model as the standard version, delivering output speeds of approximately 180 Token/s for regular coding tasks and up to 260 Token/s for short-context

According to a July 10 Chosun Daily report, Samsung Electronics supplied prototypes of its GAIA AI PC accelerator to Lenovo and HP for performance testing ahead of planned mass production in 2027. GAIA is built on a 4-nanometer process and features a neural processing unit designed for generative A

OpenAI’s latest flagship model, GPT-5.6 (Sol), was “held” by the U.S. government for 12 days between completing training and its public release; Sam Altman said on CNBC that the process involved multiple rounds of talks with officials including Commerce Secretary Lutnick, Treasury Secretary Bessent, and National Cyber Director Cairncross, but OpenAI refused to disclose specific details. Sam Altman: The review involved multiple rounds of talks with the commerce secretary, treasury secretary, and

According to Goldman Sachs, Zhipu Technology (02513) surged over 11% following its announcement of a rights offering yesterday, but pulled back significantly this morning, trading down 7.4% to 1,882 yuan by 10:44 a.m. HKT, with intraday losses reaching as much as 9.7%. Goldman Sachs initiated

Kartilyah.ph, a legal search platform built in the Philippines, has recently launched a free website allowing users to search Philippine statutes, codes, and Supreme Court decisions in plain language. Built by lawyer and technologist Karlo Dizon, the platform provides direct answers with each

Researchers from Tel Aviv University, the Israel Institute of Technology, and Intuit have revealed a new attack method called "Adversarial HalluSquatting" in the paper "Beware of Proxy Zombie Networks: Achieving Scalable Unidirectional Promptware Attacks via Universal and Transferable Adversarial HalluSquatting," which exploits AI hallucination phenomena to trick AI agents into downloading malicious code. HalluSquatting Attack Mechanism: Technical Principles of Predicting AI Hallucination Resour

According to Critini Research analyst Jukan, heard at ICML, Meta has completed an internal model with capabilities reaching Anthropic's Mythos 5 level, with potential deployment within the coming months. Mythos 5 is Anthropic's most advanced model to

According to Meta, Muse Spark 1.1 launched on July 9, marking the company's first AI model offered to developers on a paid basis. The model features API pricing of $1.25 per million input tokens and $4.25 per million output tokens—approximately one-quarter the cost of OpenAI and Anthropic's

After SpaceXAI released Grok 4.5, Musk described it on X as an "Opus-level product, faster and cheaper," and later added that "it seems to rank first in at least some benchmark tests"; however, Musk later admitted that this comparison was against the previous generation Claude Opus 4.8, not the current one. Grok 4.5 costs $2 for input, while Claude Opus 4.8 costs $5 According to SpaceXAI's official pricing and Artificial Analysis data, Grok 4.5's API pricing is $2 per million tokens for input an

According to Beating, Meta announced yesterday that it will launch its self-developed AI chip Iris in September and expand compute capacity from 7GW this year to 14GW by 2027. The company also began charging developers for access to its Muse Spark 1.1 large language model via API, marking its

Meta Platforms, Inc. released the Muse Spark 1.1 AI model on Thursday, while Cathie Wood's ARK Investment Management purchased 34,080 Meta shares valued at $20.6 million. The AI model represents Meta's most capable release yet for coding and agentic tasks, marking the company's push to compete in th

Meta released the Muse Spark 1.1 AI model on Thursday, its most capable model yet for coding and agentic tasks. ARK Invest disclosed late Thursday that it acquired 34,080 Meta shares valued at $20.6 million. META stock rose 4.7% on Thursday and added 1.3% in overnight trading, moving closer to its 2

According to UBS, global memory chip sales reached $74.6 billion in June, hitting an all-time high with a month-over-month increase of 31.7%. DRAM sales accounted for approximately $48 billion (up 27.7%), while NAND Flash surged to $25.8 billion (up 40.7%), driven by rising AI data center demand. U

According to ChainCatcher, on July 5, ANTON held a closed-door Web4.0 ecosystem meeting in Hong Kong. MY Group CEO Zhang Dingwen attended and announced that MY Group will provide ecosystem resources, global market capabilities, and capital support to ANTON. The two entities will deepen

According to Beating, OpenAI's AGI deployment CEO Fidji Simo announced she is transitioning from full-time to part-time advisor role. Simo, who joined OpenAI in 2025, took a three-month medical leave beginning in April due to a neurological immune condition and determined recovery will take longer t

According to CEO Kim Yeon-seop's statement at a July 9 investor relations day, Lotte Energy Materials will expand AI chipset copper foil production capacity to 16,000 tons by the first half of 2027, up from the current 3,700 tons. The company aims to capitalize on structural supply shortages in AI-d

According to Decrypt, researchers from Tel Aviv University, Technion-Israel Institute of Technology, and Intuit discovered a new attack method on July 10 called 'adversarial hallucination injection' that exploits AI model hallucinations to trick AI agents into downloading malicious code and

According to Bernstein's July 9 report on memory interface chips, AI's shift from training to inference is driving server CPU shipment growth of 24% compound annual growth rate through 2030, while expanding memory interface chip TAM threefold to $20

According to TechCrunch, on July 10, Sequoia partner David Cahn updated his analysis on AI infrastructure investment returns. He calculated that global AI infrastructure spending will reach approximately $1.5 trillion in 2026. When accounting for data center operating costs and operator profits,