Flash News

TREE NEWS update: Intel has started shipping samples of its GPU D-Series to customers, part of the chipmaker’s push to advance open infrastructure for the agentic AI era. The company did not disclose which customers are receiving the samples or a timeline for volume production. The move marks Intel’s latest effort to compete in the data-center accelerator market dominated by Nvidia.

TREE NEWS reports: The Trump administration plans to reassign staff from the federal agency responsible for testing artificial intelligence to the White House. The move would shift AI testing personnel into the executive office. No timeline, number of staff affected, or further details were provided.

TREE NEWS update: Binance has rolled out Binance AI, a free generative-UI trading interface, to all English-language users, following a beta that drew roughly 150,000 testers. The tool is available at no cost on the exchange. Binance did not disclose further product details in the rollout.

TREE NEWS reports: Andreessen Horowitz is in talks to expand its equity allocation in a funding round for startup TypeSafe AI. No amount, valuation or round stage was disclosed. TypeSafe AI has not commented on the reported discussions.

TREE NEWS reports: Global smartphone shipments fell 20% year-on-year in the third quarter, the weakest Q3 since 2012. Memory chip costs surged and supply tightened as AI infrastructure projects absorbed supply, pushing up production costs. Manufacturers responded by raising retail prices, cutting shipment volumes and prioritizing profitability over unit sales.

TREE NEWS reports: Meta AI head Alexander Wang said the company’s Muse Wonder models are “true frontier models” that can compete with the current best, particularly in personal and professional agent work. Wang urged skeptics to try them and judge for themselves, saying Meta is deliberately differentiating by focusing on consumer-facing models rather than coding agents, where other ecosystems have concentrated.

TREE NEWS update: Meta Platforms CEO Mark Zuckerberg said the company will further increase spending because it is “already preparing for the next step.” He said the company is catching up with rivals in frontier AI models. No figure or timeline for the additional spending was given.

TREE NEWS update: Meta released two AI models, with CEO Mark Zuckerberg describing Muse Wonder as a larger, more advanced class of model. Muse Wonder is available through Muse Code and the Meta model API, and will be integrated into other products soon. Meta did not disclose model sizes, benchmarks or pricing.

TREE NEWS reports: Bitcoin miners need roughly $50 billion to complete planned AI infrastructure projects, with IREN, Riot Platforms and HIVE Digital facing major funding gaps. Lenders are scrutinizing the deals as financing costs widen across the sector.

TREE NEWS update: Google Cloud announced a preview release of a Looker data application that it describes as a generative AI application with native governance and security capabilities. The preview gives users access to the product ahead of general availability. No pricing, availability date or customer details were disclosed.

TREE NEWS update: Asus unveiled the XA P2N-E2, a 2U server built on NVIDIA’s MGX architecture and powered by the NVIDIA Vera CPU, targeting agentic AI and reinforcement learning workloads. Debuting at the OCP Global Summit 2026, the platform handles execution-layer tasks such as code execution, tool calling, sandboxing, data processing and orchestration. It supports up to two dual-slot NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs.

TREE NEWS update: Nvidia and Cisco have jointly introduced USFP (Ultra Small Form Factor Pluggable), a new pluggable optical module form factor designed to scale networking in AI factories. USFP is the first module to support 400Gb/s electrical signaling per lane and delivers up to 12.8Tb/s per module, letting AI builders connect more computing systems without adding network tiers while cutting power per bit. Bandwidth density is 4x and energy efficiency 1.3x versus existing 200Gb/s-per-lane modules. Nvidia and Cisco are working with TE Connectivity and Amphenol on a multi-source hardware ecosystem.

TREE NEWS update: Nvidia and Cisco jointly introduced USFP, an ultra-dense small form-factor pluggable optical module specification aimed at scaling AI factory networks. USFP is the first module to support 400Gb/s electrical signaling per lane, with single-module capacity up to 12.8Tb/s, letting AI builders connect more compute systems without adding network tiers while cutting energy per bit. It claims 4x bandwidth density and 1.3x energy efficiency versus existing 200Gb/s-per-lane modules. Nvidia and Cisco are working with TE Connectivity and Amphenol on the ecosystem.

TREE NEWS update: Beijing police said they have placed two men under criminal coercive measures for fabricating online rumors that several people at a tech company had been taken away. Guo, 50, and Chang, 51, posted the false information for traffic and industry attention, causing widespread circulation and seriously disrupting public order. The case was confirmed to involve rumors about Moonshot AI’s KIMI.

TREE NEWS update: AI compute operator KONST has closed a $30 million Series B round led by ADATA Technology, with M Mobility and Pegatron Venture Capital among strategic participants. The company said the new capital will fund expansion of its AI data center buildout and accelerate its Token Factory strategy.

TREE NEWS reports: OpenAI is exploring the possibility of relaunching its Sora video-generation model and has held talks with a stealth startup focused on AI video tools for film production, The Information reported on October 12. OpenAI had previously paused development of Sora and its companion TikTok-style video app as part of a plan to cut high-server-cost side projects. Rivals including xAI and Google continue to push ahead with AI video generation.

TREE NEWS update: Flo Health announced a multi-year strategic partnership with Google Health. The two companies did not disclose financial terms or a specific start date for the collaboration. Flo Health operates a women’s health app.

TREE NEWS update: AMD has launched Helion, a rack-scale system designed for agentic AI workloads. The product marks the chipmaker’s push into rack-level infrastructure for running autonomous AI agents at scale. No pricing, availability date or technical specifications were disclosed.

TREE NEWS update: Marvell Technology released its second-generation Structera X, the company’s latest chip for accelerating data-center infrastructure. The product launch follows the first Structera X generation and targets the compute and memory demands of AI-era data centers. Marvell has not disclosed pricing or availability details for the new part.

TREE NEWS reports: Microsoft is preparing to install the newest version of its in-house chips across its data centers, aiming to power its AI software mainly with its own hardware. The company expects to stockpile multi-gigawatt volumes of Maia chips, with Maia 300, designed for inference, rolling out next year. Microsoft says Maia 300 can deliver 40% more tokens per megawatt than leading GPU-based systems, and it currently does not plan to rent the underlying compute directly to cloud customers.