AI News

AI News Brief: French AI startup ZML has officially launched its inference server software LLMD, aiming to let major open-source LLMs run at near-native or beyond-native peak speeds across chips from various vendors. Supported hardware covers Nvidia GPUs, AMD, Google TPU, Apple Metal, and Intel Arc. Founder Steeve Morin said AI inference is gradually surpassing model training in importance, but the software-architecture divide locks enterprises into specific vendors. ZML’s core pitch is to let enterprises and cloud providers mix and match chips from different sources, cutting costs or energy use to make large-scale AI adoption possible.

AI news brief: Former OpenAI CPO Kevin Weil has joined the board of Seattle-based rocket startup Stoke Space as a formal director, following his earlier role as an angel investor. Weil previously held senior roles at Twitter, Meta, and Planet Labs, and most recently led OpenAI’s accelerated science research program before stepping down in October 2025. Stoke has raised 1.34 billion dollars to date and is targeting first flight this year with its fully reusable Nova rocket.

AI News Flash: Singapore’s sovereign wealth fund Temasek Holdings announced plans to expand its AI-related investment portfolio by approximately 2.5 times over the next five years, with core focus on data centers and semiconductors, reflecting strong confidence in long-term AI infrastructure demand. Temasek stated that there are currently compelling investment opportunities in AI, and despite growing external concerns about a tech bubble, the fund is increasing its bets against the trend, demonstrating an optimistic stance on long-term AI industry development.

AI News Brief: Microsoft has begun replacing some third-party models from OpenAI and Anthropic with its in-house MAI models in core Office software like Excel and Word to handle a portion of user prompts. This strategic shift, reported by Bloomberg, signals Microsoft is actively reducing its reliance on external AI vendors and the associated licensing costs. In fact, Microsoft had previously…

AI News Brief: Decagon CEO Jesse Zhang recently posted a counterintuitive take — the more mature enterprise AI deployments become, the more they tend to switch to lightweight open-source models, yet overall spending on frontier models has barely dropped. He argues frontier and open-source models aren’t competitors but two stages of the same lifecycle — expensive frontier models validate new use cases, and once a scenario matures it’s handed off to cheaper open-source alternatives…

AI News Flash: Meta officially launched Muse Image on Tuesday, a brand-new AI image generator developed by its Meta Superintelligence Labs, internally codenamed ‘Mango’. It’s now available for free through the Meta AI app, Instagram Stories, and WhatsApp. Muse …

AI News Brief: The global AI wave is reshaping a key battleground in the semiconductor industry — the CPU market. CPUs were once viewed as general-purpose computing cores, but as demand for LLM training and inference surges, their role in AI infrastructure is quietly shifting. They’re no longer just supporting players; they’re working alongside GPUs to handle more complex data preprocessing, model scheduling, and system control tasks. Nvidia recently…

AI news brief: MUFG Bank has announced plans to integrate satellite imagery with AI technology to build a brand-new real estate collateral review system, scheduled to officially launch in fiscal year 2027. The core function of this system is to remotely evaluate the real estate collateral provided by borrowers, replacing the traditional approach of having bank staff visit sites in person for inspections. According to the bank’s estimates, the new system can save up to approximately 10,000 hours of manual on-site inspection time per year, significantly reducing the cost of labor-intensive appraisal work.

AI News Brief: Anthropic’s general-purpose knowledge work AI agent tool, Claude Cowork, has officially expanded to web and mobile platforms, now available to Max subscribers. Launched in January this year as a desktop application, this update allows users to initiate tasks on their computer and track progress on their phone. Even when the laptop is closed, the agent continues running in the background and delivers results upon completion. …

AI News Brief: Google announces new features for Gemini API’s Managed Agents, directly responding to developer feedback, with the goal of making agents more suitable for production deployment. The core architecture uses a single API endpoint, with Gemini automatically handling reasoning, code execution, package installation, file management, and web information retrieval in an isolated cloud sandbox, so developers don’t need to manage the underlying infrastructure themselves.