Daily AI News | Naver signs $9 billion AI infrastructure agreement with Brookfield
# Daily AI News | Naver signs $9 billion AI infrastructure agreement with Brookfield
## Lead Story
### Naver signs $9 billion AI infrastructure agreement with Brookfield
Naver signed a $9 billion agreement with Brookfield for AI computing infrastructure. The deal reportedly satisfies a prerequisite for Nvidia’s planned $1 billion strategic equity investment and advances the partners’ 1-gigawatt global AI factory project.
**AI take:** The agreement links compute construction with equity investment; signing, closing and operation remain separate milestones.
## More News
### Google releases Genie 4 world model in research preview
Google DeepMind released Genie 4 in research preview, enabling text-driven scene changes during simulation and synchronized music, ambient sound and interactive dialogue. TechCrunch reports that scenes can remain consistent for up to two minutes, while details can still drift during longer runs.
**AI take:** Live editing broadens creative experimentation, while persistent scene consistency remains a practical constraint.
### Oracle and Nvidia expand Vera Rubin infrastructure deployment
Oracle and Nvidia plan to expand OCI Superclusters using Vera Rubin NVL72 systems to multi-gigawatt scale for AI training, inference and agent applications. They also announced general availability of Vera Rubin NVL72 and Vera CPUs on OCI.
**AI take:** Available products and future construction are distinct; the planned scale does not mean the expansion is already complete.
### Nvidia and Meta expand collaboration on Muse models
Nvidia and Meta announced expanded collaboration on Muse and its model family to improve training and inference performance. The companies linked the work to growing demand for useful AI agents and outlined the direction of their continuing partnership.
**AI take:** Joint model and compute optimization is the focus; subsequent results will show how the goals translate into practice.
### Baidu open-sources Unlimited OCR 2 for long documents
Baidu released and open-sourced Unlimited OCR 2 for multi-page and very long document parsing. Launch coverage describes a demonstration processing all seven Harry Potter books, totaling more than 3,600 pages; the reported results come from the release materials.
**AI take:** Long-document support can reduce batching, while accuracy should be assessed on the documents a user actually handles.
### Liquid AI introduces the LFM3 multimodal model family
Liquid AI introduced LFM3, bringing text, vision, audio, omni and decision models onto a shared backbone in the roughly 2-billion-parameter range. The company says developers can fine-tune on their own data using one GPU and deploy where their application runs.
**AI take:** A shared backbone offers options for specialization; single-GPU fine-tuning claims still depend on the task and configuration.
### Ant Bailing releases Ling-3.1 enhanced and base weights
Ant Bailing released weights for Ling-3.1-flash-1019 alongside Ling-3.1-flash-base. The report states that both are available on Hugging Face and ModelScope under the MIT license, providing open weights for use and further development.
**AI take:** Releasing base and enhanced versions gives developers different starting points for adaptation.
### Anthropic expands government agreement with 40% Claude discount
Anthropic expanded its GSA partnership with a 40% discount on Claude Enterprise, Claude for Government and the Claude API for government partners. It extends to state, local, tribal and territorial governments, covering direct purchases now and cloud-provider purchases in coming months.
**AI take:** The offer spans products and procurement channels, so eligibility and channel availability need to be checked separately.
### OpenRouter announces CSAM screening across inference requests
OpenRouter said it will screen all inference requests passing through its service for child sexual abuse material during processing. Potentially flagged content may be retained for human review, with reporting and preservation where legally required.
**AI take:** The change affects request handling and retention, making the scope of human review relevant to service users.
### Opera introduces WebAI for local browser models
Opera introduced WebAI, a proposed web standard intended to let users run a variety of AI models directly in the browser without relying on cloud inference. The announcement presents it as a route to deploying models locally within the browser.
**AI take:** Local execution changes where computation happens; developers still need to check support for their chosen models and browsers.
### Apple rolls out iOS 27.2 with expanded Siri AI languages
Apple began rolling out iOS 27.2. TechCrunch reports expanded language support and improved memory for Siri AI, alongside message suggestions and Apple Pencil support for iPhone Duo.
**AI take:** Language coverage and memory affect everyday assistant use; availability should be checked on the user’s own device.
### Apple unveils M6 iMac starting at 2.49 million won in Korea
Apple unveiled a new iMac powered by the M6 chip. Launch reporting describes improved AI and graphics performance plus Wi-Fi 7 and Bluetooth 6 connectivity, with a Korean starting price of 2.49 million won.
**AI take:** The update combines compute and connectivity changes; the Korean starting price is not a quote for other markets.
### KT schedules Token Factory model platform for October 30
KT unveiled Token Factory, scheduled for October 30, integrating more than 500 domestic and international AI models with usage and cost management. Reports describe task-based model selection and token-based billing; the announcement precedes the planned launch date.
**AI take:** A unified entry point can simplify model and cost comparisons; routing quality remains to be tested after launch.
### LINE Yahoo launches Agent i Biz for businesses
LINE Yahoo launched Agent i Biz and a Business Manager product-management feature. The agent uses conversational input in a dedicated web app and company data to support market analysis, advertising and LINE official-account operations; product management helps expose registered product information in search and ads.
**AI take:** The agent targets concrete business workflows, where usefulness depends on connecting business data with operational tasks.
### Doosan plans 399.2 billion won investment in AI data-center materials
Doosan plans to invest 399.2 billion won in Gimje, South Korea, in a dedicated facility for high-value copper-clad laminate used in optical modules. The project aims to expand material supply for high-speed data transmission in AI data centers and remains an investment and construction plan.
**AI take:** Compute expansion also creates demand for interconnect materials; new capacity depends on construction and delivery.
### Shanghai issues action plan for AI-enabled manufacturing
Shanghai’s economic and information technology commission published a new intelligent-manufacturing action plan centered on AI integration and smart factories. It targets smart-factory coverage and widespread AI application across key manufacturers by 2028, alongside humanoid robot deployment in factories.
**AI take:** These are policy targets; realized factory applications and equipment deployment will show the implementation progress.
### GXO expands humanoid robot deployment to two more US warehouses
GXO Logistics is expanding its partnership with Agility Robotics to deploy humanoid robots at two additional US warehouses. The expansion follows test operations in Georgia and extends the application beyond the earlier site.
**AI take:** Multi-site deployment moves beyond a single demonstration; operational results will help show its repeatability.
### Rokid unveils Color AI glasses starting at 4,999 yuan
Rokid unveiled full-color Rokid Color AI glasses and the Rokid Ring controller. Launch prices are 4,999 yuan for black 2GB+64GB glasses, 5,499 yuan for tortoiseshell 4GB+64GB glasses, and 1,699 yuan for the ring.
**AI take:** The glasses and ring offer a combined interaction setup, but their purchase prices are separate.
### Mazda adds Google Gemini support to the new CX-5
Mazda announced Google Gemini support for new CX-5 models specified for the US, Canada, Europe and Japan. The required update is offered free through Google Play, adding Gemini to Google services on the supported vehicle versions.
**AI take:** Support covers specified versions of the new vehicle; a free update does not imply compatibility with every older model.
### AiMOGA unveils Mobo companion robot and opens pre-orders
AiMOGA Robotics unveiled Mobo, described as its first consumer-grade family companion humanoid robot, and announced worldwide pre-orders. The announcement establishes the debut and ordering stage rather than completed delivery across all markets.
**AI take:** Pre-orders start the commercialization process; delivery arrangements and use in homes remain the next practical checks.
### Vuzix signs distributor and starts shipping LX1 glasses order
Vuzix signed a distribution agreement with Denver-based Commercial Electronics Systems and said it had received and begun shipping an initial six-figure order for LX1 smart glasses. The announcement describes an order value, not a six-figure number of devices.
**AI take:** Shipment marks a concrete commercial step, but order value alone does not establish unit sales.
### General Compute raises $100 million Series A
General Compute raised a $100 million Series A led by Blackbird Ventures and Abstract Ventures. TechCrunch reports that a major customer deal worth $100 million in annual recurring revenue helped enable the round, which values the company at $800 million post-money.
**AI take:** Funding, contracted recurring revenue and valuation are different measures; the deal value is not the same as realized revenue.
### Nvidia backs Simplismart in $14 million funding round
Indian AI software startup Simplismart raised $14 million in a round backed by Nvidia. TechCrunch describes software intended to improve model-serving efficiency and, in some cases, reduce GPU requirements; that does not establish savings for every deployment.
**AI take:** Software optimization is distinct from adding hardware, and resource savings need to be checked for each deployment.
### Calix releases communications language model on Hugging Face
Calix released a communications language model on Hugging Face and plans task-specific Calix Gems for Calix One customers in 2027. Initial planned uses include security threat detection, subscriber-experience optimization and network troubleshooting; the later products remain planned.
**AI take:** The release targets operational network tasks, while the available model and next year’s product plans are separate developments.
### Crunchbase reports $9.9 billion in third-quarter robotics funding
Crunchbase reports global venture funding for robotics startups of $9.9 billion in the third quarter of 2026, versus $4.3 billion a year earlier, with most investment going to companies in China and the US. The comparison concerns quarterly funding rather than industry revenue.
**AI take:** Funding measures investment activity, not product revenue or proof of commercial success.
### Atlas Bio raises $12.5 million for AI drug-trial simulation
Atlas Bio raised $12.5 million in a round led by Creandum to simulate drug trials using AI. The report establishes the funding and development direction, without showing that the simulations have replaced real clinical trials or received regulatory acceptance.
**AI take:** Funding for simulation development is distinct from clinical validity and does not establish that trials can be replaced.
### Invesco works with Anthropic on Claude tools for financial advisors
Invesco announced work with Anthropic to make portfolio intelligence and practice-management data and capabilities available through Claude for Financial Advisors. Its planned connector is intended to provide an advisor toolkit; the announcement describes collaboration and future integration.
**AI take:** Specialized data can bring assistants closer to advisor workflows, while a partnership announcement does not make every feature immediately available.
## Rumors
