Retail News CRM

Tag: cloud computing

  • Five Firms Sign Agreement to Develop Satellite Mega-Constellation

    Five Firms Sign Agreement to Develop Satellite Mega-Constellation

    NTT DOCOMO BUSINESS signed an agreement with four major corporations to explore developing and operating a satellite mega-constellation for commercial communications and space-based computing.

    The five-company partnership unites NTT DOCOMO BUSINESS, Mitsubishi Electric Inc., NTT Inc., MUFG Bank, Ltd., and Development Bank of Japan Inc. To pool technology, customer networks, and financing.

    Under the memorandum of understanding, the group is evaluating the commercial and technical viability of orbital network infrastructure. Target services include telecommunications, Earth observation, positioning, navigation, and timing (PNT), alongside space data centers designed to process data directly in orbit.

    Splitting Infrastructure Across Five Balances

    The consortium divides responsibilities along established corporate lines. NTT and NTT DOCOMO BUSINESS handle end-to-end network architecture, routing satellite traffic into ground facilities, and testing NTT’s proprietary Innovative Optical and Wireless Network optical technology against international specifications.

    The alliance aims to sell services to domestic and overseas enterprise clients across manufacturing operations.

    In order to turn the vast amounts of data collected in space into value for industry and society, it is essential to ensure smooth coordination between the communications networks that deliver the data and the ICT infrastructure used to process and analyze the data. Using the expertise we have developed through our IOWN® initiative and our AI-Centric ICT Platform® concept for a next-generation ICT platform optimized for the AI era, NTT DOCOMO BUSINESS looks forward to working with the other participating companies to explore commercial opportunities in areas including satellite communications and space-based data centers. By creating communications and data-utilization infrastructure that connect space and Earth, we hope to support AI adoption by enterprises for new business development.

    What Orbital Computing Changes for Operators

    Telecom operators face flattening domestic mobile average revenue per user and rising terrestrial infrastructure expenses. Building an independent sovereign satellite network allows NTT to bypass foreign constellation operators for enterprise backhaul, maritime fleet monitoring, and rural redundancy.

    For logistics operators, space-based data centers offer lower-latency observation feeds for trade flows and factory asset tracking without depending on fragmented local fiber routes across developing markets. The central commercial risk rests on unit launch economics, where Western constellation operators hold cost advantages through reusable rocket fleets that heavy industry has yet to match at scale.

    From Ground Fiber to Orbital Routing

    NTT had already mapped its optical network architecture toward low-power data center routing, making space payloads a natural technical extension rather than a pivot.

    The five companies are now formulating technical baselines and enterprise partner frameworks, with initial feasibility studies for constellation communications and Earth observation services to follow.

  • South Korea Secures 10,000 Nvidia GPUs as Contracted Deliveries Stay on Schedule

    South Korea Secures 10,000 Nvidia GPUs as Contracted Deliveries Stay on Schedule

    South Korea will receive its contracted supplies of Nvidia graphics processing units on schedule despite an investigation involving chief executive Jensen Huang, Seoul’s science ministry confirmed on Monday.

    A batch of 10,000 GPUs funded under a 2025 supplementary budget has arrived and entered service in local networks.

    Vice Minister of Science and ICT Ryu Jemyung updated a parliamentary budget committee on the supply line, stating that additional hardware planned for 2026 remains under contract and will deploy this year.

    “We expect production and the delivery schedule for contracted volumes to proceed without disruption,”

    Ryu told lawmakers. The ministry is monitoring corporate developments while working through pricing and delivery timetables with Nvidia Korea.

    Contract Security for Digital Infrastructure

    The clarification gives enterprise platforms and state compute facilities immediate operational certainty. South Korean commercial cloud providers, e-commerce platforms, and logistics networks rely heavily on Nvidia architecture to run automated sorting, predictive inventory systems, and customer-facing algorithms. A pause in hardware allocations would force operators to delay software rollouts and stall data centre buildouts.

    Hardware backlogs in East Asia typically hit mid-tier tech firms hardest, since sovereign contracts and hyperscalers secure priority delivery queues. With government-backed allocations safe, Seoul avoids an immediate bottleneck across its state-funded research initiatives. The risk now concentrates around spot market availability and future volume commitments if corporate governance issues distract supplier roadmaps.

    Executive Disruptions in Seoul

    Nvidia’s board opened an independent investigation on Sunday following reports that chief executive Jensen Huang was targeted by a Chinese intelligence operation through businesswoman Alice Huang. The company board stated it has no reason to suspect US national security or corporate trade secrets were breached, and Huang has pledged full cooperation.

    The fallout hit Nvidia’s commercial engagements in Seoul immediately. Madison Huang, an Nvidia senior director and the chief executive’s daughter, cancelled multiple scheduled appearances across the capital city citing unavoidable internal circumstances.

    South Korean artificial intelligence startup Rebellions called off a planned headquarters tour and an executive dinner with her. Transport operator Kakao Mobility also dropped a planned photo call and media session, though it said a business agreement with Nvidia would proceed with senior director Gaurav Agarwal attending instead.

    Enterprise Engagements Hold Schedule

    The operational calendar across the domestic tech sector remains largely intact despite the cancelled executive briefings. The Nvidia AI Day Seoul industry conference was proceeding as planned with about 2,000 participants expected to address autonomous infrastructure, agentic systems, and physical computing frameworks.

    Domestic software developers and autonomous mobility platforms need steady access to hardware roadmaps to validate next-generation vehicle fleets and warehouse robotics. Kakao Mobility’s decision to preserve its signing ceremony indicates corporate transport operators will protect technical integrations even when partner optics turn volatile.

    Government procurement officers are now locking down final dispatch dates for the remaining 2026 allocations with Nvidia Korea to keep municipal computing centres on schedule.

  • China Mobile Cloud Signs Zhipu to Link GLM Models and Build Coding Tools

    China Mobile Cloud and Zhipu signed a cooperation pact on November 6 to connect the GLM model family to the operator’s MoMA service platform.

    The agreement establishes a joint development center focused on token generation efficiency, targeting domestic hardware adaptations alongside financial services and government contracts.

    Technical teams from both companies plan to tune Zhipu’s models specifically for Chinese-made computing processors. The partnership outlines shared research into inference performance, yet neither company published baseline efficiency figures or an operational schedule for customer deployments. Access arrangements, service-level commitments and supported versions within the GLM lineup were also omitted from the joint filing.

    Hardware Adaptation and Coding Tools

    Beyond platform hosting, the two companies plan to develop joint AI coding applications for commercial software engineers. The announcement outlined an integrated product program without releasing software names, development roadmaps or public testing dates.

    For enterprise buyers, the lack of a timetable leaves procurement teams waiting on functional specifications. Developers evaluating enterprise assistants must weigh unreleased software against tested tools already running on commercial clouds across mainland China.

    Rival Hyperscalers Build Token Tariffs

    Commercial cloud operators in China are locking enterprise accounts into aggressive subscription packages rather than unpriced development agreements. Alibaba Cloud already offers its AI Token Plan starting at $6 a month on its Model Studio service, bundling its proprietary Qwen lineup alongside DeepSeek and GLM models. Alibaba also sells dedicated enterprise engineering assistants, such as Qoder, directly to corporate IT departments.

    State-backed telecom operators often hold strong connections to public sector budgets, but commercial developers demand transparent pricing and tested inference pipelines. China Mobile Cloud must show that domestic silicon can match commercial cloud economics on cost per million tokens. Without published benchmarks, enterprise procurement officers will treat the venture as an exploratory project rather than an immediate supplier switch.

    From Transport Compute to Platform Hosting

    China Mobile has been steering its infrastructure divisions toward specialized vertical computing workloads. The carrier previously launched an automotive AI foundation featuring 112 exaflops of computing power and six specialized agents for vehicle manufacturers. Adding Zhipu’s models to MoMA extends that infrastructure strategy into general business software, banking risk engines and municipal administration.

    Commercial validation now hinges on the release of concrete access tiers, supported GLM versions and billing schedules on China Mobile Cloud’s MoMA catalog.

  • Stolen Cloud AI Keys Hit Retailers and Tech Groups with Million Dollar Bills

    Stolen Cloud AI Keys Hit Retailers and Tech Groups with Million Dollar Bills

    Intruders who breach corporate cloud accounts are quietly activating dormant artificial intelligence tools and running models at company expense, running up bills exceeding 100,000 dollars in a day.

    Investigations carried out across 2026 by Mandiant and the Google Threat Intelligence Group revealed intrusions across Asia Pacific, North America and South America that reached cumulative costs of up to 1 million dollars before victims spotted the leaks. The compromises struck retail, transportation, financial services and technology companies that had adopted large language model architectures into their systems.

    How Illicit Resellers Drain Foundation Models

    The campaign, dubbed LLMJacking, relies on a distinct commercial division of labour within cybercrime rings. One group specializes in penetrating corporate environments to harvest credentials, while a separate tier operates illicit platforms known as LLM Mixers to resell that raw model compute to third-party subscribers.

    One service, trading under the name Jet Night, offered paying customers unlimited subscriptions for about 60 dollars a month to access premium foundation models. Comparable commercial access carries legitimate pricing between 100 and 250 dollars. Security investigators determined the syndicate could only sustain that price differential by powering the backend with stolen enterprise credentials.

    During a single intrusion linked to the network, an unauthorized actor channeled more than 500,000 queries to foundation models in under eight hours. The targeted enterprise absorbed an unauthorized token bill approaching 1 million dollars before discovering the drain. Jet Night subsequently recorded sudden outages on its Telegram broadcast channels whenever compromised businesses identified the breach and revoked the active keys.

    Another unauthorized syndicate set up a lookalike domain disguised as an internal corporate host, routing the victim organization’s application programming interface quotas into a commercial resale front operating as newapi.pro.

    The Shift From Cryptojacking to Token Hijacking

    Retailers and consumer tech operators have spent years guarding compute fleets against illicit cryptocurrency mining, but model hijacking works at vastly accelerated speed. Cryptojacking trickles away compute slowly to evade system detection, whereas LLMJacking generates massive spikes in fraudulent cloud charges in a matter of hours.

    Investigators noted that actors reach their goals through token proxying, automated model streaming and unauthorized graphics processing unit hardware provisioning. Rather than manipulating consumer-facing chatbots to spit out rogue lines of code, the attackers bypass the store front-end entirely to tap backend infrastructure.

    For retail and digital commerce platforms in Asia that connect real-time product search, automated customer query desks and dynamic inventory to foundation models, the financial exposure sits directly in the operating budget. Cloud quotas linked to corporate credit accounts or open enterprise billing agreements allow perpetrators to cash out instantly through black-market resale while the operator absorbs the invoice.

    Stolen Keys and Exposed Gateways

    Initial penetration rarely requires novel zero-day attacks. Threat groups scrape publicly exposed GitHub repositories for stray application programming interface keys, running minute test queries designed to avoid triggering usage alerts.

    Most intrusions stem from compromised cloud infrastructure credentials, including unprotected service account keys, active session tokens and corporate GitHub authorization tokens. A group tracked as UNC6993 gained access to a commercial cloud instance via a misconfigured service account credential, quickly bridging across into internal model endpoints. The same operators established a fraudulent registration portal for model developers, tricking visiting engineers into submitting live access tokens that circumvented multi-factor authentication controls.

    Once inside a corporate network, attackers search database tables and data warehouse systems for deeper administrative secrets, occasionally disguising backdoor accounts as internal cost-optimization utilities. In another instance, attackers targeted a vulnerability logged as CVE-2026-40217 inside a LiteLLM routing gateway, escalating basic user credentials into root administrative control.

    Dormant Services Turned on by Intruders

    In multiple forensic investigations conducted by Mandiant, the targeted companies had left their high-tier machine learning services deactivated. Intruders used administrative cloud credentials to switch on the dormant model engines, initiating high-frequency automated requests within minutes.

    The findings indicate that monitoring routine endpoint activity is no longer sufficient for digital commerce operators running hybrid cloud stacks. IT departments now face the operational reality that unmonitored API limits directly threaten gross margins, turning routine cloud administration into an immediate balance-sheet liability.

    Enterprises running production workflows across multiple public cloud providers must track anomalous model activation alerts and enforce hard spend caps on machine learning billing accounts to catch breaches before usage tallies reach six figures.

  • Huawei Cloud Retires 2 openPangu Models from Model-as-a-Service Platform

    Huawei Cloud Retires 2 openPangu Models from Model-as-a-Service Platform

    Huawei Cloud will shut down commercial access to its openPangu-2.0-Flash and openPangu-2.0-Pro AI models on its Model-as-a-Service platform, ending sales on November 6 and terminating all hosted calls on November 30.

    The cut comes four months after Huawei open-sourced the 505-billion-parameter Pro version, forcing enterprise customers running consumer analytics and internal search workloads in its Southwest Guiyang cloud region to switch tools before December.

    New customers cannot buy or activate either model service after midnight on November 6, while existing accounts retain operational access only through the end of the month. Huawei Cloud directs accounts using openPangu-2.0-Flash to DeepSeek-V4.1-Flash-1017 as their replacement, while routing openPangu-2.0-Pro users to Zhipu AI’s GLM-5.3.

    Migration Demands and Technical Friction

    Engineering teams face immediate rewrite costs. Huawei Cloud noted that the replacement models carry different API parameters, cost structures and output profiles, requiring developers to retest task pipelines rather than swap endpoints blindly.

    Software setups built on the company’s early MaaS standard API V1 must be rewritten to work with an OpenAI-compatible endpoint. Teams already on API V2 or an OpenAI-style interface need to alter model identifier calls.

    Retail commerce platforms relying on openPangu for live inventory categorisation or chatbot assistance now have less than a month to validate replacement models. Changing baseline models often alters latency profiles and customer response formats across live digital storefronts.

    Rival Platforms and Vendor Strategy

    The decision to retire homegrown models while recommending independent rivals highlights the fierce pace of platform rationalisation across China’s cloud ecosystem. Rather than carrying the operational cost of maintaining earlier foundation models, cloud providers now drop their own releases rapidly in favour of top-performing external open weights.

    DeepSeek and Zhipu AI gain immediate enterprise volume from the shift. Huawei Cloud’s public embrace of DeepSeek-V4.1-Flash and GLM-5.3 confirms that third-party open architectures are setting operational benchmarks on Chinese sovereign server hardware, even inside Huawei’s proprietary Ascend computing clusters.

    Platform developers bear the stability risk. By cycling through model versions within single-digit months, platform operators transfer the cost of code maintenance, integration testing and deployment downtime onto enterprise software clients.

    Rapid Model Turnover on Ascend Hardware

    Huawei made openPangu-2.0-Flash public on June 30 with 92 billion parameters, followed by the 505-billion-parameter openPangu-2.0-Pro sparse model on July 31. Both variants were tailored for Ascend neural processing units, accompanied in late September by the release of complete pretraining, supervised fine-tuning and reinforcement learning source code.

    The company’s hosted platform has cleared out earlier systems at a comparable pace throughout the autumn. Huawei Cloud scheduled GLM-5.1 for removal on November 16, switched DeepSeek-V4-Flash accounts to version 4.1, set Kimi-K2.6 for retirement on November 8, and pulled DeepSeek-V4-Pro from hosting in September alongside older Qwen and Wan models.

    The retirement schedule applies strictly to managed API services hosted out of the Guiyang data centre, leaving the underlying open-source weights and training code available in public repositories for teams running their own compute stacks.

    Software teams on Huawei Cloud must finish transitioning their active workloads to the DeepSeek and GLM services before the Guiyang cluster unplugs both openPangu endpoints at midnight on November 30, while Huawei prepares a successor generation of Pangu models without a public release date.

  • Energy North Pulls June 2028 Target for 1 GW Outback Data Centre

    Energy North Pulls June 2028 Target for 1 GW Outback Data Centre

    Australian developer Energy North dropped its planned June 28, 2028 operational target for a 1-gigawatt outback data campus after the Northern Territory environment regulator suspended its statutory review deadline.

    The referral proposed building the off-grid facility across a 191-square-kilometre cattle station located 683 kilometres from the nearest city, combining dedicated solar arrays with gas turbines.

    Work on the schedule stopped after the Northern Territory Environment Protection Authority issued an October 7 ruling pausing its standard seven-week evaluation period. The regulator stated it will seek guidance from federal authorities, who are drafting national policy frameworks for artificial intelligence infrastructure developers.

    Land Size Meets Regulatory Resistance

    Project Ares stands out as the largest Australian computing site forced to withdraw its commissioning schedule. In regulatory filings lodged four months ago, the company presented the June 2028 connection as an indicative benchmark before the local watchdog raised red flags.

    The agency identified 10 distinct environmental threats, citing possible depletion of local water tables and physical disruption to Indigenous sacred sites. In its formal finding, the regulator noted a low level of confidence in the developer’s proposed risk mitigations.

    Plans for the facility call for self-contained infrastructure, including an airfield, dedicated gas turbines, and temporary housing quarters for 4,300 construction personnel. Company management declined to identify the external lenders providing project finance.

    Grid Constraints Drive Developers Into Remote Basins

    Tech infrastructure investors face worsening bottlenecks near primary metropolitan hubs, driving developers toward isolated acreage despite logistical barriers. Goodman abandoned a planned facility in suburban Sydney following community objections, while NextDC postponed a Melbourne expansion due to power grid limitations.

    Hyperscalers and regional platform operators require continuous baseload power that established urban distribution networks can no longer supply without extensive upgrades. Going off-grid into remote pastoral regions solves metropolitan power queuing, but it exchanges utility disputes for complex cultural heritage consents and environmental oversight.

    Regional enterprise tenants in commerce, logistics, and retail banking face higher hosting overheads as pipeline capacity stalls. Operators competing for regional AI workloads cannot easily bypass standard administrative review simply by choosing empty land.

    Pastoral Leases and Indigenous Approvals

    Northern Territory authorities have pitched the jurisdiction’s 260,000 residents and desert climate to capture part of the $7 trillion in global AI data centre capital forecast by McKinsey by 2030. Commercial gas supplies from Tamboran Resources’ Shenandoah South project in the Beetaloo Basin began flowing to Darwin on September 1, 2026, offering developers a local source of fuel.

    Scale remains a major technical hurdle. A separate proposal from Beetaloo Energy near Darwin aims for 2 gigawatts, a footprint requiring up to 400 terajoules of gas daily, or ten times current production from the basin’s only active project.

    Securing project consent across the Territory requires prolonged community rounds with Indigenous traditional owners, who make up a quarter of the local population. Representatives from the Northern Land Council warned federal inquiries that accelerating consultations fails to yield lawful consent, leaving final project approvals exposed to court challenges.

    The Northern Territory EPA has set no revised date to establish terms for the project’s formal environmental impact statement.

  • Alibaba Connects Qoder Desktops Without SSH for 8 Million Users

    Alibaba Connects Qoder Desktops Without SSH for 8 Million Users

    Alibaba added a desktop interconnect feature to its Qoder developer suite on October 30, linking separate computers without requiring SSH or remote desktop configurations.

    The tool lets engineers launch jobs, inspect outputs and monitor workflows on a secondary machine while working entirely inside their local application interface.

    Developers handling code refactoring or validation often distribute workloads across dedicated desktop rigs and portable laptops. The new functionality bypasses terminal networking setups by pairing two instances directly through a temporary pairing code and device address. Users inspect files generated by the remote assistant, including spreadsheets and presentations, directly on their primary display.

    How the Network Link Functions

    Both devices must sign into the identical user account to validate the link. The software restricts access through time-limited codes, allowing engineers to pick specific project directories and assign distinct synchronization rules before starting work.

    Operators can sever active authorizations immediately from the client window. The update builds on earlier mobile features that allowed users to monitor desktop tasks from smartphones.

    Alibaba counts more than 8 million users and over 100,000 corporate clients across the Qoder ecosystem. The lineup includes a desktop workspace, a command-line utility, cloud agents, a JetBrains plugin and the QoderWake digital employee platform.

    Enterprise Workstations and Cloud Margins

    Software development efficiency has become a primary testing ground for Chinese cloud operators trying to sell generative artificial intelligence subscriptions to corporate departments. Tech teams want tools that bypass manual terminal friction without creating security holes in enterprise networks.

    By managing device handshakes inside proprietary client software, Alibaba keeps corporate development workflows bound to its ecosystem rather than external terminal utilities. The commercial risk sits in security governance, where corporate IT administrators often restrict non-standard peer-to-peer device handshakes across internal enterprise subnets.

    Building Out Domestic Hardware Compatibility

    The interconnect rollout follows a series of compatibility releases aimed at expanding Qoder across alternative hardware architectures. Alibaba recently ported the coding software to domestic systems, introducing support for Loongson LoongArch processors alongside China-developed desktop operating systems Kylin and UOS.

    The engineering team also rolled out a fast mode for the client designed to generate first responses within three seconds while cutting developer credit consumption by 20 per cent.

    Alibaba is testing deeper integration between local desktop execution and its remote cloud agent infrastructure, with enterprise adoption figures across Chinese software vendors serving as the performance metric to watch.

  • Tencent Unifies Workplace Apps Under WorkBuddy Agent with 700 Yuan Monthly Tier

    Tencent Unifies Workplace Apps Under WorkBuddy Agent with 700 Yuan Monthly Tier

    Tencent rolled out Buddy One on October 29, linking Tencent Docs, Tencent Meeting, and its knowledge software under a single WorkBuddy artificial intelligence agent with unified accounts and shared credits.

    A premium consumer tier costs 700 yuan per month, bundling WorkBuddy’s top tier with Tencent Meeting Pro, Tencent Docs premium access, and one terabyte of cloud storage.

    Shared Credits and Cross-App Search

    Users link accounts through version 5.9.0 or later of WorkBuddy’s desktop client. The integration eliminates the need to manually toggle tools or paste context between applications.

    When a user submits a prompt, WorkBuddy runs an automated search across approved files, meeting recordings, and knowledge bases. Returned answers label the originating software for each piece of retrieved context. Credits purchased inside WorkBuddy convert directly into universal Buddy One credits functional across Docs, knowledge tool ima, and the company’s new storage system.

    Cloud Storage Built for Agents

    Alongside the consumer bundling, Tencent introduced Tencent Drive, a cloud storage infrastructure layer designed specifically for autonomous software agents. The product provides the shared data layer for WorkBuddy and connects with external agents using a command-line interface and the Model Context Protocol.

    This protocol lets workers complete an initial draft in one third-party artificial intelligence tool and reopen it inside another without manually uploading files. WorkBuddy’s internal knowledge base already operates on this infrastructure.

    For corporate clients, the storage layer relies on Tencent Cloud Object Storage architecture. The setup includes virtual private cloud deployment, off-site disaster recovery backups, and end-to-end data encryption. Corporate administrators can assign distinct permission rules to human employees and automated software tools. Tencent expanded corporate seat storage within WorkBuddy from 5 gigabytes to 50 gigabytes.

    Platform Interoperability and Market Pressure

    Enterprise software platforms across China are racing to prevent corporate clients from switching between disconnected point solutions.

    Tencent is positioning WorkBuddy as an open framework rather than an enclosed ecosystem. Enterprise clients can connect proprietary models, external autonomous agents, and outside messaging networks directly into the storage architecture. Tencent cloud chief Dowson Tong stated that enterprise buyers must retain final discretion over their chosen models, agents, and workplace communication tools.

    Building Out the Hunyuan Architecture

    The product launch follows an active expansion of Tencent’s proprietary model stack. In August, the company introduced its Hy4 preview under the Hunyuan model family, aiming to compete with open-source systems deployed by domestic rivals.

    Corporate deployments have expanded across consumer sectors during the same period. Domestic cosmetics manufacturer Shanghai Jahwa deployed WorkBuddy across nine separate operating departments in late August, testing agent workflows before Tencent finalized this cross-application unified credit tier.

    Enterprise software managers will track how quickly Tencent expands memory sharing between individual applications, which remains scheduled for upcoming client revisions.

  • DeepSeek Releases Model V4.1 Flash Update and Opens Serving Framework

    DeepSeek Releases Model V4.1 Flash Update and Opens Serving Framework

    DeepSeek announced an updated V4.1 Flash artificial intelligence model alongside an open-source inference framework, pricing enterprise tasks down to roughly $0.20 each.

    The developer posted a 77 per cent Pass@1 score on the DeepSWE v1.1 coding benchmark, cutting compute demands through shorter operational outputs.

    Its release of V4.1-Flash-1017 on October 17 raised the system’s Terminal-Bench 4.0 score from 31.2 to 42.3, while DeepSWE v1.1 gained almost three points to 77.0. Benchmark evaluations show the system completing tasks for $0.11 to $0.14 when running lower-cost operational settings, scoring between 72 and 73 per cent. Independent tracking on Arena’s Agent Arena leaderboard placed the updated model first among open-weight competitors and 11th overall across 7,498 real-use sessions, with median execution costs between $0.18 and $0.19 per task.

    Open Source Engine for Enterprise Chips

    Three days after the model release, DeepSeek published DeepInfer, the inference engine running its production services, alongside a request-handling library named deepseek-recipe. The framework manages hardware allocation, latency schedules and computing kernels to serve queries at scale. It runs directly on Nvidia SM100 graphics processing units and Huawei Ascend 950 neural processing units, targeting V4.1 Flash deployments.

    The open-source code does not provide an end-to-end service stack out of the box. Production teams still need to supply their own key-value cache infrastructure, load balancing across parallel replicas, and complete disaggregation systems. Developers gain internal scheduling kernels, but enterprise engineering teams will absorb the integration overhead.

    Shifting the Cost of Business Automation

    For online marketplaces, customer support platforms and software houses across Asia, the release attacks the cost ceiling that keeps autonomous agents out of routine commercial workflows. Running thousands of real-time inventory adjustments, code fixes and customer transactions through commercial western models remains too expensive for thin-margin Asian e-commerce. If an open-weight system matches proprietary grade accuracy at 20 cents a task, mid-sized digital retailers will self-host rather than pipe data through expensive closed APIs.

    Hardware exclusivity remains the primary roadblock for regional operators seeking to replicate these results. While support for Huawei’s Ascend 950 gives Chinese domestic firms an alternative to export-controlled American silicon, Southeast Asian developers operating standard cloud setups will face hardware configuration bottlenecks. Running pruned agent frameworks demands deep systems engineering that many consumer brand operators do not maintain in-house.

    Building Out the Infrastructure Layer

    DeepSeek spent the previous quarter trimming resource requirements across its stack. The company cut key-value cache memory to a quarter and solid-state drive storage to an eighth of previous requirements during the September rollout of V4.1 Flash. It introduced off-peak API access at a 50 per cent discount to balance round-the-clock server loads.

    Commercial access expanded in October with Search and Fetch APIs billed at 5 yuan per 1,000 successful queries. The firm is maintaining free API connections for developer platforms including Claude Code and Codex until per-request billing starts on November 1.

  • Tencent Deploys Retail AI Suite Across 600 Asian Merchants

    Tencent Deploys Retail AI Suite Across 600 Asian Merchants

    Tencent Smart Retail rolled out its artificial intelligence enterprise software suite to 600 retail clients across Asia during an industry summit in Hong Kong. The deployment targets merchants across 30 retail sectors, connecting automated operational software directly into Tencent Cloud and the broader Weixin, WeCom and Mini Programs network.

    Terry Li, vertical general manager of Tencent Smart Retail, detailed the push alongside enterprise partners at the 2026 World Retail Lab Asia event. The software line-up introduces five specialized tools for merchant operations: WorkBuddy for workforce productivity, CodeBuddy for software development, DataBuddy for enterprise data workspaces, Miora for automated creative design, and Ardot for turning design assets into code.

    Connecting Weixin to the shop floor

    The system links merchant back offices directly to front-line sales teams on WeCom and shoppers on Weixin. Rather than forcing store associates to toggle between stand-alone software suites, the setup feeds automated inventory updates, design files and customer prompts into Tencent’s messaging channels. That reduces friction for floor staff who already run client relationships through corporate chat accounts.

    Enterprise software vendors across Asia frequently stumble on floor-level adoption. Tencent is betting that tying generative design and analytics agents to the chat apps clerks already use daily will bypass the standard training hurdles.

    Jewellery chains and supply taggers

    Chow Tai Fook Jewellery Group is using the infrastructure across customer management, store operations and jewellery design. The jewellery retail chain manages a database of nearly 60 million global loyalty members, who generate more than 95 per cent of its total sales volume. The group links unified customer insights to floor operations across its store network to track individual buying habits and stock preferences.

    Industrial supplier SML Group has embedded the software into its product development, market intelligence and supply management. The firm provides RFID tracking, retail identification and digital product solutions to roughly 600 consumer brands across global markets.

    Sinong Duan, vice president of strategy and transformation at SML Group, said: “Customer engagement is one of the areas where we see significant potential for AI. It can help businesses better understand customer needs and improve product recommendations, creating new opportunities to enhance both conversion and service experiences. We have also had the opportunity to work with Tencent WorkBuddy, and its flexibility, user experience and support capabilities have enabled our teams to explore AI adoption more effectively.”

    The platform battle against Alibaba

    The aggressive push into merchant operations pits Tencent directly against Alibaba Group’s cloud and DingTalk ecosystem across Greater China. While Alibaba dominates pure-play e-commerce marketplaces like Taobao and Tmall, Tencent controls private-domain traffic where brands cultivate direct shopper relationships through bespoke Mini Programs. Brand operators increasingly favor private-domain channels because customer acquisitions carry lower marketplace take-rates.

    For consumer brands, committing to proprietary cloud agents locks enterprise data deep inside Tencent’s walled garden. Shifting between competing cloud environments becomes punitive once product catalogs, customer service histories and design archives run on platform-specific models. Retailers gain rapid workflow speed today at the price of long-term software dependency tomorrow.

    Measuring actual returns on automation

    Chinese consumer spending has remained selective throughout the year, forcing retail executives to audit back-office overhead and technology spending closely. Multinationals and domestic retail giants, including Decathlon China, Estee Lauder Group and toy maker Pop Mart, have increasingly trimmed speculative software experiments in favor of tools that cut headcount costs or lift repeat store visits.

    Tencent began assembling its smart retail unit in 2018 to bridge social messaging with physical store sales counters. The group spent earlier iterations refining digital payment terminals and loyalty program Mini Programs for department stores and supermarket chains across tier-one Chinese cities. This latest enterprise suite represents the next operational tier: replacing manual design, coding and inventory forecasting with automated agents.

    Merchant adoption rates for tools such as Miora and Ardot over the upcoming Golden Week and Lunar New Year retail peaks will provide the first real test of whether generative design can shave measurable weeks off corporate fashion and merchandising cycles.

  • Google Defends US$15 Billion Indian AI Data Hub Against Community Scrutiny

    Google Defends US$15 Billion Indian AI Data Hub Against Community Scrutiny

    Google is pushing forward with a US$15 billion artificial intelligence data centre hub in Visakhapatnam despite community protests over power grids and water supplies in southern India.

    The gigawatt-scale development represents the search giant’s largest AI infrastructure project outside the United States and is scheduled to enter service in 2028.

    Local groups and environmental activists in Andhra Pradesh have called for a blanket moratorium on data centre construction across the state until authorities assess collective ecological impacts. The development footprint spans construction sites at Tarluvada village and Rambilli in the Anakapalli district, placing heavy computing loads in an area already experiencing water stress.

    Air Cooling Replaces Freshwater Systems

    Google altered its baseline engineering for the campus after pre-development studies identified acute pressure on the surrounding watershed. The facility will rely on advanced air cooling instead of evaporative water cooling towers to chill server racks, avoiding the millions of litres of daily freshwater extraction typical of large-scale server farms.

    Facility maintenance and non-cooling operations will still draw municipal water. The company has pledged to replenish 120 per cent of the volume consumed by non-cooling facility needs across its operational footprint by 2030.

    Grid capacity presents a parallel bottleneck. While Google declines to publish site-level electricity numbers, company executives state that the group pays for all necessary grid substation and transmission upgrades upfront while funding local renewable energy projects above statutory quotas.

    Subsea Cabling and Regional Edge Networks

    Visakhapatnam offers direct coastal access for international subsea fibre connections linking India to landing stations in Australia and South Africa. That routing establishes Andhra Pradesh as India’s third core subsea landing hub alongside Mumbai and Chennai, giving enterprise networks faster bandwidth to Southeast Asia and Indian Ocean routes.

    Hyperscale cloud operators across the Asia-Pacific region face rising friction over energy allocations. Authorities in mature data hub markets like Singapore have enforced strict sustainability caps, while Finnish regulators recently suspended construction on two Google sites pending mandatory environmental assessments. Digital platforms rolling out enterprise AI models across Asia must now compete directly with residential populations and manufacturing zones for grid access.

    For enterprise software vendors, regional retailers and cloud customers, infrastructure constraints in primary metros make secondary coastal hubs essential. Yet when global operators cluster in emerging tier-two corridors, they import the same resource disputes that delayed builds in Western Europe and Southeast Asia. The financial exposure rests with regional governments that offer tax concessions before local utilities can guarantee steady baseload power.

    Economic Payoffs and Local Employment

    Data centre developments generate large short-term construction payrolls but run with lean permanent staff once servers go live. Google estimates the Visakhapatnam hub will support approximately 4,200 construction workers annually across a three-to-five-year build cycle.

    Company projections place the recurring economic contribution at US$688 million each year through supply chain activity, facility services and maintenance contracts. Alexander Smith, Google’s head of infrastructure and energy for the Asia-Pacific region, defended the long-term community footprint during project reviews.

    “For every job inside a Google data centre we support nine additional jobs in the surrounding community,” Smith said.

    Andhra Pradesh state planning officials must now review activist petitions demanding an environmental audit before granting additional clearances for companion tech facilities scheduled for the same coastal corridor.

  • Telcos Struggle to Scale AI as Only 13 per Cent Reach Production

    Telcos Struggle to Scale AI as Only 13 per Cent Reach Production

    Communications service providers have deployed artificial intelligence across dozens of operational trials, but only 13 per cent of operators have successfully scaled those deployments into full production environments.

    The deployment bottleneck persists even though nearly 75 per cent of companies report positive results during initial pilot phases, exposing structural friction inside legacy telecom IT stacks.

    Industry consortium TM Forum and regional operators will address the operational divide on November 5 during a joint technical briefing on autonomous network infrastructure and agentic customer experience. The session brings together Joann O’Brien, vice president of composable IT and ecosystems at TM Forum, alongside Seetha Gurunathan, global head of strategy and transformation at Vodafone, Shekhar Kulkarni, chief IT architect at Deutsche Telekom, and Seshan Krishnamurthi, vice president for Africa and Central Europe at Covalense Digital.

    The Bottleneck in Monolithic Billing Systems

    Legacy business support systems remain the central roadblock for carrier monetization. Traditional billing architectures were engineered for static monthly invoicing rather than real-time micro-transactions, non-terrestrial satellite network access, or dynamic algorithmic pricing.

    When carriers attempt to bolt generative or agentic intelligence onto these rigid systems, integration costs surge and transaction speeds falter. The industry push now centers on replacing monolithic enterprise software with modular, cloud-native components communicating through standardized application programming interfaces.

    Moving from Reactive Chatbots to Agentic Systems

    Customer service operations represent the initial testing ground for carrier autonomy. Early deployments relied on basic script-based chatbots that answered recurring billing queries, but network operators are now testing autonomous software agents capable of executing account modifications and diagnosing network faults without human intervention.

    Giving software agents execution authority introduces operational liabilities. Carriers must balance automated problem resolution against transactional accuracy, data compliance, and consumer trust across diverse regulatory jurisdictions.

    Standardizing Open Architecture Across Vendors

    Multi-vendor integration requires common software standards across global supply chains. TM Forum is pushing its Open Digital Architecture framework to establish interoperable, plug-and-play IT components that prevent telecom operators from locking themselves into single-vendor software suites.

    Standardized interfaces allow operators to deploy machine-learning models from independent software vendors directly into existing network workflows. Without open architecture standards, every software upgrade requires bespoke systems integration that slows time-to-market.

    Measuring Returns on AI Capital Expenditures

    Telecom operators face mounting investor pressure to justify heavy capital allocations to digital transformation projects. Evaluating returns requires a metrics framework that tracks real operational expenditure cuts, customer churn rates, and service provisioning speeds alongside headline revenue gains.

    Network operators will test whether these modular architectures can deliver measurable financial returns when the industry reviews production deployments in the upcoming operational cycle.

  • Microsoft Secures Approval for $902M Western Sydney Data Centre

    Microsoft Secures Approval for $902M Western Sydney Data Centre

    Microsoft won New South Wales government approval for an A$1.3 billion ($901.9 million) data centre in western Sydney on 1 October. The five-storey facility in Huntingwood will deliver 96 megawatts of IT capacity across 40 data halls.

    State planners signed off on the project three days after Goodman Group abandoned plans for an A$798 million facility on Sydney’s north shore following neighborhood protests. Industrial corridors and suburban fringes continue to drift apart as cloud infrastructure expands across Australia.

    Power Caps and Grid Conditions

    The development consent caps total power consumption at 140 megawatts. Before operations begin, state authorities require Microsoft to match all electricity demand with additional, firmed renewable supply. This power must come through power purchase agreements for generation within New South Wales.

    First-year carbon dioxide equivalent emissions are projected at 1,326 kilotonnes, driven largely by operational electricity draw. Onsite backup systems include 46 diesel generators holding up to 1.5 million litres of fuel, alongside 158,763 kilograms of lithium-ion batteries. Cooling will rely on 36 rooftop evaporative cooling towers drawing approximately 3.9 megalitres of water daily.

    Federal Biodiversity Clearance

    Planning consent gives Microsoft no immediate path to break ground. The company must still secure federal environmental sign-off under the Environment Protection and Biodiversity Conservation Act.

    Clearing the site requires removing 0.33 hectares of a conservation area containing critically endangered woodland and potential wildlife habitats. Construction will take 24 months once federal approvals land. It will generate 250 construction jobs and 74 permanent operational positions.

    Site Selection Shifts West

    Securing the Huntingwood site required extensive negotiations with Blacktown City Council over building bulk, drainage, and landscaping before reaching a formal planning agreement in September 2025. The proposal attracted only a single public objection. Residential properties sit 90 metres away.

    Goodman faced a harsher reception at its proposed Project Mars site in Lane Cove West. Homes stood 20 metres away, with a primary school within 160 metres. That proposal drew 374 public submissions, with only nine in support, before Goodman withdrew it on 28 September.

    The policy environment for large data centres had evolved significantly since planning started, Goodman head of planning Guy Smith told the planning department.

    Hyperscalers and real estate developers in Sydney are concentrating capital in western industrial clusters to avoid urban friction. Huntingwood borders Eastern Creek, where Goodman has filed plans for a 500-megawatt, A$5 billion campus. Nearby in Minchinbury, Starwood, Doma, and Telstra InfraCo are building a 62-megawatt facility.

    National Standards on the Horizon

    Engineering consultancy Lehr Consultants International lodged Microsoft’s initial application in 2023 without naming the end user. The project ties into Microsoft’s October 2023 commitment to invest A$5 billion over two years to expand Australian cloud and computing infrastructure.

    State planners also approved Stockland’s A$630.4 million Road 1 Data Centre in Macquarie Park on the same day, permitting a 34.3-megawatt IT load alongside future residential towers. Developers now face tighter rules. Victoria is proposing mandatory 150-metre buffer zones from homes, while the Australian federal government prepares national data centre legislation for introduction in early 2027.

  • Goodman Faces Community Resistance over 3 Trillion Yen Tsukuba Data Center

    Goodman Faces Community Resistance over 3 Trillion Yen Tsukuba Data Center

    Goodman Group faces local opposition at its 3 trillion yen ($20 billion) data center campus in Tsukuba, Japan. The pushback follows the cancellation of a similar project in Sydney.

    The 45-hectare site in Ibaraki Prefecture will reach an eventual capacity of 1 gigawatt. It ranks among the largest foreign-backed digital infrastructure projects in the country.

    Work began earlier this year on the initial 50-megawatt phase in the Oho district, an area of about 20,000 residents roughly 50 kilometers northeast of Tokyo. The site sits next to a residential neighborhood, a hospital and a nursing home. Heavy construction machinery arrived on location in February. Residents say that gave them their first real indication of the project’s scale.

    Dispute over site purpose and consultation

    Neighborhood representatives say initial planning filings obscured the data facility. Municipal documents in 2023 originally described the project as a logistics warehouse. References to a data center appeared later in the planning process.

    Hiroyuki Yanagimachi, who formed a residents group to challenge the project, gathered roughly 800 signatures demanding a public town hall. His group sent two formal letters to Goodman requesting an open session. Yanagimachi said the group will protest at the Tsukuba site and outside the developer’s Tokyo office if open talks do not happen.

    Goodman said in a statement that its door remains open to meet the neighborhood association head. It plans to hold an information meeting after consulting with local authorities. The developer previously hosted two invitation-only evening dinners in August for several dozen attendees at a local restaurant. Participants said the format felt like promotional hospitality rather than an open public forum.

    Grid demand and foreign capital in Japan

    Foreign property groups are pouring record capital into Japanese digital infrastructure to capture demand for artificial intelligence workloads and cloud capacity. Singapore-based GLP and London-based Colt Data Centre Services are building large campuses across Greater Tokyo and regional hubs. Both frequently buy industrial plots close to existing residential districts.

    Power limits and land scarcity force operators to build closer to population centers where local zoning rules offer few formal hurdles. The friction in Tsukuba exposes an operating vulnerability for international developers. Many rely on standard logistics zoning permits to secure land before converting designs to power-heavy computing hubs.

    The Japan Data Center Council, an industry association that includes Goodman, issued guidance in May calling for genuine consultation. Its guidelines state that operators must treat public engagement as a platform for discussion rather than a one-way notification process.

    Sydney cancellation and project pipeline

    In Australia, the developer recently withdrew plans for a major Sydney data center following sustained resistance from local neighborhood groups. That setback raises the operational stakes for its Japanese pipeline, where land acquisition costs and grid allocations are tied to long-term investment timelines.

    Company officials say Goodman has complied with all legal standards and surpassed baseline regulatory guidelines for the Tsukuba site. The project remains on schedule for its initial phase while the developer coordinates with municipal officials on the format for future public meetings.

  • Global AI Data Centre Spending to Hit Thirty Trillion Dollars by 2050

    Global AI Data Centre Spending to Hit Thirty Trillion Dollars by 2050

    Cumulative global spending on artificial intelligence data centres will reach 30 trillion dollars by 2050, exceeding the capital poured into any previous technological transition in modern economic history.

    That outlay projection from PwC puts infrastructure commitments on par with the entire market value of outstanding United States Treasuries. The scale of investment eclipses the capital deployments seen during the nineteenth-century railroad expansion and the late-1990s internet boom, even after adjusting for historical inflation.

    The Capital Equation for Infrastructure

    Tech companies and sovereign funds are committing capital to server capacity, specialised chips, and power generation at a pace that demands rapid commercial returns. The sheer volume of hardware spending means enterprise software buyers now face pressure to prove that automation cuts operating costs faster than the infrastructure depreciates.

    Data centre developers are also running into physical bottlenecks across regional power grids. In major markets, server farms are competing directly with residential housing sectors for high-voltage electricians, base-load electricity contracts, and industrial water allocations required for cooling compute clusters.

    Cost Shifts Across Enterprise Operations

    Corporate adoption is altering cost structures across regional businesses, though initial savings frequently push expenses into other parts of the balance sheet. In retail and supply chain operations, automated inventory tracking and customer service systems reduce direct headcount while increasing technical compliance and software licensing overhead.

    The labour impact has hit external service providers first. Enterprise data compiled by fintech platform Ramp shows companies that spent heavily on freelance contractors cut those budgets by 97 per cent after switching task workflows to generative software. The transition eliminated costs for translation, graphic design, and basic coding without requiring formal corporate restructurings or severance payouts.

    Dispute Volume and Operational Frictions

    Lower barriers to generating technical and legal text are creating fresh operational strains for regulators and public tribunals across the Asia-Pacific region. Australian legal services provider Maurice Blackburn noted in regulatory filings that automated tools alter access costs across the market.

    “AI will let law firms deliver high-quality services at scale,”

    Cheap generative drafting has also driven an influx of poorly verified submissions into administrative bodies. Australia’s Fair Work Commission has recorded an increase in automated unfair dismissal filings, forcing public registries to allocate extra staff time to screening low-quality claims.

    Consumer Agents and Capital Allocation

    New consumer tools are targeting deposit inertia at traditional financial institutions. Meta has introduced its Muse personal assistant, designed to help retail users automatically identify higher-yielding non-bank deposit alternatives and shift balances out of standard accounts offering sub-one per cent returns.

    These consumer-facing agent deployments mark the next phase of enterprise software rollouts following two years of pure model training and server farm buildouts. The Australian Parliament’s Joint Select Committee on Artificial Intelligence is scheduled to deliver its final findings on regional economic impacts in November 2026.