LeituraPesquisaRadarFramework de investimento
Entrar / Cadastrar
TudoTecnologiaProdutosInvestimentoBolsaOutros

Leitura

2026-10-29198 posts

Atualizar

US Carriers Oppose Starlink Spectrum Expansion as Satellite Direct-to-Device Services Launch in Brazil

Tópico · 星链直连蜂窝业务Material
2026-10-29 09:31 GMT+8

As rumors of SpaceX's Starlink building its own terrestrial mobile network in the US intensify, traditional telecom operators are ramping up regulatory pushback. According to ZDNet Korea, Verizon, AT&T, and T-Mobile have jointly submitted comments to the US Federal Communications Commission (FCC), demanding a review of plans to expand frequencies for satellite direct-to-device (D2D) communications. They argue that the 2.5GHz and 3.45GHz bands, critical for 5G services, should be excluded from D2D regulations. T-Mobile warned that interference between satellite and terrestrial signals could result in up to 96% loss of base station uplink coverage.

Simultaneously, the cost barriers to market entry have been quantified by incumbents. Dan Schulman, CEO of Verizon, estimated that for Starlink to build an independent nationwide mobile network, it would take at least 4 to 6 years and require an investment of $300 billion to $500 billion over the next 7 to 8 years. He noted that beyond base stations, backhaul networks, and Radio Access Network (RAN) equipment, new entrants must also bear the costs of spectrum acquisition, user management, and billing systems, expressing skepticism about MVNO partnerships with satellite operators.

Despite regulatory friction in the US, commercialization is accelerating globally. In Brazil, the largest mobile operator Vivo launched 'Vivo Satélite', a satellite direct-to-device service in partnership with Starlink on October 26. The service covers approximately 50 million users, offering SMS, maps, and app-based communication in dead zones, with the first 12 months free for individual customers. On the same day, competitor AST SpaceMobile announced it had secured usage approval from Brazilian regulator Anatel for S-band 30MHz and up to 20MHz in the 700MHz band, planning to collaborate with four local operators and build four ground gateways.

For investors, these developments highlight two parallel trends in the satellite communications sector: traditional operators protecting their spectrum assets and infrastructure ROI through lobbying, and emerging markets and equipment vendors benefiting from actual deployment. Bank of America analysis suggests that Starlink's potential ground network construction could positively impact wireless access network equipment makers like Nokia and Ericsson, as overall capital expenditure is unlikely to shrink even if competition intensifies.

Fontes:zdnet.co.kr

Pesquisa

Korea Elderly Living Support Center Partners with Intune Labs to Launch AI Voice Agent Care Pilot

Tópico · Intune LabsVisão rápida
2026-10-29 10:57 GMT+8

According to a report by South Korean media outlet ETNews, the Korea Elderly Living Support Center has signed a Memorandum of Understanding (MOU) and a proof-of-concept contract with South Korean voice artificial intelligence company Intune Labs to officially introduce an AI agent-based "elderly care platform." This initiative aims to address the worsening shortage of caregiving personnel in public welfare settings as South Korea enters a super-aged society.

The core of this partnership lies in utilizing advanced AI agents to respond immediately to field needs. The platform primarily offers three services: first, conversational companionship through sophisticated voice AI to alleviate emotional isolation among elderly individuals living alone; second, automated check-in calls and health status assessments tailored for seniors, asking daily about medication, meals, and sleep conditions; third, a smart management system for caregivers, where the AI analyzes and records daily conversations and health data in real time, significantly reducing the administrative burden on life support workers and social welfare professionals.

Under the signed proof-of-concept contract, the two parties plan to conduct a three-month service pilot targeting 50 elderly individuals managed by the center from November 2026 to January 2027. During the trial period, the focus will be on evaluating the actual effectiveness of voice AI interactions in alleviating loneliness among the elderly, as well as assessing quantitative and qualitative indicators of how AI agents reduce frontline workload and expand the number of people one caregiver can effectively manage.

Fontes:미래를 보는 창 - 전자신문

Pesquisa

ByteDance's Jimeng Launches Short Drama Studio to Streamline AI Video Series Production

Tópico · 即梦短剧工作室Material
2026-10-29 12:02 GMT+8

ByteDance's AI creation platform, Jimeng, has officially launched the "Short Drama Studio," designed to address challenges in transitioning from single AI-generated clips to multi-episode series, such as character consistency, scene coherence, and high trial-and-error costs. The tool integrates proprietary models—DreamScript for writing, Anycook for visual assets, and DreamShot for storyboarding—to create a complete pipeline from inspiration to final cut.

Unlike traditional workflows that rely on repeated prompt-based "gacha" generation, the new studio emphasizes asset accumulation and reuse. After importing or generating a script, the system automatically parses characters, scenes, and props into a manageable digital asset library. These assets can be directly called upon for subsequent shot generation, sequel development, or style transfers, preventing the visual drift common when re-setting parameters each time.

To mitigate the cost of failed generations, the studio introduces a "3D Director Stage" and "Smart Pre-visualization." Creators can adjust camera angles, character positioning, and motion trajectories in a 3D space for low-cost dynamic reference before committing compute resources to formal video generation. Additionally, it supports partial re-shooting, audio separation, and frame-level annotation editing for local flaws like incorrect costumes or unnatural movements, eliminating the need to regenerate entire segments.

According to hands-on testing by iFanr, combined with Double Eleven promotions, the unit price for Seedance 2.5 at 720P resolution drops to approximately 0.19 RMB per second. However, the report notes that true cost reduction lies not just in per-second pricing but in minimizing invalid attempts through asset reuse and pre-visualization. This shift enables smaller teams to sustainably produce serialized content, marking a transition of AI video from one-off experiments toward industrialized production workflows.

Fontes:爱范儿

Pesquisa

Samsung SDS Q3 Revenue Up 11.2%, Cloud Business Grows 18% on Enterprise AI Demand

Tópico · 三星SDS AI转型Material
2026-10-29 09:28 GMT+8

Samsung SDS, a major South Korean IT services and logistics company, released its third-quarter earnings report on October 29. Consolidated revenue reached 3.77 trillion KRW, up 11.2% year-over-year, while operating profit rose 3.7% to 241 billion KRW. The performance was primarily driven by rapid expansion in cloud services and increased air and sea freight volumes.

Within the IT services division, the cloud business served as the core growth engine, generating 796.1 billion KRW in revenue, an 18% increase year-over-year. Samsung SDS attributed this growth largely to expanding demand for Artificial Intelligence Transformation (AX) among enterprises. Specifically, the Cloud Service Provider (CSP) segment grew 15% due to increased demand for Samsung Cloud Platform (SCP) and expanded GPU-as-a-Service (GPUaaS) supply to public and software sectors. The Managed Service Provider (MSP) segment grew 25%, driven by financial sector AX projects, ERP expansions in shipbuilding, and increased adoption of global AI services.

The logistics business also performed strongly, with revenue reaching 2.09 trillion KRW, up 16% year-over-year, marking the first time quarterly revenue exceeded 2 trillion KRW. This growth was supported by higher air and sea cargo volumes and rising freight rates. Additionally, Samsung SDS recently acquired 100% of Ujeong Aviation, an air cargo consolidation specialist, to further expand its global aviation logistics capabilities.

To sustain cloud growth momentum, Samsung SDS is accelerating its 'Multi-Dimensional AI Full Stack' strategy, covering everything from data center infrastructure to practical AI service applications. In AI infrastructure, the company has been selected as the preferred bidder for Pacific Asset Management's large-scale data center Design-Build-Operate (DBO) project exceeding 200MW. Meanwhile, the National AI Computing Center is targeted for completion in 2028, and the initial 60MW AI data center in Gumi, North Gyeongsang Province, is scheduled to break ground in early November.

In the AX services domain, Samsung SDS has secured contracts for AI and data platform construction in the financial sector and was selected as the preferred bidder for AI platform projects at insurance companies and financial holding firms. As demand for AI platforms expands in finance, related business opportunities are increasing. In the public sector, its AI collaboration solution Brity Works has reached 330,000 cumulative users, and the company launched Brity Works Cowork, an AI assistant supporting internal document search and analysis.

Fontes:zdnet.co.kr

Pesquisa

Tencent Launches Buddy One: Unifying Accounts and Credits Across Docs, Meeting, Drive, and ima

Tópico · Buddy OneMaterial
2026-10-29 11:16 GMT+8

On October 29, at the Tencent Global Digital Ecosystem Conference, Tencent officially launched Buddy One, a cross-application AI collaboration system for individual users. The system integrates Tencent Docs, Tencent Meeting, Tencent Drive, and ima, sharing the WorkBuddy Agent base while unifying account and credit pools. Users must update the PC version of WorkBuddy to 5.9.0 or higher to link these services.

Buddy One addresses the issue of repeated authorization previously faced when invoking different products. In the past, WorkBuddy required re-authentication whenever connector permissions expired. Now, users only need to link their accounts for the four supported products once, allowing them to directly search and integrate authorized documents, meeting records, and knowledge base content within WorkBuddy. The system generates answers with cited sources, eliminating the need for manual context copying. Search results clearly indicate which application each piece of information originated from.

Regarding the credit mechanism, joining Buddy One upgrades WorkBuddy credits to Buddy One credits, which are shared across WorkBuddy, Tencent Docs, ima, and Tencent Drive. Credit consumption in any one app synchronously reduces the balance in others, covering package credits, promotional gifts, and purchased credits. Notably, Tencent Meeting does not currently consume credits but remains searchable via WorkBuddy after linking. Existing independent credit rules within each app remain valid and will not be merged or cleared.

Launched concurrently, 'Tencent Drive' is positioned as a new memory space product for the Agent era, providing unified storage for the WorkBuddy suite. Currently, Buddy One offers two main benefits: interconnected credits and interconnected knowledge, with memory sharing yet to go live. Tencent stated that more products will join this system in the future.

Fontes:凤凰网

Pesquisa

Lee Sun-woong, CEO of Cloudike: For Public Institutions Adopting AI, Data Environment Design Is as Critical as Model Selection

Tópico · 클라우다이크AI安全框架Material
2026-10-29 10:00 GMT+8

Many enterprises wish to use AI to process internal documents, yet they often overlook a critical issue as vital as model selection: where exactly is the data stored, where is it processed, and who has access? In an interview with ZDNet Korea, Lee Sun-woong, CEO of the Korean enterprise cloud SaaS company Cloudike, identified this factor as one of the primary bottlenecks preventing public institutions from adopting generative AI.

By 'data location,' he does not merely refer to which server room houses the hardware. In his framework, this question encompasses where documents are processed, who has access rights, and where Q&A records are ultimately stored. He argues that the requirements for using public information differ significantly from those for processing sensitive internal documents. Organizations with high security standards must first distinguish which data can be used in which services before considering model performance comparisons.

For institutions unable to export data to external environments, he proposes an alternative: running AI within the environment where the data resides—deploying models on intranets or dedicated infrastructure controlled by the institution to complete document retrieval and summarization within that setting. However, he notes that simply installing the model inside the intranet is insufficient: every step, including text extraction, search indexing, Q&A log recording, and any connectivity features that might send data to external services, must be individually verified.

Regarding regulatory boundaries, he highlighted a common misconception: CSAP certification applies only to the certified cloud service itself and does not automatically guarantee the security of the entire AI system connected to it. While South Korea's National Network Security Framework (N2SF) classifies security levels based on business information and system importance to impose corresponding controls, even if N2SF applies, it does not mean all internal data can be input into external AI services. Data classification, the institution's own security policies, and necessary control measures still require item-by-item review.

He advises organizations preparing to introduce AI to first determine the specific business scenarios they intend to use, confirm which documents are needed for those scenarios and how existing permission systems operate, and then validate effectiveness on a small scale in a secure environment before gradually expanding. For data difficult to export externally, AI should operate on that data within an environment controllable by the institution.

It should be noted that Cloudike itself is a supplier of CSAP-certified public cloud file storage and collaboration services, while also supporting customers in deploying their own infrastructure. Lee Sun-woong's arguments align closely with his company's business positioning: positioning data management infrastructure as a key layer for AI implementation, equally important to model selection. This commercial background does not diminish the rationality of his technical judgment, but readers adopting his advice should consider data environment design as one of the planning dimensions, evaluated in parallel with model assessment.

Fontes:zdnet.co.kr

Pesquisa

Seoul Proposes 26.6 Billion KRW Budget for Youth Generative AI Access and Education

Tópico · 首尔青年AI券计划Visão rápida
2026-10-29 09:54 GMT+8

The Seoul Metropolitan Government announced on October 29, 2026, that it will allocate 26.6 billion KRW (approx. $18.5 million USD) in its 2027 budget proposal for the "Youth AI Growth Pass" project. The initiative aims to provide 500,000 young residents of Seoul with vouchers for generative artificial intelligence services and customized, level-based education.

According to Yonhap News Agency, Seoul Mayor Oh Se-hoon stated that the measure is designed to prevent disparities in employment and income caused by unequal access to AI tools and training costs. This marks one of the first major policies introduced since the start of his ninth-term administration.

Notably, the pilot budget for this project was fully cut during the supplementary budget review by the Seoul City Council in September. The city claims that surveys targeting youth and experts confirmed the necessity of the program, leading to its inclusion in the new fiscal year's main budget. The proposal now awaits deliberation by the City Council, where the opposition Democratic Party holds a majority, leaving its final approval uncertain.

Fontes:미래를 보는 창 - 전자신문

Pesquisa

Marvell Reported to Be in Advanced Acquisition Talks with Lumentum

Material
2026-10-29 13:35 GMT+8

According to Bloomberg, U.S. networking chipmaker Marvell Technology is in advanced acquisition talks with Lumentum Holdings, a supplier of optical components for data centers. Reuters, citing sources, reported that the deal could be announced as early as next week. Neither Marvell nor Lumentum immediately responded to requests for comment, and the price and transaction structure remain undisclosed.

Market reactions were mixed: Lumentum shares closed up 5.52% in regular trading and rose approximately 9% in after-hours trading, while Marvell fell more than 8% after hours. Analysts suggest the potential merger aims to combine Marvell's optical signal processing chips with Lumentum's lasers and photodetectors, addressing the shift from copper to optical interconnects within AI computing clusters. Bernstein analysts noted that given Lumentum's estimated $83 billion market cap, the deal size makes an all-cash payment nearly impossible and highlights significant integration challenges.

Fontes:INSIDE

Pesquisa

South Korean Startup OpenLabs Launches On-Premise AI Development Service Nuvida

Tópico · NuvidaVisão rápida
2026-10-29 11:58 GMT+8

South Korean AI agent startup OpenLabs launched the on-premise version of its AI development automation service, 'Nuvida', on October 29, 2026. The service allows enterprises to utilize multiple AI agents for collaborative software development within internal networks disconnected from the internet.

Nuvida employs a 'specification-based development' model. Unlike traditional methods where developers query AI chatbots for code snippets piecemeal, users first define standardized functional specifications. A project management agent then coordinates specialized agents for database, infrastructure, frontend, backend, and quality assurance to generate design documents, database schemas, and code. Developers review and approve results at each stage.

OpenLabs stated that this approach addresses issues arising from fragmented use of AI coding tools, such as inconsistent code styles and discrepancies between design documents and actual code. By synchronizing designs and code based on central specifications, the company aims to reduce rework.

The newly released on-premise model processes all tasks within the corporate intranet without communicating with external servers. This mitigates security concerns for regulated industries like finance and large enterprises, preventing the leakage of core business logic or transaction code. Additionally, OpenLabs introduced 'Nuvida Code', a helper tool ensuring individual developers adhere to central specifications, and supports batch conversion of legacy C-language systems to Java.

Fontes:스타트업 생태계의 모든 것 \'플래텀(Platum)\'

Pesquisa

Chinese AI Hardware Suppliers Biwin and Moore Threads See Stocks Plunge Despite Q3 Revenue Surge

Tópico · 佰维存储与摩尔线程Material
2026-10-29 13:35 GMT+8

Shares of Chinese semiconductor suppliers Biwin Storage Technology and GPU maker Moore Threads tumbled more than 10% on Thursday following the release of robust third-quarter earnings, reversing Wednesday's rallies. The market retreat highlights investor concerns over near-term margin pressures and cash burn outweighing long-term AI promises.

Beijing-based Moore Threads reported a nearly 1,774% surge in third-quarter revenue to 1.55 billion yuan (US$232.4 million), with net profit reaching 49.3 million yuan, or 20.1 million yuan excluding non-recurring items. The company attributed this rapid expansion to robust domestic demand for GPUs used in AI model training and inference across cloud providers, telecommunications operators, and computing centers. However, profitability remains fragile due to steep product development overheads; adjusted net loss for the first nine months was 130.7 million yuan, as R&D expenditure rose approximately 55% to 1.34 billion yuan, consuming nearly 41% of total revenue.

Meanwhile, Shenzhen-based Biwin posted a net profit of 1.6 billion yuan for the quarter ended September 30, up 525% year-on-year, while revenue more than doubled to 6.14 billion yuan, according to an exchange filing. The storage module maker cited resilient AI computing demand and a strong memory market cycle. Sales of storage products for emerging AI devices reached 4.82 billion yuan in the first nine months, up nearly 390%, serving tech giants including Google and Meta Platforms.

Despite the annual growth, sequential slowdowns raised caution. Calculations based on disclosures show Biwin's net profit slumped 62.5% from the second quarter, and revenue dropped nearly 30%. Both companies faced strained cash reserves due to capacity expansion: Biwin recorded a nine-month operating cash outflow of 10.04 billion yuan, while Moore Threads posted an outflow of 3.31 billion yuan, attributing the burn to rising procurement costs and strategic stockpiling.

The broader market context exacerbated the sell-off, with the Shanghai STAR 50 index dropping more than 4%, deepening a slide that had already seen the bourse lose more than 40% of its value from a July peak. Nomura analysts projected in an October 19 report that memory supply shortages could persist through at least 2028, arguing that the market underestimates the breadth of AI demand, though they warned that sustained cuts to computing commitments or significant delays to funded projects would challenge this outlook.

Fontes:South China Morning Post

Pesquisa

Microsoft Discloses $10 Billion Quarterly Revenue from OpenAI, Nearly 11% of Total

Tópico · 微软OpenAI收入分成Material
2026-10-29 05:34 GMT+8

U.S. technology giant Microsoft disclosed in its regulatory filing for the first quarter of fiscal year 2027, ended September 30, 2026, that it generated $10 billion in revenue from commercial arrangements with OpenAI, including revenue-sharing payments. This figure represents nearly 11% of Microsoft’s total quarterly revenue of $92.3 billion, underscoring the enormous financial ties that remain between the two companies even as they increasingly compete for the same customers.

According to the filing cited by TechCrunch, Microsoft’s revenue from OpenAI hit $24.1 billion for the full 2026 fiscal year. At the latest quarterly pace, Microsoft would pull in approximately $40 billion annually from OpenAI-related arrangements. Additionally, Microsoft reported $7.2 billion in outstanding receivables from OpenAI at the end of September, up from $6 billion at the end of June. This money represents amounts OpenAI owes for previous transactions, not additional revenue.

It remains unclear how much of the $10 billion came from revenue-sharing payments versus Azure cloud computing services. In November 2025, reports citing leaked documents indicated Microsoft had received $865.8 million in net revenue-sharing payments from OpenAI during the first nine months of that year. OpenAI reportedly shares 20% of its revenue with Microsoft as part of a previous deal where Microsoft invested more than $13 billion, though neither company has publicly confirmed this percentage. Under an April 2026 agreement, OpenAI is obligated to continue sharing revenue with Microsoft through 2030, subject to a cap, even as it gains freedom to work with competing cloud providers.

Microsoft’s ownership stake in OpenAI also declined slightly from about 25% in June to 24% in September, continuing a gradual drop from roughly 27% following the October 2025 recapitalization. The company did not specify what caused the dilution. This disclosure provides investors with critical data points for assessing Microsoft’s revenue stability and its exposure to a single major partner.

Fontes:TechCrunch

Pesquisa

South Korean Survey: 81% of Job Seekers and 92% of HR Staff Support Banning AI Smart Glasses in Interviews

Tópico · 韩国AI智能眼镜面试禁令Visão rápida
2026-10-29 09:52 GMT+8

According to a report by Korean tech media outlet Platum, HR technology company SuperCoder and PR agency Dongseo Nambug jointly released the results of a survey on "wearing AI smart glasses during interviews" on October 29. The online survey was conducted from October 12 to 26, targeting 478 participants of an AI interview competition hosted by SuperCoder and 348 HR personnel from SuperCoder's partner companies. It should be noted that the sample was not randomly selected but focused on specific groups.

The results showed that 81% of job seekers and 92% of HR staff agreed to ban the wearing of AI smart glasses in face-to-face interviews. However, the reasons for supporting the ban differed significantly between the two groups. 51% of job seekers cited "fairness and equality" as the primary reason, arguing that not everyone can afford expensive smart glasses; followed by corporate and personal privacy leaks (32%) and difficulty in assessing true capabilities (16%). In contrast, 67% of HR staff were most concerned about corporate and personal information leaks, pointing out that the camera and recording functions of the glasses could lead to the exposure of interviewer identities or interview questions; only 26% of HR staff listed fairness issues as their top concern.

Yoo Seung-hoon, CEO of Dongseo Nambug, stated that cases of cheating using AI glasses have recently been detected in South Korean college entrance exams and national qualification tests, suggesting that new types of cheating enabled by technology could shake trust in the recruitment market. Choi Jae-woong, CEO of SuperCoder, said that some large South Korean companies have already begun prohibiting the entry of AI smart glasses and other assistive devices into interview venues, and are reviewing procedures to identify and handle such devices.

Fontes:스타트업 생태계의 모든 것 \'플래텀(Platum)\'

Pesquisa
Próxima →
Entrar / Cadastrar
LeituraPesquisaRadarFramework de investimento