China’s computing‑power industry is rapidly maturing as national policies, massive infrastructure investments, and domestic chip development converge to create a strategic high‑ground that fuels AI, data centers, and digital‑economy transformation, with clear upstream, mid‑stream, and downstream value chains.
The U.S. AI Action Plan outlines a three‑pillar strategy—accelerating AI innovation, building robust AI infrastructure, and asserting leadership in international AI diplomacy and security—to secure America’s technological dominance, protect national interests, and ensure AI benefits American workers and society.
The AI Adoption Survey H1 2025 reveals that nearly half of organizations have deployed AI in production, engineering and R&D lead usage, Chinese LLMs gain overseas interest, and cost, reliability and intelligence remain the top challenges, while tool preferences and multimodal trends reshape the market.
Driven by exploding AI workloads, the data‑center networking landscape is shifting toward four distinct networks—Compute Fabric, Backend, Front‑end, and DCI—with forecasts showing AI backend network spend surpassing $100 billion by 2029, Ethernet outpacing InfiniBand, and massive port‑speed upgrades reshaping the market.
This article explores the evolution of AI infrastructure, comparing it with traditional backend systems, and details how hardware shifts to GPU-centric designs, software adaptations like deep learning frameworks, and engineering challenges in model training and inference can be addressed using established backend methodologies.
AI infrastructureGPU computingInference Optimization
This article examines how Volcano Engine's Elastic Instant Cache (EIC) tackles the memory bottleneck, high‑concurrency latency, and cross‑node coordination challenges of large language model inference by decoupling storage and computation, pooling resources, and applying layered optimizations, ultimately boosting AI inference efficiency, scalability, and cost‑effectiveness across various deployment scenarios.
This article explores how AI infrastructure has shifted from CPU‑centric designs to GPU‑driven architectures, detailing hardware evolution, software changes, and the engineering challenges of large‑model training and inference, while offering practical insights for traditional backend engineers transitioning to AI systems.
This investigative article traces Scale AI’s evolution from a MIT‑dropout’s data‑annotation startup to a $29 billion AI infrastructure leader, detailing its founder Alexandr Wang, core products, government contracts, competitive advantages, and the strategic shift toward defense‑focused AI solutions.
This article summarizes Alibaba Cloud’s key technical contributions at KubeCon China 2025, covering AI‑focused Kubernetes optimizations, Argo Workflows enhancements, storage strategies for large models, Fluid’s data orchestration, multi‑tenant security, and the RoleBasedGroup framework for PD‑separated AI inference.
This article explains the technical definitions, core differences, and practical use cases of Scale‑Up and Scale‑Out networking in AI systems, highlighting how they impact latency, bandwidth, and cost, and illustrates their combined application through NVIDIA's NVL72 supernode case study.
AI infrastructureGPU networkingHigh-Performance Computing
The article explains the concept, background, definition, challenges, roadmap, and key application scenarios of China's Computing Power Internet, highlighting its role in unifying fragmented compute resources, enabling on‑demand AI services, and driving nationwide digital transformation.
AI infrastructurecloud computingcomputing power internet
This article explains EasyRec's end‑to‑end recommendation system, covering its training‑inference architecture, a series of CPU/GPU and distributed optimizations, and a real‑time online‑learning pipeline that together improve throughput, latency, and model freshness.
AI infrastructureInference Optimizationdistributed computing
In a detailed AICon talk, Ant Group’s Baoling team leader Zhou Jun outlines their latest large‑model training techniques, MoE architecture optimizations, multimodal breakthroughs, open‑source releases, and the strategic roadmap needed to turn AI into a ubiquitous, “scan‑code‑level” everyday assistant.
AI infrastructureMixture of Expertslarge language models
In a Bloomberg Originals interview, OpenAI CEO Sam Altman discusses the massive “Stargate” infrastructure project, exploding demand for AI compute, multi‑partner collaborations, a projected $500 billion investment, GPU bottlenecks, and his vision for AI’s role in science, employment and humanity’s future.
This article explains how Baidu’s Kunlun supernode, built on high‑density liquid‑cooled cabinets and a modular 1U 4‑card design, breaks traditional 8‑card limits, boosts compute density four‑fold, improves power and cooling efficiency, and provides a scalable foundation for large‑model AI training and inference.
AI infrastructureGPU clusterHigh-Performance Computing
The article examines the rapid growth of unstructured data, explains the fundamentals and resource‑intensive nature of vector retrieval, presents Ant Group’s engineering practices—including hybrid HNSW‑DiskANN indexing, performance tricks like BSA pruning and memory prefetching, sparse‑vector and feedback‑driven recall improvements—and outlines the open‑source VSAG roadmap and ecosystem integrations.
At the 2025 Taipei International Computer Expo, Nvidia CEO Jensen Huang outlined the company's shift from a chipmaker to an AI infrastructure leader, introduced the concept of physical AI, and detailed upcoming hardware, software, and strategic initiatives that could reshape data centers, robotics, and autonomous driving.
The article analyzes China’s rapid expansion of intelligent computing centers, covering market forecasts, government policies, regional deployment patterns, major telecom operators' strategies, the industry value chain, and future trends that together signal a sustained surge in AI‑focused compute infrastructure across the nation.
The article analyses Nvidia’s shift from a chip supplier to a full‑stack AI infrastructure provider called AI Factory, explains the token‑economics framework that measures intelligent output, details the hardware‑software stack and network fabric, quantifies token consumption of advanced agents, and evaluates the strategic opportunities and risks for Nvidia.
The article examines Databricks' acquisition of the incremental computation engine Fennel, detailing how its unified batch‑stream processing, incremental updates, Python‑native development, and built‑in data governance can eliminate data silos, cut costs by up to 90 % and accelerate real‑time feature engineering for AI models, while also discussing industry impact and future roadmap.
PaddlePaddle Framework 3.0 delivers five breakthroughs—dynamic‑static unified automatic parallelism, integrated training‑inference pipelines, high‑order scientific differentiation, a neural‑network compiler with automatic operator fusion, and streamlined heterogeneous chip adaptation—drastically reducing development effort, boosting training speed, and expanding compatibility for large‑scale AI models.
AI infrastructureModel Inference OptimizationPaddlePaddle
The podcast recap dissects a year of rapid AI change, highlighting surprise‑fast open‑source model releases, shifting foundation‑model dynamics, the rise of GPT wrappers, over‑hyped agents, undervalued memory, product‑market fit debates, infrastructure opportunities, and lingering mysteries like RL in non‑verifiable domains.
This article analyzes Deepseek's EP‑based inference architecture for V3/R1 models, comparing it with TP, detailing how EP reduces memory and compute overhead, boosts batch size, cuts GPU memory usage, and introduces reliability, scalability, and maintainability challenges for large‑scale deployments.
AI infrastructureExpert ParallelismGPU memory optimization
This article provides a comprehensive analysis of DeepSeek, covering its foundational AI research, technology stack, product offerings, and the broader upstream, midstream, and downstream AI industry landscape, including hardware, server, cloud, and market trends.
This article provides an in‑depth analysis of DeepSeek's open‑source 3FS distributed storage system, focusing on its network communication module, RDMA‑based design, core classes such as IBSocket, Listener, and IOWorker, and how these innovations advance high‑performance AI infrastructure.
The article traces the evolution of embedding technology from a secret weapon of tech giants to a mainstream developer tool, explains the rapid rise and subsequent integration of vector databases into traditional search engines, and argues that vector databases are essentially search engines with added vector capabilities.
The surge in demand for private AI deployment has prompted multiple vendors to launch DeepSeek one‑stop machines—integrated hardware solutions that support the full DeepSeek model family, offering higher stability, easier setup, customization, cost savings, and data security across diverse industry scenarios.
The 2024 report of Alibaba Cloud's Infrastructure Network team details AI‑driven network breakthroughs, high‑performance protocol stacks, large‑scale monitoring systems, numerous top‑conference paper acceptances, open‑source ecosystem initiatives, and extensive industry outreach, highlighting the evolving AI infra landscape.
AI infrastructureConference PapersHigh-Performance Computing
The article examines the massive AI data‑center spending surge, highlighting Microsoft's $80 billion pledge, Amazon's $75 billion capex plans, market forecasts that AI servers will dominate the server market by 2025, and the sustainability concerns surrounding this rapid growth.
This presentation introduces the challenges of the large‑model era, describes the Colossal‑AI architecture—including N‑dimensional parallelism, heterogeneous storage, and zero‑code experience—shows benchmark results and real‑world use cases, and answers audience questions about its integration with PyTorch and advanced parallel strategies.
At AWS re:Invent 2024, Amazon unveiled its self‑designed AI hardware trio—Graviton 4 CPU, Nitro 5 DPU, and Trainium 2 accelerator—explaining the innovation, efficiency, and cost advantages driving the strategy, and detailing how these chips power next‑generation cloud services, ultra‑high‑performance servers, and massive AI super‑computing clusters.
Mooncake, an open-source KVCache-centric inference architecture co-developed by Alibaba Cloud and Tsinghua University's MADSys lab, dramatically improves large-model throughput and reduces cost by decoupling resources, standardizing cache pooling, and integrating with frameworks like vLLM, sparking broad industry interest.
The article analyzes Elon Musk’s Colossus AI supercomputer—its 100,000 NVIDIA H100 GPUs, record‑fast 122‑day construction, vertical‑integration strategy, and the broader implications for U.S. AI infrastructure dominance and China’s competing challenges in funding and chip supply.
This article provides a comprehensive analysis of China's AI server industry, covering upstream component markets, midstream shipment and revenue trends, downstream application demand, server classifications, major players, and future policy and technology drivers, all backed by recent market data and charts.
The article analyzes the growing demand for high‑bandwidth, low‑latency GPU Scale‑Up interconnects in AI clusters, explains why existing Ethernet and RDMA solutions fall short, and examines the industry‑wide UALink alliance and Alibaba's Alink System as a new open‑ecosystem solution.
eBay upgraded its AI infrastructure by adopting Ray, cutting model development and deployment time by roughly 50% and boosting GPU utilization from about 10% to over 75% through automated cluster scaling and high‑throughput batch inference.
The talk outlines Baidu Smart Cloud's comprehensive cloud‑native redesign—including ultra‑elastic compute, AI‑focused storage, high‑performance networking, AI‑driven operations, and edge‑distributed services—illustrated with automotive and fintech case studies that demonstrate how enterprises can accelerate digital transformation in the AI‑native age.
This article explains the motivations behind RDMA, describes its architecture, key components, and protocols such as RoCEv2, and discusses future technical challenges for scaling RDMA in large AI and HPC data‑center networks.
The article details the design and deployment of 360's AI Compute Center, covering GPU server selection, high‑performance networking, Kubernetes‑based cluster management, advanced scheduling, training and inference acceleration techniques, and a comprehensive AI development platform with visualization and fault‑tolerance features.
Alibaba Cloud's server R&D team presented three research papers at IEEE EPEPS 2024 and PCB West 2024 covering high‑density PCIe 6.0 crosstalk optimization, 100G‑PAM4 Ethernet performance under air and immersion cooling, and sustainable low‑cost PCB materials for immersion‑cooled computer systems, highlighting their relevance to AI infrastructure and data‑center design.
This article details the design and implementation of 360’s AI Computing Center, covering server selection, network topology, Kubernetes scheduling, training and inference acceleration, and the AI platform’s core, visualization, and fault‑tolerance capabilities for large‑scale AI workloads.
This article analyzes Baidu's Baige 4.0 AI infrastructure, detailing its four‑layer architecture, XMAN 5.0 hardware, HPN network, BCCL communication library, and AIAK inference upgrades, and explains how these innovations address large‑model training and inference challenges while boosting performance, utilization, and cost efficiency.
AI infrastructureGPU accelerationHigh-Performance Computing
The article details Baidu Baige 4.0’s four‑layer AI infrastructure—hardware, cluster components, training‑inference acceleration, and platform tools—highlighting its heterogeneous computing, high‑performance networking, fault‑tolerant communication library, and optimizations that boost large‑model training and inference efficiency.
AI infrastructureheterogeneous computinghigh performance networking
The article discusses the challenges of training ever‑larger AI models on internet‑scale data, critiques traditional batch ETL pipelines, and proposes a streaming data‑flow architecture with dynamic data selection and a shared‑memory/Alluxio middle layer to decouple data processing from model training, improving efficiency and scalability.
In this speech, Wang Jian reflects on the evolution of artificial intelligence, arguing that modern AI is fundamentally different from its early concepts, emphasizing the pivotal roles of data, models, and infrastructure, and exploring the transformative impact of AI+, transformers, and cloud platforms on future innovation.
The article explains what an Intelligent Computing Center (智算中心) is, analyzes its extensive upstream and downstream industry chain, describes the cutting‑edge AI computing architecture that powers it, forecasts massive growth in AI compute capacity by 2028, and outlines regional deployment strategies and service models such as leasing, data, operation, and talent cultivation.
AI computingAI infrastructureIntelligent Computing Center
In this speech, academician Wang Jian reflects on the short, seven‑year history of modern AI, distinguishes AI, AI+ and AI infrastructure, explains how data, models and compute power have become the new foundational layer, and examines the roles of Google, OpenAI, transformers, and cloud services in shaping today’s AI revolution.
This article dissects Meta’s Llama 3 405‑billion‑parameter model, covering its dense Transformer design, data‑mixing strategy, two‑stage scaling‑law prediction, 4‑D parallelism, custom hardware clusters, training schedules, post‑training alignment methods, and the extensive evaluation results that benchmark it against leading LLMs.
The article traces the evolution of large‑model training and inference infrastructure from the early “black‑iron” era (2019‑2021) through the “golden” boom (2022‑2023) to the emerging “silver” phase (2024‑), highlighting key research breakthroughs, open‑source frameworks, hardware trends, market dynamics, and practical challenges for engineers entering the field.
This article examines the Model-as-a-Service (MaaS) paradigm, tracing its origins, defining its expanded capabilities for large‑model ecosystems, outlining the full‑stack services it offers, and analyzing current industry adoption, deployment models, and the technical and regulatory challenges that must be addressed for scalable AI rollout.
This article discusses the historical background, major challenges such as massive compute and memory demands, and the technical ecosystem—including data parallelism, pipeline parallelism, and optimization strategies like DeepSpeed and 1F1B—to enable efficient distributed training of large language models.
AI gateways have emerged as essential infrastructure for modern AI applications, offering specialized security, load balancing, cost management, and observability that go beyond traditional API gateways, and understanding their differences and deployment considerations is crucial for developers and ops teams.
This analysis explores how large model technology drives explosive AI compute demand, enables cost-effective MaaS offerings, and forces cloud providers like Baidu, Alibaba, and Tencent to pivot from traditional services to AI-native ecosystems amid market saturation.
DeepRec Extension enhances large‑scale sparse model training by adding automatic elastic training, resource‑aware scheduling, real‑time monitoring, and efficient fault‑tolerance mechanisms, enabling lower cost, higher throughput, and more reliable distributed training for AI workloads.
Baidu Baige’s AIAK‑LLM suite accelerates large‑model training and inference by boosting Model FLOPS Utilization through techniques such as TP communication overlap, hybrid recompute, zero‑offload, automatic parallel‑strategy search, multi‑chip support, and inference‑specific optimizations, achieving over 60 % speedup and seamless Hugging Face integration.
The talk outlines how Baidu’s Baige AIAK‑LLM suite tackles the exploding compute demands of trillion‑parameter models by boosting Model FLOPS Utilization through advanced parallelism, memory‑saving recompute, zero‑offload, adaptive scheduling, and cross‑chip orchestration, delivering 30‑60% training and inference speedups and a unified cloud product.
The article explains the scaling challenges of ever‑larger LLMs, introduces the MFU performance metric, surveys industry parallelism and memory‑saving techniques, and details Baidu’s AIAK‑LLM suite—including resource, component and acceleration layers—as well as concrete training and inference optimizations that raise MFU by 30‑60% and cut deployment costs.
This weekly tech roundup covers Microsoft’s upcoming mobile game store, Alibaba Cloud’s Tongyi Qianwen 2.5 AI model, Google DeepMind’s AlphaFold 3 for drug discovery, TikTok’s AI‑content labeling, 神州信息’s AI‑native product, Apple’s on‑device AI chips, expert views on scaling laws, and news on Fei‑Fei Li’s startup, Apple’s China tax, and Buffett’s Apple stake reduction.
The article examines how the surge of generative AI (AIGC) is fueling rapid growth in AI server demand, detailing the emerging AIGC ecosystem, server hardware composition, model scaling, heterogeneous computing, training vs. inference workloads, market size forecasts, and the competitive landscape of AI server manufacturers.
This interview‑style guide explores how vector databases handle massive, high‑dimensional time‑series data for quantitative stock trading, detailing data scaling challenges, selection criteria, and why the research team chose LanceDB over alternatives for efficient, scalable financial analysis.
The article analyzes the computational characteristics and system challenges of training large AI models on domestic platforms, examines framework parallelism and future algorithms, and proposes six strategic measures—including scaling compute, improving data management, building a national R&D team, and boosting AI‑chip investment—to accelerate China’s AI leadership.
This article reviews the rapid development of large language models since 2019, outlines the historical background, identifies key challenges such as massive compute demand, memory constraints, and system complexity, and then details distributed training technologies—including data parallelism, pipeline parallelism, and advanced optimization strategies—while also discussing future research directions and answering common questions.
The rapid growth of AI model sizes has created unprecedented demands on network bandwidth, latency, stability, and automation, making efficient RDMA‑based interconnects, advanced congestion control, and intelligent deployment essential for scaling distributed training clusters to thousands of GPUs.
The article analyzes the compute, memory, and communication resources required to train and run large language models, quantifies bottlenecks such as the massive FLOP demand, terabyte‑scale GPU memory, and high‑bandwidth interconnect needs, and evaluates parallelism strategies and bandwidth estimates to guide hardware and software design for scaling LLMs.
The article examines how large AI models are moving toward a unified, low‑knowledge‑density paradigm that raises computational efficiency challenges across model, algorithm, framework, and infrastructure layers, while also highlighting NVIDIA's GTC 2024 China AI Day sessions that showcase practical solutions and upcoming training opportunities.
The article explains why collective communication is critical for distributed large‑model training, outlines the new requirements for system reliability, and introduces Baidu’s Collective Communication Library (BCCL), detailing its enhanced observability, fault‑diagnosis, stability, and performance optimizations that raise effective training time to 98 % and bandwidth utilization to 95 %.
AI infrastructureCollective CommunicationFault Diagnosis
Baidu’s Collective Communication Library (BCCL) enhances large‑model distributed training by improving real‑time bandwidth monitoring, fault diagnosis, network stability, and performance, leveraging RDMA networks and GPU‑specific optimizations to increase effective training time to 98% and bandwidth utilization to 95%.
AI infrastructureCollective CommunicationFault Diagnosis
The article details JD Retail's evolution from TensorFlow‑based sparse training to a custom high‑performance parameter server and a fully GPU‑accelerated, multi‑node, multi‑card synchronous training framework that leverages GPU‑RDMA, two‑level CPU‑DRAM/GPU‑HBM caching, and pipeline parallelism to overcome storage, I/O, and compute challenges of trillion‑parameter recommendation systems.
This article introduces Colossal‑AI, an open‑source platform that tackles the challenges of training, fine‑tuning, and deploying massive AI models by leveraging efficient memory management, N‑dimensional parallelism, and high‑performance inference to dramatically reduce cost and improve scalability across thousands of GPUs.
In a detailed interview at the 2023 Wuzhen Internet Forum, Alibaba Cloud’s infrastructure lead Cai Dezhong outlines the three‑phase IPv6 rollout, highlights organizational and technical innovations, stresses the need for global cooperation, and explains how IPv6 underpins the next generation AI infrastructure and predictable high‑performance networking.
The article summarizes a Baidu Cloud Intelligence conference speech and whitepaper, explaining how AI foundations and large‑model infrastructure reshape applications, boost enterprise digital transformation, and drive regional economic development, offering a roadmap for high‑quality digital economy advancement.
AI infrastructureEnterprise Transformationdigital economy
At the 2023 Baidu Cloud Intelligence Conference, Baidu AI Cloud outlined a comprehensive, four‑layer solution—spanning distributed cloud infrastructure, AI‑focused compute, data compliance, and end‑to‑end toolchains—to address the challenges of electric, intelligent vehicles, large‑model deployment, and regulatory compliance in autonomous driving.
AI infrastructureAutonomous Drivingcloud computing
BladeLLM, Alibaba Cloud’s large‑model inference engine, pushes the limits of LLMs by supporting ultra‑long context lengths up to 70 K tokens, leveraging novel RaggedAttention and a DNN‑based AutoTuner to deliver superior performance, memory efficiency, and low‑latency inference across diverse workloads.
The article explains how AIGC models demand massive GPU compute, why network bandwidth and latency become the critical limiting factors, and how the Distributed Disaggregated Chassis (DDC) architecture addresses these challenges with scalable, high‑throughput networking solutions.
The May 15 Beijing forum gathered experts from leading tech firms and research institutes to discuss immersion liquid cooling as a vital solution for AI infrastructure's growing compute and thermal challenges, presenting current trends, technical designs, material research, and future sustainable development directions.
The article details Baidu's development of a massive high‑performance GPU/IB cluster, its architectural design, the challenges of training trillion‑parameter models, and how the integrated AI stack—spanning hardware, framework, and resource management—overcomes compute, memory, and communication bottlenecks to accelerate large‑model training.
Over the past two decades, AI infrastructure has evolved from early distributed storage and MapReduce to GPU programming, modern package managers, in‑memory processing, deep‑learning frameworks, parameter servers, AI compilers, synthetic data pipelines, open‑source model hubs, and today’s large‑scale Kubernetes‑based clusters, forming the essential foundation for every breakthrough.
AI CompilersAI infrastructureDeep Learning Frameworks
This article explains how Baidu built a massive, high‑performance GPU/IB cluster, optimized its architecture and software stack, and integrated AI frameworks and resource management to overcome compute, memory, and communication bottlenecks, enabling efficient training of trillion‑parameter large models.
The article presents Baidu's comprehensive approach to federated‑learning interoperability, covering the underlying principles, the JinKe Alliance bottom‑layer solution, the high‑level HIGHFLIP protocol, and a comparative discussion of white‑box, gray‑box, and black‑box integration strategies.
Alibaba Cloud’s PAI platform unveils TePDist, an HLO‑based automatic distributed deep‑learning system that decouples strategy search from model code, offers client/server architecture, supports SPMD and pipeline parallelism, delivers high performance on GPT, MoE and other models, and is now open‑source.
SenseTime introduced its visual‑centric multimodal large‑model platform SenseNova, detailing model scaling, extensive AI infrastructure, diverse industry deployments such as autonomous driving and generative content, and the challenges of compute efficiency and data acquisition in the race for advanced AI.
The article analyses the computational resources, hardware costs, and human investment needed to train and serve large language models such as GPT‑3, discusses practical cost calculations, highlights the challenges faced by Chinese AI teams, and argues for sustained, long‑term funding to achieve meaningful breakthroughs.
ColossalChat, an open‑source project built on LLaMA, offers a full RLHF pipeline—including supervised fine‑tuning, reward‑model training, and reinforcement learning—enabling low‑cost, bilingual ChatGPT‑like models with 4‑bit quantized inference, detailed code, dataset, and performance optimizations.
This article analyzes Tencent's AngelPTM framework, detailing its ZeRO-Cache strategy, unified storage management, multi‑stream async execution, SSD tiered storage, and performance benchmarks that show up to 95% larger model capacity and over 44% speedup compared to community solutions.
The article explains how the massive compute and storage demands of today’s large language models create a “compute wall” and “storage wall,” and describes Baidu Intelligent Cloud’s four‑layer full‑stack infrastructure—combining advanced parallelism techniques, optimized GPU networking, static‑graph compilation, and cost‑model‑driven placement—to train trillion‑parameter models efficiently.
The article explains that reproducing ChatGPT in China is not just a matter of funding but requires extensive expertise in large‑scale language model training, massive compute resources, optimized cloud infrastructure, and deep AI research, as demonstrated by Alibaba's DAMO Academy efforts.
This article explores Hulu’s five‑year‑old machine‑learning training platform, detailing its three‑layer architecture, the shift from single‑node to distributed training, and the technical solutions—including parameter servers, Ring AllReduce, Kubernetes, Volcano, and Horovod—that enable scalable AI workloads across GPU, CPU, and storage resources.
In this conference talk, He Zhengyu of Ant Group outlines the challenges of large‑model AI, proposes a green, shared, model‑centric infrastructure built on foundation models, cloud‑native MLOps, and Model‑as‑a‑Service (MaaS) to lower cost and accelerate AI adoption across industries.
This article details Tencent's development of the 1‑trillion‑parameter HunYuan‑NLP model, covering its MoE architecture, cost‑effective pre‑training strategies, distributed training framework, model compression toolkit, and successful deployment across advertising, gaming, and other Tencent services.
This article explains how Baidu's intelligent cloud overcomes the compute and storage walls of large‑scale model training by combining hardware design, network topology, and software optimizations such as pipeline, tensor, and expert parallelism, cost‑model‑driven placement, and future‑proof AI infrastructure evolution.
The DataFun Summit 2022 brings together leading AI researchers and industry experts to discuss deep‑learning frameworks, ultra‑large model training, AI chips, compilers, MLOps, and end‑to‑end AI infrastructure, offering live streaming of six thematic forums and dozens of technical talks.
Xiaohongshu’s in‑house LarC platform powers real‑time, multimodal recommendation, life‑search, and generative‑AI commercial content for its 200 million‑user community by processing billions of daily feedback samples, employing conflict‑free parameter servers, diversified sequence modeling, and large‑scale representation learning to deliver personalized, fresh, and diverse user experiences.
AI infrastructureMachine Learning PlatformRecommendation System
This article explores Baidu’s cutting‑edge GPU container virtualization architecture, detailing the challenges of low GPU utilization in AI workloads, the dual‑engine (user‑space and kernel‑space) isolation mechanisms, various mixing strategies, performance evaluations, and best‑practice recommendations for maximizing resource efficiency in large‑scale AI deployments.
AI infrastructureContainerizationGPU virtualization
The article examines the industry trend toward ever‑larger AI models, compares their parameter scale to the human brain, outlines the computational and memory challenges of training such models, and details advanced parallelism techniques and Baidu's high‑performance cluster solutions that enable efficient, stable large‑scale model training.
The article examines how monolithic AI toolchains hinder innovation, recounts the historical fragmentation of software in the 1990s, highlights LLVM's modular architecture as a turning point, and argues for a new, composable AI infrastructure to make machine learning more accessible and scalable.
DeepRec, developed since 2016 by Alibaba, is a specialized sparse‑model training engine that addresses feature elasticity, training performance, and deployment challenges through dynamic elastic features, optimized runtimes, distributed training frameworks, incremental model export, and multi‑level storage, and is now being open‑sourced for broader industry collaboration.
This article details Weibo's large‑scale multimodal content understanding platform, covering its background, data and model heterogeneity challenges, the end‑to‑end workflow, GPU‑heterogeneous cluster design, resource scheduling, performance optimization for distributed training and online inference, and comprehensive monitoring to ensure stable, low‑latency AI services.
This presentation explains how knowledge graphs serve as the foundation for new‑infrastructure initiatives, detailing the evolution of AI from perception to cognition, the role of big‑data centers, DIKW modeling, intelligent data governance, and the construction of a cognitive AI middle‑platform for industry applications.
TensorNet is a TensorFlow‑based distributed training framework that tackles the challenges of massive data and billions of sparse parameters in advertising and recommendation systems by enabling near‑infinite sparse feature dimensions, drastically reducing synchronization overhead, and delivering up to 35% inference speed improvements.
JD Digital Science introduced its fast secure federated learning framework, highlighted two pioneering technologies—a kernel‑based nonlinear federated learning algorithm and a distributed fast homomorphic encryption method—both accepted at KDD 2020, and discussed their industrial applications, privacy benefits, and regulatory relevance.