Tagged articles
TPU
51 articles · Page 1 of 1
This article explains why specialized AI chips like Google's TPU and mobile NPUs are needed, details their architectures (systolic arrays, SoC integration), compares generations and vendors (Ascend, Cambricon, Edge TPU, Habana), and covers key metrics (TOPS, TOPS/W) and market landscape across training, inference, and edge.
AI AcceleratorsAI chip marketASIC
0 likes · 11 min read

Google open-sourced TorchTPU, a software stack enabling native PyTorch execution on TPUs via device_map="tpu", achieving 97% numerical parity with GPU across 4,917 models and performance within single-digit percentages of JAX, with integrations across vLLM, SGLang, Hugging Face, and TorchTitan.
AI hardwareCUDAGoogle
0 likes · 13 min read

Google's Gemini 4 Argon achieves breakthrough 1M output tokens, top benchmarks, and aggressive pricing via vertical TPU integration, but the article argues Google remains trapped in a reactive cycle due to innovator's dilemma, organizational inertia, and competitor ecosystem lock-in, making its lead potentially temporary.
AI competitionAgentic workflowsGemini 4 Argon
0 likes · 30 min read

Google's eighth-generation TPU introduces two specialized chips: TPU 8t with 3D Torus, 9,600-chip pods, 2PB shared HBM, and SparseCore for massive pre-training, and TPU 8i with Boardfly topology, 1,152-chip pods, CAE, and high HBM bandwidth for low-latency Agentic AI inference.
3D TorusAI AcceleratorsAgentic AI
0 likes · 10 min read

Anthropic has hired former Google TPU chief Amir Salek to lead its in‑house chip effort, signaling a shift from relying on external accelerators to building custom hardware as AI companies intensify competition over compute efficiency and cost.
AI hardwareAmir SalekAnthropic
0 likes · 5 min read

Google is intensifying its AI arms race by urging founder Sergey Brin to champion Recursive Self‑Improvement, reshaping Gemini development, reorganizing DeepMind leadership, and confronting TPU resource limits, as the company seeks to accelerate AI research while balancing commercial pressures and internal structural challenges.
AIAI strategyDeepMind
0 likes · 10 min read


21CTOAug 10, 2026 · Industry Insights
The article reviews Jeff Dean’s 27‑year career at Google, highlighting six transformative systems—MapReduce, BigTable, Spanner, TensorFlow/DistBelief, TPU, and other foundational tools—that together reshaped large‑scale data processing, storage, and AI infrastructure, cementing Google’s dominance in the digital world.
BigtableGoogleJeff Dean
0 likes · 6 min read

In a candid YC Startup School interview, Jeff Dean explains how AI is moving from ever‑larger models to long‑running, self‑evolving agents, emphasizing inference hardware, data‑movement costs, context engineering, and the strategic 1% rule that guides startups to focus on problems where generic models still fail.
AIAgentsStartup Strategy
0 likes · 26 min read

The article forecasts ASIC shipments reaching 7.7 million units in 2026 (45% market share) and surpassing GPUs by 2027 (58% share), analyzes TSMC's CoWoS capacity growth, details AWS's Nitro‑Graviton‑Trainium roadmap and network redesign, examines Google's TPU v1‑v8 evolution, and compares Alibaba's Pingtouge chips and revenue impacts across the cloud AI market.
AI AcceleratorsASICAWS
0 likes · 11 min read


SuanNiMay 21, 2026 · Artificial Intelligence
Google’s I/O 2026 keynote announced a full‑scale shift to the Gemini agent era, detailing new 8th‑gen TPUs, the Gemini 3.5 Flash model with higher Elo scores and lower cost, multimodal Omni Flash, expanded Agent tools like Antigravity and Spark, revamped search, commerce protocols, creative suites, and AI‑driven scientific applications.
AI agentsGeminiGoogle AI
0 likes · 13 min read

Google I/O 2026 unveiled Gemini 3.5 Flash—a faster, cheaper flagship model now fully open—alongside the multimodal Gemini Omni video generator, the 24/7 personal AI agent Gemini Spark, the biggest search overhaul in 25 years, upgraded Antigravity 2.0, new TPU 8 chips and refreshed AI subscription plans.
AI agentsGeminiGoogle I/O
0 likes · 15 min read

The article breaks down the seven major AI‑focused processors—CPU, GPU, NPU, TPU, LPU, DPU, and VPU—explaining each one's architectural strengths, typical workloads, representative vendors, and trade‑offs, then summarizes which role each chip excels at in modern AI systems.
CPUDPUGPU
0 likes · 9 min read

Anthropic’s move from GPU‑based training to a dual compute pool of Google TPUs and Amazon Trainium promises up to 40% lower training costs, while the article compares the hardware efficiencies, market shares, and strategic risks across Google, OpenAI, Nvidia, and Chinese open‑source AI chip camps.
AI hardwareAnthropicClaude
0 likes · 6 min read

The article examines Nvidia’s competitive moat beyond CUDA, detailing how its token‑factory model, extensive supply‑chain commitments, and a flexible accelerator ecosystem contrast with Google’s TPU ASIC approach, while also exploring the impact of AI agents on future compute demand.
AI hardwareCUDANVIDIA
0 likes · 7 min read


SuanNiApr 29, 2026 · Artificial Intelligence
Google’s Cloud Next 2026 reveal splits the 8th‑generation TPU into training‑focused Sunfish and inference‑focused Zebrafish, highlighting Ironwood’s record‑breaking performance, a multi‑vendor supply chain, Anthropic’s multi‑gigawatt order, and a broader industry shift toward custom AI chips that promise far higher profit margins than generic GPUs.
AIGoogleTPU
0 likes · 8 min read

Google unveiled its eighth‑generation TPU line at Cloud Next 2026, introducing the training‑focused TPU 8t with a 2.7× performance boost and massive scaling, and the inference‑optimized TPU 8i featuring three‑times more on‑chip SRAM and an 80% performance uplift for agentic AI workloads, while positioning the chips as a complement—not a replacement—to Nvidia's offerings.
AI hardwareAgentic AIGoogle Cloud
0 likes · 9 min read

Anthropic has signed a multi‑year agreement with Google and Broadcom to lock in multiple gigawatts of next‑generation TPU capacity starting in 2027, a move driven by Claude’s soaring demand, revenue surpassing $30 billion and a rapid doubling of high‑spending enterprise customers.
AI computeAnthropicClaude
0 likes · 5 min read

This article explains why TPU programming differs from GPU, describes the explicit HBM‑VMEM‑register data movement required on TPU, introduces the Pallas grid‑BlockSpec‑Ref model, and walks through four progressively more complex kernels—including element‑wise add, tiled dot product, fused RMSNorm with scratch memory, and a production‑grade FlashAttention implementation—showing how each kernel maps to the TPU memory hierarchy and leverages Pallas features such as input_output_aliases and PrefetchScalarGridSpec.
FlashAttentionJAXPallas
0 likes · 20 min read

Google’s Tensor Processing Unit (TPU) uses a systolic array architecture and low‑precision quantization to overcome the Von Neumann bottleneck, delivering orders‑of‑magnitude higher throughput and energy efficiency for matrix‑multiplication‑heavy AI workloads—from AlphaGo’s inference to today’s massive language models.
AI hardwareGoogleTPU
0 likes · 15 min read

The article explains Google’s TPU data‑center architecture, detailing the vertical Scale‑Up strategy within a SuperPod, the horizontal Scale‑Out across SuperPods, the 3D Torus topology with Twisted variants, and the multi‑layer network design that enables petabyte‑scale AI training and inference.
AI hardwareScale-outSuperPod
0 likes · 8 min read

The article examines the shifting AI‑chip landscape, explaining how Google’s TPUv7, backed by massive pod architecture and optical circuit switching, challenges Nvidia’s GPU dominance by offering superior system‑level efficiency and lower total cost of ownership for large‑scale model training.
AI hardwareGPULarge-scale AI training
0 likes · 12 min read

The article explains how Google’s TPU v7 supercomputer uses a simple yet powerful networking scheme—1.5 optical modules per TPU for intra‑rack communication and an additional 2.6 modules per TPU for inter‑rack high‑speed links—enabling massive AI model training with balanced cost and performance.
AI supercomputingGoogleTPU
0 likes · 13 min read

The article examines Google’s Gemini 3 Pro launch, highlighting its full‑stack vertical integration, advanced System 2 reasoning, dynamic compute budgeting, native multimodal architecture, TPU cost advantages, the Antigravity IDE platform, generative UI capabilities, and the strategic implications for Google’s AI ecosystem and competitive positioning.
AI infrastructureAntigravityGemini 3 Pro
0 likes · 32 min read

This article introduces Google’s Tunix library for JAX‑based LLM post‑training, explains its core features such as supervised fine‑tuning, reinforcement learning and knowledge distillation, and provides detailed installation steps and a complete TPU‑accelerated QLoRA fine‑tuning workflow on the Gemma 2B model, including code snippets and inference testing.
AIJAXLLM
0 likes · 8 min read

Google reveals that a single Gemini query now consumes only 0.24 Wh of electricity, emits 0.03 g CO₂e and uses about five drops of water, thanks to a comprehensive measurement framework and aggressive optimizations across model architecture, quantization, hardware design, and data‑center operations.
AI energyAI sustainabilityGoogle Gemini
0 likes · 8 min read

The article examines the limitations of traditional von‑Neumann CPUs and power‑hungry GPUs for modern AI workloads, explains the rise of ASIC and FPGA based AI accelerators, compares major industry solutions, and highlights why reconfigurable, low‑power AI chips are becoming essential for robotics and edge computing.
AI chipsASICFPGA
0 likes · 11 min read

The article traces the evolution of AI hardware over the past ten years, outlining three key phases—from early chip limitations that sidelined neural networks, through CPU advances that still fell short, to the rise of GPUs and specialized AI chips that finally unlocked rapid AI deployment, while also highlighting the parallel impact of algorithmic breakthroughs and massive data growth.
AI hardwareGPUTPU
0 likes · 5 min read

The article analyzes Google’s Made by Google 2025 event, showing how the Pixel 10 lineup, the Tensor G5 chip, Gemini Nano, and a full‑stack AI infrastructure—including custom TPUs, AI Hypercomputer, and Vertex AI—form a coordinated on‑device AI strategy that challenges Apple and builds a long‑term economic moat.
AI strategyGeminiGoogle
0 likes · 25 min read

This article breaks down the hardware bottlenecks of large‑scale LLM training, explains the Roofline performance model, arithmetic intensity, and how computation and communication costs interact on GPUs and TPUs, offering concrete formulas and examples for efficient scaling.
Arithmetic intensityGPULLM training
0 likes · 12 min read

This article explains what ASIC chips are, how they differ from CPUs, GPUs and FPGAs, classifies them by customization level and function, outlines their performance and cost advantages, discusses their drawbacks, and reviews current products and market trends driving AI hardware adoption.
AI hardwareASICFPGA
0 likes · 11 min read


21CTOMay 19, 2025 · Artificial Intelligence
In a candid All‑In podcast interview, Google CEO Sundar Pichai explains how the company is reshaping search with AI‑driven assistants, leveraging its custom TPU infrastructure, expanding into quantum computing, robotics, and new hardware while confronting energy constraints and fierce global competition.
AIGoogleRobotics
0 likes · 23 min read

Jeff Dean’s 2024 ETH Zurich talk traces fifteen years of AI breakthroughs—from the rise of neural networks and back‑propagation, through large‑scale distributed training, TPUs, Transformers, sparse MoE models, and advanced prompting techniques—showing how scaling compute, data, and clever software have driven today’s powerful Gemini models.
AIChain-of-ThoughtSparse Models
0 likes · 18 min read

This article outlines the development of Google's Tensor Processing Units (TPU) from the first generation to the latest seventh‑generation chip, detailing architectural improvements, performance specifications, integration into data‑center pods and mobile devices, and concludes with references to related AI‑hardware resources and promotional material.
AI hardwareGoogleTPU
0 likes · 10 min read

This article explains the rise of AI‑specific processors, defines AI chips, compares their architectures, and examines the distinct requirements of training versus inference chips while outlining the main technology routes (GPU, FPGA, ASIC) and future outlook.
AI chipsASICDSA
0 likes · 9 min read

This article analyzes Google’s Tensor Processing Unit (TPU) as a purpose‑built AI ASIC, tracing its evolution from early GPGPU and FPGA solutions, detailing its MXU systolic‑array design, low‑precision advantages, performance benchmarks, power efficiency, cluster interconnect innovations, and software integration with TensorFlow.
AI hardwareASICGoogle
0 likes · 15 min read

This article explains the fundamental performance indicators of AI chips—TOPS, TFLOPS, and precision formats like FP16, FP32, and INT8—while comparing GPU, ASIC, and TPU architectures, highlighting Tensor Core advantages and TPU's superior efficiency over CPUs and GPUs.
AI chipASICFP16
0 likes · 4 min read

While GPUs have driven AI acceleration for years, this article analyzes their architectural constraints, compares emerging alternatives such as Google's TPU and high‑end FPGAs, and explores future application niches like VR/AR, cloud gaming, and military systems where GPUs may still thrive or be replaced.
AI hardwareFPGAGPU
0 likes · 15 min read

This article explains the essential metrics of AI chips—including TOPS and TFLOPS compute, precision formats like FP16, FP32 and INT8, and the roles of GPUs, ASICs and TPUs—while highlighting how Tensor Cores boost deep‑learning performance and comparing TPU efficiency to CPUs and GPUs.
AI chipsASICFP16
0 likes · 4 min read

Google has released the open‑source Gemma large language model in 2 B and 7 B parameter versions, claiming superior performance to Llama 2 and Mistral across 18 benchmarks, especially in math and code, while running on laptops, desktops, IoT and cloud devices.
AIGemmaLlama 2
0 likes · 10 min read

Google’s Gemini, a trillion‑parameter native multimodal model trained on TPU v4/v5e, was launched overnight and, according to its technical report, surpasses GPT‑4 on 30 of 32 academic benchmarks, achieves the first human‑level score on MMLU, and powers the new AlphaCode 2 code‑generation system.
AlphaCode 2GPT-4Gemini
0 likes · 11 min read

The article explains the various AI‑capable chips such as CPUs, GPUs, FPGAs, NPUs, and TPUs, compares their performance and efficiency, describes heterogeneous CPU+xPU solutions, and provides market share data while highlighting the growing adoption of specialized AI accelerators.
AI accelerationAI chipsCPU
0 likes · 7 min read

The article surveys the rapid entry of leading vendors into the AI ASIC market, detailing Google’s TPU generations, Intel’s acquisition of Habana Labs and the Gaudi 2 chip, IBM’s upcoming AIU, Samsung’s Warboy NPU, and the performance, architectural, and future trends of ASICs for AI inference and training.
AI ASICGaudiTPU
0 likes · 11 min read

The article analyzes how leading vendors such as Google, Intel, IBM, Samsung, Nvidia and AMD are racing to develop AI ASICs, compares their architectures and performance, and projects a rapid rise in ASIC market share for both data‑center and edge AI workloads by 2025.
AI ASICGaudiTPU
0 likes · 13 min read

The article analyzes how GPUs, once dominant in accelerating AI workloads, now face limitations in precision, energy efficiency, and on‑chip networking, prompting a shift toward specialized accelerators like Google's TPU and FPGA solutions, while also exploring emerging GPU‑friendly scenarios such as VR/AR, cloud gaming, and military applications.
FPGAGPUTPU
0 likes · 11 min read

This article reviews a decade of Google TPU development, highlighting ten technical and architectural lessons, the hardware's impact on the AI industry, performance and energy‑efficiency improvements, and strategies for reducing machine‑learning carbon footprints.
Domain-specific ArchitectureGoogleMachine Learning Hardware
0 likes · 19 min read

The article reviews how deep learning models have grown deeper and wider, discusses the memory and bandwidth limits of single GPUs, and explains pipeline and sharding techniques—including GPU clusters and TPU pods—to efficiently train large‑scale models in industrial settings.
GPUPipelineTPU
0 likes · 6 min read

This article analyzes how system architects can select the optimal heterogeneous compute cores—CPU, GPU, FPGA, ASIC, VPU, or TPU—for edge AI deployments, weighing performance, size, weight, power, and cost to maximize inference efficiency and security.
AI edge computingASICCPU
0 likes · 7 min read

The article explains why dedicated AI chips (ASICs) are needed, compares their performance and power efficiency to traditional CPUs and GPUs, describes the architecture of Google's TPU and other AI accelerators, and provides historical context for the evolution of AI hardware.
AI chipASICCPU vs GPU
0 likes · 10 min read

This article compares various AI server architectures—CPU, GPU, FPGA, TPU, and ASIC—by evaluating performance versus programmability, and outlines practical guidelines for choosing GPU servers based on workload, cost, power, and deployment scenarios.
AI serversASICCPU
0 likes · 8 min read

Deep neural networks, propelled by breakthroughs such as AlexNet and advances in GPU and TPU hardware, are rapidly moving from academic research into diverse applications—including earthquake prediction, medical imaging, and autonomous driving—driving massive industry investment, new semiconductor designs, and intense competition among tech giants and startups.
AI hardwareComputer VisionGPU
0 likes · 9 min read


21CTOJul 26, 2018 · Cloud Computing
Google Cloud NEXT 18 in San Francisco unveiled the alpha‑tested Cloud TPU 3.0, major AutoML enhancements, and the Contact Center AI solution, while CEO Diane Greene highlighted AI and security investments and the cloud’s rapid revenue growth, signaling Google’s push to outpace AWS, Azure, and IBM.
AIAutoMLGoogle Cloud
0 likes · 7 min read
