SemiPulse | AI-Powered Semiconductor Supply Chain Intelligence & Market Signals

According to a Fubon Research report, Google plans to produce 12 to 15 million TPU v9 chips by 2028, potentially surpassing NVIDIA’s AI GPU shipments in the same period. This marks a significant shift in the AI hardware landscape, as Google’s self-developed chips are rapidly scaling in deployment. T

AI accelerator Google TPU NVIDIA GPU Semiconductor manufacturing Artificial Intelligence chips TSMC Intel Foundry AI infrastructure

Google could build more AI accelerators than Nvidia sells in 2028, analyst claims — could push the company to use Intel Foundry to meet its goals - Tom's Hardware

0.85

www.tomshardware.com 2026-07-30

Samsung, Broadcom to Jointly Develop Next-Generation HBM-Based AI Accelerators - thelec.net

0.95

news.google.com 2026-07-27

Meta to use custom AMD Instinct MI400 accelerators with 144GB of HBM4 for select workloads, report claims

0.85

tomshardware.com 2026-07-22

According to SemiAnalysis, Meta plans to use a customized AMD Instinct MI400 accelerator for recommendation systems, featuring 144GB of HBM4 memory—significantly less than the full-fledged Instinct MI455X. This design aims to reduce costs by cutting hardware expenses, particularly beneficial for soc

Meta AMD AI accelerator recommendation system HBM4 memory cost optimization semiconductor technology artificial intelligence

Chinese companies are ditching Nvidia’s advanced accelerators for domestic AI suppliers - Fortune

0.92

fortune.com 2026-07-08

Chinese companies are increasingly shifting away from NVIDIA's advanced AI accelerators toward domestic silicon suppliers, reflecting intensifying US-China tech rivalry and Beijing's push for technological self-reliance. According to a Bloomberg Intelligence survey, Chinese firms plan to increase th

Artificial Intelligence Semiconductor NVIDIA Chinese Technology AI Accelerators Domestic Substitution Data Centers Chip Supply Chain

Parasail to Combine NVIDIA AI Infrastructure with d-Matrix Accelerators to Achieve 10x Faster Token Generation - Morningstar

0.92

www.morningstar.com 2026-07-08

Parasail has announced a collaboration with d-Matrix to combine NVIDIA AI infrastructure with d-Matrix's Corsair accelerators, achieving up to 10x faster token generation. This marks one of the first large-scale commercial deployments of heterogeneous disaggregated inference in production. By pairin

AI Inference GPU Acceleration AI Infrastructure Token Generation Heterogeneous Computing Data Center Optimization Low Latency Inference Semiconductor Technology

Parasail to Combine NVIDIA AI Infrastructure with d-Matrix Accelerators to Achieve 10x Faster Token Generation - PR Newswire

0.92

www.prnewswire.com 2026-07-08

On July 8, 2026, Parasail, an AI inference service provider, and d-Matrix, a pioneer in low-latency AI inference platforms, announced a collaboration to combine d-Matrix's Corsair inference accelerators with NVIDIA Hopper and Blackwell GPUs, achieving up to a 10x increase in token generation speed.

AI Inference GPU Acceleration Token Generation Heterogeneous Computing Data Center Optimization Low Latency Inference NVIDIA Hopper NVIDIA Blackwell

Parasail to Combine NVIDIA AI Infrastructure with d-Matrix Accelerators to Achieve 10x Faster Token Generation - Yahoo Finance UK

0.92

uk.finance.yahoo.com 2026-07-08

Parasail has partnered with d-Matrix to combine NVIDIA's AI infrastructure with d-Matrix's Corsair accelerators, achieving up to 10x faster inference performance. This integration targets token generation and aims to enhance both speed and cost-efficiency in AI inference services. By pairing NVIDIA

AI Inference GPU Acceleration Heterogeneous Computing NVIDIA Hopper NVIDIA Blackwell d-Matrix Corsair Token Generation Cloud Computing

China’s Huawei to enter South Korean AI chip market with new Atlas SuperPods, clusters pack 8,192 Ascend 950 accelerators per deployment — reportedly challenges Nvidia dominance with 'tripled inference performance' of H20 at one-quarter the cost - Tom's Hardware

0.95

news.google.com 2026-07-06

China’s Huawei to enter South Korean AI chip market with new Atlas SuperPods, clusters pack 8,192 Ascend 950 accelerators per deployment

0.92

tomshardware.com 2026-07-06

Huawei plans to enter South Korea's AI chip market in Q4 2026 with its Ascend 950 processors and Atlas 950 SuperPod platform, marking a major move into one of NVIDIA's strongest overseas AI markets. Huawei aims to compete on price and performance, offering a cost-effective alternative to NVIDIA's H2

Huawei South Korea AI chip Ascend 950 NVIDIA Artificial Intelligence Semiconductor Industry Chip Competition

Neurotechnology Integrates NVIDIA Blackwell Accelerators Into AI Infrastructure - Security Today

0.92

securitytoday.com 2026-07-02

Neurotechnology, a biometrics and deep learning company, has integrated NVIDIA Blackwell Ultra B300 accelerators into its existing infrastructure to enhance large-scale data processing and model training capabilities. This move strengthens the company's AI research and enterprise applications, suppo

Neurotechnology NVIDIA AI Infrastructure Biometrics GPU Upgrade Cloud Computing Deep Learning Artificial Intelligence Research

NVIDIA, Australia's Firmus to jointly deploy 170,000 AI accelerators in Indonesia, targeting $30B in revenue - TNGlobal

0.92

technode.global 2026-06-29

NVIDIA and Australian AI firm Firmus Technologies have announced a strategic partnership to deploy up to 170,000 AI accelerators in Indonesia’s Batam campus, aiming for $25–$30 billion in revenue over the first six years. The collaboration leverages Firmus’s HyperCube liquid-cooled AI architecture a

AI accelerators NVIDIA Firmus Technologies Indonesia AI infrastructure cloud computing AI platform integration data center digital infrastructure

Qualcomm plans China-specific data center chips — new Dragonfly lineup will include nerfed AI accelerators that comply with export thresholds

0.87

tomshardware.com 2026-06-25

Qualcomm has announced plans to introduce its full Dragonfly data center product lineup to China, including custom AI accelerators designed to comply with U.S. export thresholds. This move aims to navigate China's unique regulatory environment while leveraging Qualcomm's existing relationships with

Qualcomm Data Center Chips AI Accelerators Export Controls China Semiconductor Market Dragonfly Lineup HBC Near-Memory Design NVIDIA

Qualcomm reveals HBC near-memory AI architecture, AI250 and AI350 accelerators — touts 6x higher bandwidth-per-watt compared to HBM, 200x capacity compared to on-chip SRAM - Tom's Hardware

0.92

www.tomshardware.com 2026-06-25

Qualcomm has unveiled its High-Bandwidth Compute (HBC) near-memory AI architecture, designed to break the memory wall that limits performance in many AI workloads. By decoupling the AI accelerator from the SoC and placing it beneath the LPDDR DRAM stack, Qualcomm claims HBC delivers 6x higher bandwi

Qualcomm Near-memory computing AI accelerator HBC architecture LPDDR Memory wall Bandwidth-per-watt Data center

Qualcomm reveals HBC near-memory AI architecture, AI250 and AI350 accelerators

0.92

tomshardware.com 2026-06-25

Qualcomm unveiled its High-Bandwidth Compute (HBC) near-memory computing architecture aimed at overcoming the 'memory wall' that limits many AI workloads. By decoupling the AI accelerator from the SoC and placing it beneath the LPDDR DRAM stack, Qualcomm achieves high bandwidth and capacity without

Qualcomm Near-memory computing AI accelerator Memory wall LPDDR HBM Data center Artificial intelligence

Semiconductor News & Analysis Feed