Loading story…
Huawei is advancing its domestic semiconductor ecosystem with the Ascend AI chip family, positioning its CloudMatrix 384 system as a direct alternative to Nvidia's GB200 platform. The Ascend 910C, a dual-chiplet accelerator with stacked HBM2E memory and DaVinci NPU architecture, is at the core of this effort, delivering up to 780 TFLOPS of dense BF16 compute. While the chip lags behind Nvidia's Hopper and Blackwell architectures in peak throughput and power efficiency, Huawei is compensating by scaling up deployments and building a robust supplier network, with over 60 semiconductor companies backed by its Hubble investment arm. This initiative is part of China's broader push for semiconductor independence, with local partners developing design toolchains and software ecosystems to reduce reliance on U.S. vendors.
Meanwhile, AMD is preparing for the next generation of AI and HPC workloads by upstreaming Linux kernel patches for its upcoming Instinct MI400-series accelerators, likely based on the CDNA5 architecture. These patches indicate hardware enablement for new GPU IP blocks, including platform security and memory controllers, and suggest that AMD is moving forward with products intended to compete in the high-performance AI accelerator market. The developments from both Huawei and AMD highlight the intensifying global competition in AI hardware, with each company pursuing different strategies to address performance, efficiency, and supply chain independence.
Track how attackers are adapting to this technology.
3 events from the most recent confirmed update back to the earliest known activity.
AMD began upstreaming Linux kernel patches enabling new GPU IP blocks, including components such as PSP 15.0.8, IH 7.1, MMHUB 4.2, GFXHUB 12.1, and GMC 12.1. The changes are believed to support the company's next-generation CDNA5-based Instinct MI400-series accelerators for AI and HPC workloads.
Huawei developed the CloudMatrix 384 AI server system, integrating 384 Ascend 910C processors as a domestic alternative to Nvidia's high-end AI platforms. The system reportedly compensates for lower per-chip performance and efficiency through large-scale horizontal deployment.
Huawei expanded a China-based semiconductor ecosystem around its Ascend processors, backed by investments in more than 60 chip-related firms and partnerships spanning design tools, materials, and manufacturing equipment. The effort is part of China's broader push for semiconductor self-reliance amid tightening foreign technology restrictions.
Vulnerabilities, threat actors, malware, products, organizations, and breaches Mallory has linked to this story.
Follow how adversaries are adapting to this technology, and where it touches your stack today.
2 references tracked. Mallory keeps watching after this page renders.
tomshardware.com
Open sourcetomshardware.com
Open sourceMap indicators from this story to your assets and identify affected systems in minutes.
Every observed campaign, victim, and pivot linked to actors named in this story.
Malware, exploits, and IOCs connected to the activity described here.
YARA, Sigma, and Snort rules deployed to your SIEM as soon as they’re published.
Get matching new stories delivered to your team as they break — not the next morning.
Ask questions about this story and take action on the answers.
Other stories Mallory is tracking that overlap on entities, sources, or topic labels.
Nvidia’s AI hardware business faced continued uncertainty in China as CEO **Jensen Huang** said Beijing had not yet approved imports of the **H200** GPU and confirmed that **no new H200 orders** were being placed while regulators decide. In parallel, Chinese GPU vendor **Shanghai Iluvatar CoreX** publicized a four-generation roadmap positioning its architectures (including *Tianshu* and successors) as domestic alternatives intended to compete with and ultimately surpass Nvidia’s Hopper/Blackwell and target Nvidia’s future **Rubin** platform by 2027, underscoring China’s push to reduce reliance on Western AI silicon. Outside the China licensing issue, Nvidia-related coverage also highlighted expansion and productization of AI compute: a report claimed Nvidia invested **$2B** in **CoreWeave** to support buildout of a planned **5GW “AI factory”** ecosystem by 2030 using multiple Nvidia platform generations (including **Rubin**, *Vera* CPUs, and *BlueField*). Separately, a review of *Nvidia DGX Spark* described a local AI development system built around the **GB10 Superchip**, reflecting growing demand for on-prem/local inference and development hardware as model sizes and VRAM requirements increase. Intel also outlined a separate, non-Nvidia event: a planned “hybrid” AI processor concept combining **x86 CPUs**, a fixed-function AI accelerator, and programmable logic for inference-focused workloads.
Jun 29, 2026Huawei and Cambricon are increasing production of AI accelerators at Chinese fabs, aiming for over a million units in 2026. However, production is constrained by limited advanced semiconductor fab capacity and a looming shortage of high-bandwidth memory (HBM). These bottlenecks threaten China's ambitions for AI hardware self-sufficiency.
Jun 29, 2026
AI infrastructure demand is reshaping leading-edge semiconductor manufacturing economics and capacity planning. Reporting citing analysts and executive comments says **Nvidia** is on track to become **TSMC’s largest revenue customer**, driven by the high wafer and advanced packaging intensity of large AI accelerators versus smaller mobile/PC SoCs; one estimate projects Nvidia generating roughly **$33B** of TSMC revenue (about **22%** of total) versus Apple at **$27B** (about **18%**). The same demand dynamic is also stressing upstream constraints (power, wafers, and advanced packaging), with industry commentary pointing to rapid AI data-center buildouts and very large accelerator deployments as a key driver of the shift. In parallel, executives and reporting describe a broader push to expand manufacturing geographically as *capacity growth* rather than a wholesale relocation of Taiwan output. Nvidia CEO **Jensen Huang** publicly rejected claims that the U.S. aims to move 40% of Taiwan’s chipmaking to America, arguing new fabs in the U.S./Europe/Japan are needed because AI-driven wafer demand is outpacing what Taiwan’s infrastructure can support, while Taiwan remains central to the ecosystem; he also highlighted tight supply for **HBM and other memory** (HBM, DDR5, GDDR7, LPDDR5X, NAND) and coordination with **Samsung, SK hynix, and Micron** for next-gen accelerators (e.g., *Rubin*). Separately, **Samsung Foundry’s Taylor, Texas** site is reported to be approaching EUV “first light” and planning **2026 risk production**, including adoption of **EUV pellicles** and a large-scale ramp target, while **Intel** outlined a “hybrid” AI processor concept combining **x86**, a fixed-function AI accelerator, and programmable logic—signaling intensifying competition for AI inference workloads and continued pressure on advanced-node and packaging supply chains.
Jun 29, 2026