Video explains China's AI chip strategy with DeepSeek, Huawei, Nvidia
An Economic Times video breaks down how China is structuring its AI chip strategy, covering the roles of DeepSeek, Huawei and Nvidia in the domestic compute push.
Archive, page 2
An Economic Times video breaks down how China is structuring its AI chip strategy, covering the roles of DeepSeek, Huawei and Nvidia in the domestic compute push.
Chinese AI chipmaker Iluvatar CoreX debuted the Tiangai 300 IXL144, a supernode system that fits 144 GPUs into a single cabinet, unveiled at the 2026 China International event.
Reddit post says Sipeed announced MaixStation, positioning it as comparable to Strix Halo and DGX Spark; specs and pricing are not given.
TCL displayed a flagship 5G router with a built-in AI agent for network optimization, Telecompaper reports.
Alice & Bob has decoded LDPC-cat codes on GPUs using NVIDIA's NVQLink, a step toward practical quantum error correction with AI hardware.
NVIDIA's blog announces it is bringing its GTC conference to Berlin, Europe, as part of its global AI developer events.
TweakTown reports an enthusiast reached 2,010MHz on an NVIDIA Quadro RTX 4000 using a $5 liquid cooler, an extreme overclocking result.
Nvidia, AMD, Google and Meta set a new Data Center Spec v0.7, a shared specification for AI datacenter hardware. Details reported by tech-insider.org.
Nvidia celebrates the 20th anniversary of the GeForce 8800 GTX, its first CUDA-capable GPU, which launched the CUDA compute ecosystem now central to AI.
Semiconductor shares fell, dragging Korea's Kospi lower as Samsung and SK Hynix declined, after Broadcom's chief warned of a 'casino' mentality and investors fretted over the scale of AI investment.
TrendForce reports Microsoft's Maia 300 will fuel HBM4 demand while Intel bets on LPDDR5X memory, as AI chipmakers diverge on memory strategies.
Anthropic CEO Dario Amodei warned of a 'huge shortage' of AI chips and has reached out to Elon Musk about compute capacity, Firstpost reported.
Nvidia-backed data center firm Firmus scrapped its Australian IPO citing market conditions and is now seeking up to $3 billion, per reports.
China's CXMT announced progress on a new 4F DRAM memory-cell architecture and plans DDR5 server memory using it by end-2026, per an X post.
Nvidia's share of China's AI chip market has fallen to about 34% as Huawei closes the gap, Reuters reports, underscoring how export restrictions and local alternatives are reshaping the market.
Anthropic is in final talks with NTT Data Group to build its first data center in Japan, Nikkei reported; the facility could draw up to 500 megawatts, making it among Japan's largest.
China Mobile tripled the capacity of its largest intelligent computing hub in Inner Mongolia, expanding its data center amid Beijing's push to boost AI digital infrastructure.
Cisco and Nvidia developed a new pluggable optical module for AI data centers, targeting high-speed interconnects with four-times USFP density so systems wait less on data.
The Financial Times reports China faces an uphill battle getting its tech companies to stop using Nvidia chips, despite local alternatives and policy pressure.
Chinese tech companies purchased more than a million Nvidia gaming chips in a single year, repurposing hardware never designed for advanced AI to meet soaring compute demand.
Intel CEO Lip-Bu Tan and Broadcom CEO Hock Tan said at a Singapore financial conference that access to capital may become a challenge for the AI sector.
An Engadget how-to on getting higher frame rates from an older GPU by adjusting settings and forgoing newer features. A general PC tuning guide, not AI news.
Nvidia is joining other tech firms in an effort to standardize data center development, according to CoStar. Specific standards or partners were not detailed in the source.
Nvidia will no longer ship new Windows 10 drivers for RTX GPUs, per reports, pushing users toward Windows 11 for continued driver support.
Microsoft's AI chief commented on whether the custom Maia 300 chip reduces dependence on Nvidia, saying competition is good, in a Yahoo Finance interview.
Intel's D-series data center GPU, codenamed Crescent Island, is named D270 and supports up to 480 GB of LPDDR5X memory, positioning it as an Nvidia rival for AI workloads.
As agentic workloads multiply, cloud token bills are outpacing budgets, prompting enterprises to rethink where inference runs. Dell and Intel weigh deployment strategy trade-offs.
Nvidia and AMD shares declined as a wave of new AI announcements failed to re-energize investors, MarketWatch reported.
Navitas shares rose about 5% after the power-chip maker announced an AI infrastructure partnership with Nvidia.
Applied Materials announced a $5 billion EPIC center aimed at speeding up AI chip research and development.
Nvidia and Cisco announced USFP optics targeting 400Gb/s per lane, alongside MGX 1.2 and DSX OS updates at OCP 2026.
Tech Xplore reports Microsoft is betting on its in-house AI chip as an alternative to Nvidia hardware, part of its datacenter self-sufficiency push.
ASML's CEO said the AI chip boom could extend beyond 2028, while Broadcom warned of a financing bottleneck in AI infrastructure buildout.
Ambarella introduced a standalone AI accelerator aimed at physical AI designs, targeting robotics and edge perception workloads.
Intel's Crescent Island inference accelerator comes to market as the GPU D-Series, offering 1.3 peak PFLOPS of FP8 compute and up to 480 GB LPDDR5X at 1.5 TB/s bandwidth.
Yahoo Finance reports Microsoft reclaimed a $4 trillion market cap, crediting its Maia 300 AI chip for the gains.
Intel branded its Crescent Island data center AI inference cards as the GPU D-Series; the D270 delivers 1.3 PFLOPS dense FP8 and 1.5TB/s memory bandwidth.
Benzinga reports a private AI chip startup has outperformed Nvidia's stock gains, without naming valuation or funding details.
An analysis piece compares what Nvidia offers versus Marvell in the AI chip market, examining differentiation across accelerators and networking.
Yahoo Finance reports Arista shares fell 2.7% amid concerns Ethernet networking is invading Nvidia's rack designs, pressuring its positioning.