Gigawatt campuses, power grids, cooling and the trillions being poured into AI compute buildouts.
516 articles
Yunbao Smart, a DPU (Data Processing Unit) startup founded by a 24-year-old Stanford talent, is preparing for an IPO on the ChiNext board. Tencent holds a significant stake, highlighting the strategic importance of DPU infrastructure in the AI era.
钛媒体 · 77d ago
Extreme heat waves and the explosive growth of AI data centers are pushing US power grids to their limits. Major grid operators like PJM are facing record-breaking demand, leading to emergency measures and warnings of potential future power shortages.
虎嗅 · 78d ago
Meta developed a custom CXL chip called Vistara to integrate decommissioned RAM into new server architectures. This approach mitigates memory shortages and rising hardware costs without compromising performance.
Computerworld · 78d ago
Meta is shifting its strategic focus from model-level development to infrastructure. This move reflects the broader industry trend where AI value is migrating toward the underlying compute and data center stack.
钛媒体 · 78d ago
Meta has developed a technical workaround to use legacy DDR4 memory in AI servers originally designed exclusively for DDR5. This innovation helps mitigate severe global DDR5 shortages affecting data center expansion.
cnBeta (Full RSS) · 78d ago
Infineon has officially opened its Smart Power Fab in Dresden, Germany, representing a 5 billion euro investment. The facility aims to boost capacity for AI data center power supplies, automotive software, and renewable energy.
IT之家 · 78d ago
Kioxia has begun shipping samples of its 10th generation 332-layer BiCS FLASH 3D NAND, featuring 1Tb TLC capacity. These chips are designed to meet the high-performance, high-capacity, and low-power demands of AI-driven data centers.
cnBeta (Full RSS) · 78d ago
Kioxia has begun shipping samples of its next-generation flash memory chips specifically designed for AI data center operators. The company aims to increase its market share against competitors in the high-performance storage sector.
Bloomberg Technology · 78d ago
Valar Atomics successfully powered an Nvidia Blackwell AI chip using its Ward 250 advanced nuclear reactor. This milestone marks the first time a next-generation reactor has directly supported AI hardware in the United States.
Bloomberg Technology · 79d ago
Unitree Robotics received approval for its IPO on the STAR Market. Meanwhile, Douyin launched an AI portrait protection feature to combat unauthorized AI-generated content and impersonation.
36氪 · 79d ago
Elon Musk's SpaceX has acquired Mesh Optical Technologies, a startup founded by former Starlink engineers, to secure proprietary optical interconnect technology. The acquisition aims to integrate Mesh's 1.6Tbps Alpha C1 transceivers into SpaceX's Starlink and the upcoming 'Starmind' space-based AI computing network.
雷峰网 · 79d ago
Meta has introduced a custom chip called 'Vistara' that allows legacy DDR4 memory to be used in modern DDR5-based server architectures. This initiative aims to reduce hardware costs and optimize resource utilization amidst global memory shortages.
cnBeta (Full RSS) · 81d ago
ByteDance is accelerating the development of a proprietary CPU to power its AI infrastructure, with design finalization expected by early next year. The company aims for mass production and deployment by the second half of 2027 to reduce reliance on external hardware.
SCMP Technology · 82d ago
South Korea has announced a major initiative to maintain its global leadership in AI by prioritizing large-scale investments in memory chips, data centers, and robotics. Industry giants Samsung Electronics and SK Hynix are central to this national strategy.
Bloomberg Technology · 82d ago
Australian firm Firmus Technologies is partnering with Nvidia to construct a 360-megawatt AI data center in Batam, Indonesia. The project aims to capture significant market demand, with the company projecting $30 billion in future offtake deals.
The Next Web (TNW) · 83d ago
MediaTek is entering the AI data center optical chip market using CPO (Co-Packaged Optics) and Micro LED technology. This move aims to break the current duopoly held by Broadcom and Marvell in high-speed data transmission.
Pandaily · 84d ago
Elon Musk has received FTC approval to acquire Mesh Optical Technologies, a startup specializing in high-speed optical transceivers for AI data centers. This move aims to overcome bandwidth and heat limitations in GPU-to-server communication.
IT之家 · 84d ago
Qualcomm has secured Meta as the first major customer for its Dragonfly C1000 data center processor. This move signals Qualcomm's aggressive expansion into the AI infrastructure market, challenging established incumbents.
The Next Web (TNW) · 87d ago
Cerebras shares dropped following an annual sales forecast that fell short of market expectations. CEO Andrew Feldman identified capacity constraints as the primary bottleneck for the company's growth in the AI data center market.
Bloomberg Technology · 87d ago
Blackstone plans to invest $30 billion over the next three to five years to develop AI data centers in Japan. The project aims for a total capacity exceeding one gigawatt, signaling strong confidence in AI infrastructure demand.
The Next Web (TNW) · 87d ago
Reflection AI has signed a massive $6.3 billion deal to lease Nvidia chip capacity from SpaceX's Colossus 2 supercomputer cluster. This agreement highlights the growing demand for large-scale GPU infrastructure among AI startups.
The Next Web (TNW) · 88d ago
Tesla is expanding into the AI infrastructure market by offering modular 'Compute Blocks'. This move leverages their internal data center expertise for external commercial use.
量子位 · 88d ago
SpaceX has entered a significant compute partnership with open-source lab Reflection AI. Starting July 2026, Reflection AI will pay $150 million monthly for access to Nvidia's GB300 chips at the Colossus 2 data center.
TechCrunch AI · 89d ago
Chevron and Microsoft have signed a 20-year agreement to power the 'Project Kilby' data center in West Texas. The facility will require approximately 2.7 gigawatts of power, highlighting the massive energy demands of modern AI infrastructure.
cnBeta (Full RSS) · 89d ago
Groq has secured $650 million in new funding to expand its data center capacity. This capital injection supports the company's strategic pivot toward becoming a major provider of AI computing services.
Bloomberg Technology · 89d ago
Microsoft has entered a 20-year contract with Chevron to power a massive data center in West Texas using natural gas. This move marks a departure from the company's previous reliance on renewable and nuclear energy for AI infrastructure.
The Next Web (TNW) · 89d ago
Harvard spinout HyperLight has raised $80 million to address the data bottleneck in GPU clusters. They are developing optical interconnects to replace copper links in large-scale AI hardware.
The Next Web (TNW) · 92d ago
A transmission fault in Ashburn, Virginia, reportedly removed more than 3 gigawatts of data-center load from the grid in seconds. The incident follows an earlier failure that affected about 60 facilities, highlighting the fragility of concentrated AI infrastructure.
MIT Technology Review · 9d ago
Arm has launched Neoverse CSS N4, a server CPU platform designed to address rising compute demand in the agentic AI era. Each die can support up to 128 cores.
InfoQ中国 · 11d ago
Broadcom reported fiscal Q3 2026 revenue of $29.591 billion, up 86% year over year, while net income rose 216% to $13.088 billion. AI semiconductor revenue reached $16.7 billion, and the company expects $21.7 billion in AI semiconductor revenue in Q4.
IT之家 · 17d ago