
TSMC's Greener AI Chips, Huawei's 910C, and Google's Gemma 3 QAT - Silicon Brief #15
April 28, 2025 -
Silicon Brief
Hi there,
In this edition of Silicon Brief, we're diving into the latest developments in AI chips and data centers. TSMC has introduced faster and more environmentally friendly AI chips, while Huawei is preparing to ship its new AI chip, the 910C, despite restrictions from Nvidia. Meanwhile, Google has released its Gemma 3 QAT models for consumer GPUs, and Baidu has unveiled its Kunlun Chip Cluster designed for AI model training.
On the data center front, Microsoft and Du are collaborating on a $544 million data center in the UAE, and iGenius has opened the Colosseum AI Data Center with partners Vertiv and NVIDIA. Additionally, Amazon and Nvidia are experiencing strong demand for AI data centers, while Adani Group plans to invest $10 billion in Indian data centers. Stay tuned as we explore these stories and more in the world of AI and data infrastructure.
Energy & Sustainability
Ocient has secured $42.1 million in Series B funding to advance energy-efficient solutions for operational data and AI workloads, targeting data-center power costs. A report by BloombergNEF predicts that AI will double data centre energy demand by 2050, with centres accounting for nearly 9% of global energy demand. AI Green Bytes is set to open an immersion-cooled data center in Paris by June 2025, utilizing Oleon's biodegradable cooling fluid to reduce energy consumption. A study from Georgetown, Epoch AI, and Rand forecasts AI data center costs could hit $200 billion by 2030, stressing potential energy and environmental challenges.
- Ocient Secures $42.1 Million for Energy-Efficient AI Solutions
- NTT Global Data Centers' AI Cooling for Net-Zero
- AI and Data Centre Energy Demand to Double by 2050
- AI Green Bytes to Open Paris Data Center with Plant-Based Cooling
- AI Data Center Costs May Reach $200 Billion by 2030
Technology
Airedale by Modine has developed a Cooling AI Control System to optimize cooling in high-density data centers, potentially reducing energy consumption by up to 40% through AI-enhanced technology. On April 24, 2025, Dover reported that its first-quarter profits exceeded estimates, driven by a 27.4% increase in sales from its data center cooling components segment. Despite this, the company adjusted its full-year profit forecast downward. AWS has introduced the Model Context Protocol (MCP) Servers to enhance AI-assisted coding tools, aiming to improve development speed, code quality, and security by integrating with AWS services. Microsoft plans to implement LG Electronics' cooling solutions in several upcoming AI data centers to enhance thermal management and efficiency.
- Airedale by Modine AI Cooling System Saves Energy
- Dover Exceeds Q1 Profit Estimates Due to Data Center Demand
- AWS Launches MCP Servers for AI Cloud Development
- Microsoft Partners with LG for AI Data Center Cooling
Regulation & Policy
Intel has informed its Chinese clients that certain advanced AI chips will now need an export license due to new US restrictions. This follows Nvidia's announcement of a financial charge related to similar export limitations.
Austin City Council is considering a resolution to enhance AI oversight, focusing on workforce protections, transparency, and environmental impacts of data centers.
TSMC has refuted claims of sending advanced AI processors to Huawei, affirming adherence to US restrictions since 2020.
Austin City Council is considering a resolution to enhance AI oversight, focusing on workforce protections, transparency, and environmental impacts of data centers.
TSMC has refuted claims of sending advanced AI processors to Huawei, affirming adherence to US restrictions since 2020.
- Intel Requires Export License for AI Chips to China
- Austin City Council Proposes AI Policy Expansion
- TSMC Denies Shipping AI Chips to Huawei
Data Centers
Microsoft and Du are collaborating to build a $544.54 million hyperscale data center in the UAE to enhance the region's digital infrastructure. iGenius has opened the Colosseum AI data center in Italy with Vertiv and NVIDIA, featuring a large NVIDIA DGX AI supercomputer. Stonepeak launched Montera Infrastructure to manage hyperscale data centers in North America with $1.5 billion in equity backing.
- iGenius Launches Colosseum AI Data Center with Vertiv and NVIDIA
- Stonepeak Launches Montera Infrastructure for Data Centers
- Galaxy Digital and CoreWeave Expand AI Infrastructure
- Vertiv Holdings Boosts Sales Forecast Amid AI Data Center Demand
- Oracle Reports 11.59 Billion Trading Volume Amid AI Investments
- Broadcom Invests in Ethernet for AI and Data Centers
- Amazon and Nvidia See Strong AI Data Center Demand
- Adani Group Plans $10B Data Center Expansion in India
- NTT DATA and Neysa to Invest in AI Data Centre in Hyderabad
- India Assigns AI Projects to Yotta, E2E, and NxtGen
- Corscale Partners with Affinius for Hyperscale Data Centers
- Yandex to Build $193 Million Data Center in Russia
- Lenovo Launches Major Data Storage Update for AI
- La Fe Health Research Institute Building Supercomputer in Valencia
- LG CNS Reports Record Profits Driven by Cloud and AI Growth
- L&T Cloudfiniti to Triple Data Center Capacity by 2026
- Khazna Data Centers Expands with New AI Facilities in UAE
- Keppel Secures $1.53 Billion for Data Center Expansion
- Exowatt Secures $70 Million for AI Data Center Solar Solutions
- Microsoft and Du to Build Hyperscale Data Center in UAE
Chips & Hardware
Ansys and TSMC have enhanced their collaboration to optimize semiconductor solutions for AI and high-performance computing. Aion Silicon has rebranded to focus on custom chip solutions for AI, automotive, and 5G applications. Vecow launched the ECX-4000 series with Intel Core Ultra 200S processors for AI applications in smart cities and in-vehicle computing.
- Ansys and TSMC Enhance Collaboration on Semiconductor Solutions
- Aion Silicon Rebrands to Focus on Custom Chip Solutions
- Vecow Launches ECX-4000 AI System with Intel Core Ultra 200S
- TSMC Introduces Faster, Greener AI Chips
- Nextchip Licenses Ceva's Neural Processing Unit for ADAS
- Huawei to Ship New AI Chip 910C Amid Nvidia Restrictions
- Aion Silicon Joins Intel Foundry Alliance
- GigaIO Launches Portable AI Supercomputer Gryf
- Huawei Launches CloudMatrix 384 Supernode
- Google Releases Gemma 3 QAT Models for Consumer GPUs
- Atum Works Develops Nanoscale 3D Printer for Chip Production
- Baidu Unveils Kunlun Chip Cluster for AI Model Training
Subscribe to Silicon Brief
Weekly coverage of AI hardware developments including chips, GPUs, cloud platforms, and data center technology.