Pinterest Stock Falls as CFO Julia Brau Donnelly Resigns 
Alibaba Raises $10.2 Billion to Expand AI Infrastructure 
Mercer International Stock Surges 45% on C$20 Million Canada Funding 
Meta Agrees to $18 Billion Settlement, Tightens Teen Social Media Rules 
Nvidia Pauses AI Cloud Financing Deals Amid Antitrust Concerns 
Nvidia Reportedly Eyes $13 Billion Hugging Face Acquisition 
Nvidia Revenue Soars 106% as AI Demand Hits ‘Inflection Point’ 
Trian Drops Wendy’s Take-Private Bid Plans, Shares Plunge 14% 
SoftBank Plans Record $6.3 Billion Bond Sale to Fund OpenAI Investments 
BNP Paribas, KB Kookmin Eye $2 Billion Techcombank Stake 
Itochu Offers $1.56 Billion to Privatize Dentsu Soken 
BofA Names ASML Top Semiconductor Equipment Pick on Growth, Valuation 
Google Changes EU Spam Policy Amid Antitrust Scrutiny 
Samsung Shares Tumble 8% as $79 Billion Return Plan Disappoints 
Trump Administration Plans New Drug Pricing Deals With Biotech Firms 
Luxshare Shares Rise as First-Half Profit Jumps 18% 
Add as a preferred source on Google
The computing chips driving artificial intelligence demand immense amounts of electricity, posing a growing challenge for data centers worldwide. On Wednesday, Taiwan Semiconductor Manufacturing Co. (TSMC), the world’s largest chip manufacturer and primary fabricator for Nvidia, revealed a groundbreaking strategy to improve chip energy efficiency by nearly tenfold—using AI-powered software to design them.
Today’s top AI servers, such as those running Nvidia’s flagship chips, can consume up to 1,200 watts during heavy workloads. Scaled across thousands of units, that level of energy draw is equivalent to powering 1,000 U.S. homes continuously. To combat this, TSMC is turning to advanced design methods and cutting-edge software tools developed in collaboration with Cadence Design Systems and Synopsys.
A key innovation lies in packaging multiple “chiplets”—smaller computing units made with different technologies—into a single system, creating far more efficient chip architectures. However, managing the complexity of such designs requires next-generation software. AI-driven design platforms are proving faster and more effective than human engineers for certain tasks, enabling chipmakers to unlock the full potential of TSMC’s 3D integration technology.
Jim Chang, deputy director of TSMC’s 3DIC Methodology Group, emphasized that AI tools are already outperforming engineers in some areas. “This thing runs five minutes while our designer needs to work for two days,” he explained, highlighting the speed and optimization advantages AI brings to chip development.
Yet, challenges remain. Moving data between chips using traditional electrical connections is hitting physical limits. Experts like Kaushik Veeraraghavan from Meta stress that optical connections may become the future for large-scale AI data centers. These advances will be critical as the world seeks more sustainable solutions to power the accelerating demand for AI computing.
By combining AI-driven design with innovative chip architectures, TSMC and its partners aim to set new standards for efficiency, performance, and scalability in the semiconductor industry.
