NowRift — Software & Technology News Platform

HardwareRIFT LEAD

Sapphire Surface Modification Yields Wafer-Scale 2D Semiconductor Crystals

Peking University researchers developed a method to grow high-quality 2D semiconductor crystals on sapphire substrates, paving the way for large-scale production.

28 Aug 2026

Newsletter

THE DAILY RIFT

Five shifts. Five minutes. No noise.

•Verified technology intelligence
•Emerging patterns before the hype
•Direct action steps for builders & founders

No spam. Unsubscribe anytime. Powered by Beehiiv.

All Intelligence AI Models Developers Cybersecurity Hardware Founders Creators Data Science

Today's Intelligence

HardwareAug 28

UM-Dearborn Generative Design Optimizes Liquid Cooling for 2.5D/3D Packages

New physics-guided framework creates efficient liquid cooling channels for high-density chip packages, addressing critical thermal challenges.

●Builder Action:Developers working on high-performance computing, AI accelerators, or advanced graphics should be aware that traditional thermal solutions are reaching their limits. This research demonstrates how generative design, guided by physics simulations, can create highly efficient, custom liquid cooling channels for 2.5D and 3D packages. Expect to see more complex, organic cooling channel designs integrated into future high-density processors, potentially enabling higher performance ceilings and improved thermal stability.

HardwareAug 28

Ferroelectric Tuning Solves LLM Training Bottlenecks, Georgia Tech Finds

New research demonstrates a 2.7x speedup in LLM training by mitigating communication stalls with ferroelectric tuning.

●Builder Action:Developers training large MoE LLMs on wafer-scale systems can expect significantly faster training times. The Georgia Tech research validates a technique that reduces communication stalls by 2.7x, allowing more compute time. This means faster model iteration and potentially lower cloud compute costs for intensive training tasks.

HardwareAug 28

AI Hardware Race Heats Up: OpenAI, Nvidia, and Alibaba Make Key Moves

OpenAI's custom chip, Nvidia's potential Hugging Face acquisition, and Alibaba's new model signal shifts in AI cost and control.

●Builder Action:Developers face a landscape where custom inference hardware (like OpenAI's Jalapeño) promises lower latency and higher throughput, potentially impacting benchmarks. Nvidia's potential control over Hugging Face could influence the availability and discoverability of open-source models. Alibaba's Qwen3.5-Flash offers a more cost-effective alternative for deployment, encouraging experimentation with lighter models.