Global Scans · Hardware · Weekly Summary

• [New] Amazon's M3 Ultra Mac Studios only made it into two regions - US East and US West (Oregon) - so users elsewhere who fancy a cloudy Mac but need lower latency will have to endure the very on-prem experience of waiting for hardware to show up. theregister
• [New] As edge AI hardware and software ecosystem matures, on-device intelligence platform demand is expected to grow significantly across latency-sensitive and regulated sectors in Mexico. Mobility Foresights
• [New] Training only one single round of GPT-4 at a reported MFU of 35% requires the computational capacity of approximately 2515 A100 GPUs under the most plausible baseline scenario of a 2-year hardware lifespan. Nature
• [New] Training GPT-4 with an MFU of 20% and a hardware lifespan of 1 year requires 8800 A100 GPUs, decreasing to approximately 2934 GPUs with a hardware lifespan of 3 years. Nature
• [New] In a note to clients this week, analysts at Goldman Sachs flagged the inflationary potential of AI, pointing to the spike in hardware prices, software firms upcharging for AI tools, and the surge in power demand from data centers, which is feeding energy inflation. Business Insider
• [New] As AR/VR hardware becomes mainstream, Render's ability to provide real-time, AI-generated 3D environments at a 40% discount will be its primary growth driver through the remainder of 2026. KuCoin
• [New] Researchers at the University of Cambridge have developed a new brain-inspired nanoscale device that they say could dramatically reduce the enormous energy demands of artificial intelligence hardware. New Atlas
• [New] On the demand side, for enterprises buying AI hardware, expect GPU and custom chip pricing to remain high through 2026. XploraHorizons.net
• [New] Future Outlook and Strategic Implications Quantum AI evolution toward 2030 suggests expanding capabilities and broader enterprise adoption as hardware improvements address current limitations. BUSINESS 2.0 NEWS
• [New] Investors are paying closer attention to Google's next-generation TPUs as a real competitive threat to Nvidia, even as analysts still argue Nvidia retains a strong lead because of its hardware roadmap and CUDA software ecosystem. Tech Startups - Tech News, Tech Trends & Startup Fundin
• [New] TSMC unveiled its A13 and N2U process nodes for 2029 production and announced a new Arizona packaging facility, while cautioning that high demand and hardware costs could prolong chip shortages beyond 2027. Distill
• [New] On the hardware side, AMD is adding HDMI 2.1 support for Linux, which could enhance experiences for gaming and AI workloads, especially for devices like the Steam Machine source. Coaio Limited
• [New] NASA plans to repurpose Gateway hardware already built, including the Power and Propulsion Element built by Maxar and the HALO module by Northrop Grumman. / China ISDC 2026 | International Space Development Conference
• [New] Nvidia's approach is based on the idea that machine learning systems trained to predict errors, optimize performance, and control systems can actively manage and stabilize quantum machines, rather than relying solely on hardware improvements. Campus Technology
• [New] The coming iPhone 18 Pro line could bring some of the biggest camera hardware upgrades Apple has made in years, paired with new iOS 27 features centered on visual intelligence and a new Siri mode inside the Camera app. Mean CEO's BLOG
• [New] We forecast iPhone growth above 20% for the year, driven by what Apple does best: high-quality hardware (camera, display, processor), new form factors, and an ever-compelling software ecosystem. Morningstar, Inc.
• [New] As XR supply chains mature and key components remain IP-intensive, hardware differentiation will slow, leaving software, services, and onboard AI as the primary competitive levers. Next Reality
• [New] The strategic signal matters: if one of the world's top AI labs can achieve performance parity on non-NVIDIA hardware, it validates the thesis that AI compute will eventually become commoditized across multiple hardware platforms, not just NVIDIA's. o-mega
• Smartphone hardware in 2026 will be optimized for AI tasks, with advanced processors and memory enabling efficient AI computations. VERTU

Last updated: 19 May 2026