Semiconductor Engineering — Posts

64 posts on "Semiconductor Engineering" by Shlomo Strauss — bilingual writing on technology, science, and business.

• The structural flaws driving OpenAI toward collapse — Without its own infrastructure, OpenAI's growing user base deepens its debt, risking a collapse that could trigger a broader tech recession.
• Liquid metal circuit boards could slash electronics prototyping time — Itera's liquid metal technology instantly reconfigures circuit boards, drastically reducing prototyping time while protecting intellectual property.
• NASA and Microchip to build space processors 100 times faster — NASA and Microchip are developing a radiation-resistant space chip 100 times faster than current models, with future uses for critical systems on Earth.
• Google Reimagines the Laptop with Android-Powered GoogleBooks — Google's upcoming GoogleBook laptops will run a redesigned Android OS with native app support and deep Gemini AI integration to challenge Windows.
• The Geopolitical Battle Over LED Manufacturing — The transition to gallium nitride in LED manufacturing has sparked a geopolitical struggle due to its critical military and communication applications.
• How Cerebras Built a Giant AI Chip and Why It Is Going Public — While Cerebras boasts revolutionary wafer-scale AI chips, its upcoming multi-billion dollar IPO faces risks from extreme customer concentration.
• How USB-C Laptop Charging Became Safe Enough for the EU Mandate — A new EU law mandates USB-C charging for laptops, made possible by hardware solutions that prevent the short circuits that once burned motherboards.
• How a Rare Helium Isotope Cools Quantum Computers — Helium-3, a rare isotope sourced from nuclear warheads and the Moon, is essential for cooling quantum computers to near absolute zero.
• Nvidia's $9 Million Server Threatens Traditional Hardware Makers — Nvidia's upcoming NVL144 server highlights the company's push to control the entire AI supply chain, threatening traditional hardware manufacturers.
• How the Global Helium Crisis Threatens Advanced Technology — An Iranian attack on Qatari facilities and the depletion of US reserves have triggered a severe helium shortage, threatening chipmaking and MRI systems.
• The Challenges and Strategic Value of a Far Side Moon Base — Despite communication and terrain challenges, the Moon's far side offers vital resources like ice and helium-3, sparking a new geopolitical race.
• Google's TurboQuant Algorithm Cuts AI Memory Needs by 8x — Google's open-source TurboQuant algorithm uses polar coordinates to compress AI memory requirements by eightfold without losing accuracy.
• How bots tried to corner the RAM market with 10 million crafty — DataDome exposed a bot campaign that scraped RAM product pages over 10 million times, aiming to scalp inventory and deepen an AI-driven memory shortage.
• How Tesla solved the battery and heat problems that held back — From cylindrical cells with interleaved cooling tubes to a structural battery floor, a look at the engineering breakthroughs behind Tesla's thermal and…
• MacBook nano and Windows RT: why Apple succeeded where Microsoft — Apple's $599 MacBook nano vindicates Microsoft's 2012 ARM laptop vision, but succeeds where Windows RT failed thanks to mature chips, Rosetta 2, and tight
• The engineering inside Tesla's motors that most drivers never think — Dual motors, carbon-fiber sleeves, and silicon carbide inverters — a clear look at the engineering solutions Tesla built into its electric drivetrains.
• Why Nokia got Asus and Acer banned from selling PCs in Germany — A Nokia patent lawsuit over HEVC video decoding reveals the licensing chaos behind modern video codecs and why the industry is moving to AV1.
• Intel's ZAM memory aims to pack 512 GB where HBM tops out at 64 — Intel and SoftBank's ZAM technology uses diagonal drilling between memory layers to promise 512 GB per package — eight times the capacity of today's best…
• Tower Semiconductor and Nvidia bet on silicon photonics for AI — A new deal between Tower Semiconductor and Nvidia will bring silicon-nitride optical interconnects inside AI servers, promising up to 70% energy savings…
• How a Japanese toilet maker became a key supplier to the chip — Toto's mastery of ceramic shrinkage, developed for luxury toilets, led to an electrostatic wafer-holding technology now in high demand as AI chip…
• Musk's AI empire: custom chips, space servers, and a closed-loop — From Dojo 3 supercomputers in orbit to xAI merging with SpaceX, the piece maps how Musk is assembling a self-contained AI pipeline that no competitor…
• The hidden damage trail of geomagnetic storms — Beyond the aurora's beauty, geomagnetic storms accelerate pipeline corrosion, collapsed Quebec's power grid in 1989, and destroyed 38 Starlink satellites…
• Why Nvidia's $20 billion Groq acquisition is a bet on the inference — As the AI industry shifts from model training to cost-efficient inference, Nvidia's Groq acquisition signals a strategic pivot to compete with Google's…
• The glass fabric holding the chip industry together — and Intel's — Why boron-free T-Glass became a critical bottleneck for AI chip production, and how Intel is developing a pure glass substrate as a future alternative.
• A home Bitcoin miner that heats your shower water — Superheat debuted at CES 2026 a home ASIC device that repurposes Bitcoin mining heat for water heating, spotlighting a broader inefficiency in how we…
• Why Nvidia skipped new GPU launches and what DLSS 4.5 signals instead — VRAM shortages driven by AI server demand blocked new GPU launches, so Nvidia responded with a Transformer-based DLSS 4.5 upgrade that generates graphics…
• Why a 2002 PS2 game suddenly sells for $400 — and nobody plays it — A leak of PlayStation 5 boot keys revealed that exploiting a critical vulnerability requires a rare physical copy of Star Wars Racer Revenge, sending its…
• A single server ran for four months to break the pi calculation — StorageReview set a new pi digit record using a single Dell server equipped with 384 CPU cores, 1.5 TB of RAM, and 2.5 PB of Micron SSDs serving as swap…
• The technology powering the Sphere's massive LED exterior — A look at how 1,500 Nvidia A6000 GPUs, 1.2 million LED tiles, and custom sync software bring the Sphere's exterior screen to life.
• Breaking a CPU overclocking record with liquid helium — From quartz oscillators to liquid-helium cooling at near absolute zero, this post explains the physics behind pushing a CPU past 9 GHz to set a world…
• Why Google may win the AI race — and it has little to do with its — An analysis of the four economic fundamentals — chips, cloud, data, and ecosystem — that will determine which AI companies survive, and why Google alone…
• How AI's appetite for HBM memory is driving up the cost of ordinary — As manufacturers shift production lines to the more profitable HBM chips powering AI accelerators, a RAM shortage for standard computers is expected to…
• Apple's N1 chip beats Broadcom with half the bandwidth — Apple's first in-house WiFi chip, the N1, uses a narrower 4nm design to outperform leading Broadcom chips in real-world use while cutting reliance on…
• Germanium: the element inside your phone, missiles, and quantum — A look at germanium's unique optical and electronic properties, its role in 5G, thermal imaging, and missiles, and the chalcogenide glass competing to…
• Gallium: the metal that melts in your hand and corrodes aircraft — A concise look at gallium's unusual properties, its role in power electronics and military radar, the complexity of refining it, and why China's dominance
• Jevons Paradox and Baumol's Effect: what AI will do to the economy — As AI cuts the cost per token, Jevons Paradox predicts demand will surge far beyond current levels, while Baumol's Effect explains why your plumber's…
• How a little-known Dutch chipmaker triggered a global auto industry — After U.S. sanctions and a Chinese countermove disrupted Nexperia's supply chain, automakers worldwide are facing shortages of components essential to…
• The $400 million machine at the heart of modern chip manufacturing — A look at how ASML's extreme ultraviolet lithography machine works, and the startup betting X-rays can do the job better and cheaper.
• Credo Technology: the cable maker quietly powering AI data centers — Credo's active electrical cables — each carrying onboard chips that clean and amplify signals — have made it a near-$25 billion company fueling the…
• Lab-grown diamonds are becoming a critical component in chips and — Lab-grown diamonds, which conduct heat five times better than copper, may soon enable faster chips and room-temperature quantum computing qubits.
• Behind U.S. moves on Intel: intelligence, not economics — A new opinion column argues that U.S. government actions regarding Intel are likely driven by intelligence-gathering motives rather than purely economic…
• Where true randomness comes from and why it matters for encryption — A practical look at true versus pseudo-random number generation, the hardware and software methods behind it, and the real-world consequences of getting…
• ASUS's gold-plated GPU pushes luxury PC hardware to a new extreme — ASUS's ROG Astral Dhahab line, sold exclusively in the Middle East, is set to get a 5 kg solid 24K gold variant worth over half a million dollars, paired…
• From a 2 MB disk in 1966 to the quantum limits of storage — A 2.9 kg disk built for UNIVAC computers in 1966 puts six decades of storage progress in perspective — and raises the question of where quantum physics…
• Google's patent puts 3D face recognition directly under the display — A newly filed Google patent describes an under-display 3D face recognition system that could rival Apple's Face ID while eliminating the need for a notch…
• How levitation works — and why it matters for climate and quantum — From Meissner-effect superconductors to optical tweezers and the Casimir force, this piece explains how levitation works and what room-temperature…
• The 1958 physics experiment that launched a $220 billion industry — A 1958 analog tennis game built on a repurposed oscilloscope is recognized as the first real-time video game, and the unlikely ancestor of today's gaming,
• Jensen Huang sees billions of robots filling the labor gap within 20 — In a Bloomberg interview, Nvidia's CEO outlined a future shaped by robot labor, physics-based AI training environments, and fully AI-managed data centers.
• Fool's gold and its surprising potential in clean energy — Pyrite, long dismissed as fool's gold, is a natural semiconductor that researchers are investigating for use in solar cells and high-capacity lithium…
• A brain implant built from the patient's own neurons aims to outdo — Science, founded by Neuralink's former president, is developing brain implants grown from a patient's own stem cells, potentially reducing rejection risk…
• Neuralink's brain implant for the blind gets FDA approval for human — Neuralink's Blindsight device pairs cameras with a brain implant targeting the visual cortex, potentially giving sight even to those blind from birth.
• From paper-savers to power-hungry: the environmental cost of modern — A 1994 photo of Bill Gates illustrates how computing shifted from saving forests to consuming vast amounts of fossil-fuel energy in the age of AI and…
• Inside Elon Musk's GPU cluster: how large language models are trained — A step-by-step breakdown of how language models learn — using Musk's massive GPU cluster as a starting point to explain forward passes, loss calculation,…
• CPU sockets explained: how a processor actually connects to a — A concise look at LGA and PGA socket types, how pin count affects data throughput, and the differences between Intel and AMD connection standards.
• DDR4 vs DDR5: why RAM type matters when buying a new PC — A practical breakdown of DDR4 versus DDR5 RAM — covering clock speeds, CL latency, and what the differences actually mean for everyday performance.
• DDR4 vs DDR5: why your next PC's RAM type actually matters — A clear breakdown of how DDR4 and DDR5 differ in clock speed, latency, and real-world performance, and why it should influence your next PC purchase.
• Apple's new iPad Pro is impressive — and almost nobody needs to — Five takeaways from Apple's iPad event: why a product that's too good becomes a business problem, what the new OLED display actually delivers, and where…
• Apple's on-device AI strategy may finally give Apple Silicon a reason — By betting on privacy-focused, on-device AI with its new openELM model, Apple may have found the killer use case that finally justifies its powerful…
• Why economic pressure on Israel and China tends to backfire — Attempts to weaken Israel and China through war or trade restrictions have repeatedly pushed both to build stronger, more self-reliant industries — a…
• What is inside Apple's $129 Thunderbolt 4 Pro cable — An internal scan of Apple's Thunderbolt 4 Pro cable reveals a complex printed circuit supporting data, video, and up to 100W charging, alongside Intel…
• The architecture of Apple's UltraFusion processor bridge — Bypassing traditional upgrades, Apple uses an UltraFusion bridge to merge two processors into a single chip transferring 2.5 terabytes per second.
• What makes Apple Silicon so fast and energy-efficient — A technical breakdown of Apple Silicon's edge, covering its SoC design, unified memory, ARM architecture, Metal API, and Apple's end-to-end control over…
• CPU vs GPU: why core count matters more than speed — A clear breakdown of how CPUs and GPUs differ in architecture, and why massively parallel but simple cores give GPUs the edge in gaming, crypto, and AI.
• What Sam Altman was really after with his $7 trillion ask — Altman's outlandish $7 trillion fundraising figure may be a calculated negotiation tactic that makes $100 billion seem reasonable while keeping his name…