MICROSOFT EXPANDS MAIA 300 CUSTOM AI ACCELERATOR ORDERS AT TSMC
Microsoft is negotiating with foundry leader TSMC to reserve manufacturing capacity for over 300,000 next-generation Maia 300 custom AI accelerators.
2026.08.12
Search the archive
Full-text search across article titles and summaries. Refine with sector, year, content type, and reading-time filters.
Results
415 matchesBOEING SELLS WISK, INSITU, AND SKYGRID TO ARCHER FOR A 16% STAKE
Archer Aviation is acquiring Boeing's Wisk Aero, Insitu, and SkyGrid in an all-stock deal that gives Boeing roughly a 16% stake in Archer plus a board seat.
2026.08.12
SEC EXEMPTS DATA CENTER BONDS FROM KEY SECURITIZATION RULES
The SEC's Division of Corporation Finance has issued guidance confirming that data center securitizations don't count as asset-backed securities under the Exchange Act, as they are treated as operating assets rather than self-liquidating financial assets.
2026.08.12
SONY AND TSMC TO INVEST $6.4 BILLION IN JOINT JAPAN IMAGE SENSOR PLANT
Sony Group and TSMC are in joint talks to invest $6.4 billion to build an advanced image sensor manufacturing plant in Kumamoto, Japan.
2026.08.12
SAMSUNG HBM4 PRODUCTION YIELD APPROACHES 80% AS RACE WITH SK HYNIX HEATS UP
Samsung Electronics has reportedly pushed manufacturing yields for its sixth-generation High Bandwidth Memory close to 80% following mass production refinements.
2026.08.12
HOW AMAZON BUILT A $2B DATA CENTER IN CALIFORNIA WITHOUT PUBLIC NOTICE
Amazon Web Services secured municipal approvals to construct a $2 billion, 438,500-square-foot data center in Gilroy, California, without triggering public town hearings.
2026.08.12
TESLA AND SPACEX COMMIT $16.8B TO BUILD "TERAFAB," A 100-MILLION-SQUARE-FOOT CHIP PLANT IN TEXAS
Tesla and SpaceX confirmed Terafab, their jointly developed chip factory, will rise in Grimes County, Texas outside Houston, with an initial $16.8 billion investment toward what Musk calls the largest building on Earth by manufacturing footprint. The vertically integrated site will handle manufacturing, packaging, and testing of both logic and memory chips, split between edge-inference silicon for Optimus robots and self-driving Cybercabs and high-power chips for SpaceX's planned space-based data centers, with SpaceX filings suggesting total spend across the project's multiple phases could reach $119 billion.
2026.08.10
TRUMP SIGNS EXECUTIVE ORDER PROTECTING US POLYSILICON INDUSTRY
Trump signed a Section 232 proclamation setting minimum import prices on polysilicon and its derivatives (the raw material behind both semiconductors and solar panels) along with a 15% tariff on downstream products, aiming to shield US producers from Chinese dumping. The order also authorizes Commerce to create an incentive program for companies building or expanding domestic polysilicon facilities, part of a broader push to rebuild a supply chain that's shrunk from 50% of global production capacity in 2005 to under 2% today, with China now supplying roughly 90%. The remedies take effect 120 days after signing.
2026.08.10
CHINESE CHIP TOOLMAKER CLAIMS WAFER POLISHING BREAKTHROUGH
Chinese equipment maker Hwatsing Technology has announced China's first in-line metrology system integrated directly into chemical mechanical polishing tools for six-inch wafers. By capturing real-time measurements before and after polishing, the system feeds live operational data back to control systems to improve wafer surface uniformity and minimize material waste. Shipped to a leading domestic optical silicon supplier, this chemical mechanical polishing breakthrough demonstrates how domestic toolmakers are quietly closing critical manufacturing gaps beyond advanced lithography, enhancing overall yield for mature semiconductor production lines across China.
2026.08.10
THE REAL DEBATE IN AI-RAN ISN'T WHETHER AI BELONGS IN 6G, IT'S WHERE
As the 3GPP prepares to set 6G's direction this September, Nokia and Nvidia's $1 billion partnership is pushing to put GPU-grade AI hardware directly inside radio access network equipment, with early trials at T-Mobile already running live 5G connections and AI video tasks on the same server. Skeptics like analyst John Strand and telecom CTO Kim Kyllesbech Larsen argue there's no clear economic case yet for loading every cell tower with expensive AI hardware, especially since wireless efficiency is already near its theoretical limits and some AI gains are achievable in software on existing infrastructure. Researcher Merouane Debbah frames the real architecture question as hierarchical: small, fast models embedded directly in radios for real-time decisions, more capable models at edge sites, and large foundation models running centrally for reasoning and coordination.
2026.08.10
Ω-0: ONE MODEL LETS HUMANOID ROBOTS WALK AND MANIPULATE OBJECTS AT THE SAME TIME
Household chores usually require a robot to move, keep its balance, and manipulate an object all at once, but most humanoid policies still treat locomotion and manipulation as separate problems solved one after another. ω-0 instead predicts whole-body actions directly, using compact learned representations of what the robot expects to see next as a lightweight guide, rather than trying to generate full future video frames, which keeps the model fast enough for real-time control while still reasoning about how the scene will change.
2026.08.10
FOUR DECADES OF IN-FAB METROLOGY AND INSPECTION: PIVOTAL MILESTONES AND ROAD AHEAD
Research institute imec published a comprehensive overview detailing four decades of in-fab metrology and inspection evolution across logic transistor scaling roadmaps. The review analyzes four distinct industry eras: predictable happy scaling, initial device material innovations, FinFET architecture transitions, and advanced EUV lithography adoption. As dimensional scaling reaches physical two nanometer limits and artificial intelligence workloads require 3D system stack integration, imec presents a forward-looking metrology framework combining specialized hardware sensor extensions with data-driven inspection algorithms to ensure manufacturing yield and process sustainability.
2026.08.10
SPACEX MAY DEPLOY 10GW DATACENTERS IN SPACE BY 2027
Research firm SemiAnalysis projects that SpaceX is on track to build 10 gigawatts of AI compute capacity by the end of 2027. Driven by sky-high inference margins generating over $100 billion per gigawatt annually for AI labs, SpaceX's "speed first" infrastructure strategy bypasses traditional data center construction bottlenecks. Offering rapid deployment alongside flexible 90-day cancellation clauses, SpaceX is securing massive off-take commitments with major hyperscalers like Microsoft. This aggressive expansion could drive up to $300 billion in annual recurring revenue.
2026.08.10
JANE STREET'S NEW PUZZLE: REVERSE ENGINEER AN ASIC FROM ITS RAW CHIP LAYOUT
Jane Street released a public puzzle, handing out only the final GDS file for a chip it designed, the raw geometric layout of every transistor and metal layer that a foundry uses to fabricate silicon, with nothing labeled. Solvers have to recover a netlist (the underlying logic gates and how they connect) directly from that layout, figure out what the circuit actually does, then use it to extract a hidden output value, all using free open-source tools like KLayout or Magic VLSI.
2026.08.10
SK HYNIX DETAILS $38 BILLION SPEND PLANS FOR BILLION ON CHIP FACTORY EXPANSION
SK Hynix approved a 54 trillion won ($38 billion) capital investment to construct two massive memory fabrication facilities to address looming AI infrastructure bottlenecks.
2026.08.10
HADRIAN VALUED AT NEARLY $8B AFTER RAISING $1.37B TO EXPAND AI-POWERED DEFENSE FACTORIES
Hadrian, which builds automated factories using AI and robotics to mass-produce precision parts for submarines, munitions, and drones, closed a $1.37 billion Series D at a $7.87 billion valuation, more than quadrupling its worth since January as it expands its manufacturing footprint to nearly three million square feet across four US facilities.
2026.08.10
CHINESE CHIPMAKERS POISED TO WIN
Chinese AI chipmakers are poised for significant growth as Beijing pushes self-reliance and Chinese companies plan to increase their spending on domestically made AI accelerators from 30% to 46% over the next year.
2026.08.10
ATI ROBOTICS BUILT A NEARLY CHINA-FREE HUMANOID ROBOT, AND THE TIMING LOOKS PRESCIENT
Indian startup Ati Robotics assembles its humanoid robots almost entirely from local and non-Chinese suppliers.
2026.08.10
FORMER SK HYNIX EMPLOYEE JAILED FOR LEAKING INFORMATION TO A CHINESE FIRM
South Korea's Seoul High Court upheld an 18-month prison sentence for a former SK Hynix employee who printed and photographed internal documents on CMOS image sensor technology and disclosed some of it to an unnamed Chinese firm's resume submission back in 2022.
2026.08.10
AI'S LATEST ROCKET SHIP IS AN OLD SCHOOL, 28-YEAR-OLD DATA COMPANY
DataDirect Networks, a 28-year-old managed data storage provider founded by Alex Bouzari, is experiencing massive growth driven by the generative artificial intelligence boom's requirements for high-performance storage architectures and KV cache acceleration solutions.
2026.08.10
AMD TO BUY STARTUP TAALAS FOR NEW AI CHIPS
AMD announced a definitive agreement to acquire Toronto-based chip startup Taalas to expand its specialized AI inference portfolio. Taalas develops custom silicon architectures that hard-wire individual AI model weights directly into semiconductor metal layers during manufacturing. AMD plans to integrate Taalas' low-cost, ultra-efficient inference technology directly into its broader accelerator roadmap alongside Instinct GPUs.
2026.08.07
ANTHROPIC IS BUILDING AN IN-HOUSE CHIP TEAM FOR CLAUDE
Anthropic has officially confirmed the assembly of an in-house chip design team to engineer custom AI processors tailored for its Claude platform. By co-designing hardware specifically alongside software models, the AI startup aims to slash inference costs, boost execution speeds, and optimize scaling for enterprise workloads.
2026.08.07
BLUE ORIGIN TRACES MAY'S NEW GLENN EXPLOSION TO A FAULTY BE-4 ENGINE VALVE
Blue Origin CEO Dave Limp said the May explosion that destroyed a New Glenn rocket during a prelaunch engine test originated at the main oxygen valve on one of its BE-4 engines, confirmed through hardware recovery and inspection. The fix is straightforward: small modifications to the valve that can be quickly retrofitted onto existing engines, with the company aiming to return New Glenn to flight later this year. The BE-4 also powers United Launch Alliance's Vulcan rocket, so ULA has been collaborating with Blue Origin on the investigation. The finding gives both companies a clear, containable root cause rather than an open-ended engine redesign.
2026.08.07
PCIE GEN 6 STORAGE ARCHITECTURE SHOWCASED
Microchip Technology and Micron showcased a high-performance PCIe Gen 6 storage system demonstration at the Future of Memory and Storage conference. Combining Microchip's 3nm Switchtec PCIe 6.0 fanout switches with Micron 9650 NVMe SSDs, the architecture achieves transfer speeds of 64 GT/s per lane—doubling available bandwidth over PCIe 5.0.
2026.08.07
D-WAVE'S DUAL-RAIL QUBIT GATE HITS 99.9% FIDELITY, CUTTING ERROR-CORRECTION OVERHEAD
D-Wave published research in Nature demonstrating a new two-qubit entangling gate for its superconducting "dual-rail" qubit architecture, which is designed so the most common quantum errors are also the easiest kind to detect and fix. The gate hit roughly 99.9% fidelity with fast ~500 nanosecond operation times, while preserving native hardware-level error detection, meaning errors get caught directly by the qubit design rather than requiring extra correction circuitry bolted on afterward.
2026.08.07
HOW TO SCALE AI ARITHMETIC EFFICIENTLY
This technical analysis explores how arithmetic specialization is essential for scaling artificial intelligence compute performance. As traditional transistor scaling slows down, hardware designers are replacing general-purpose IEEE floating-point units with specialized arithmetic engines tailored to low-precision representations like FP8, FP4, and Microscaling (MX) formats. By optimizing number format combinations and operation depth directly within silicon arithmetic units, modern AI accelerators drastically reduce energy consumption per operation while maximizing throughput without swelling total die area or power budgets.
2026.08.07
NVIDIA AND GOOGLE ARE CROWDING TSMC'S N3 NODE, CREATING A RARE OPENING FOR INTEL
Nvidia's next-gen Rubin GPUs are moving from 4nm to TSMC's 3nm node just as Google's TPUs, which are already on N3, plan to stay there, while new AI CPUs from Nvidia, Amazon, Microsoft, and Arm are converging on the same process simultaneously. TSMC has already hit its planned N3 capacity and is still adding more, warning the node could stay "very tight" for years, with three of its four Arizona fabs reportedly already fully booked despite only one being operational, driving reported price hikes of up to 25% for some customers. That squeeze creates a real opening for Intel, whose Fab 52 in Arizona just entered full production at 40,000 18A wafer starts per month, more capacity than TSMC's original Fab 21 campus.
2026.08.07
BREAKING THE MEMORY WALL WITH CXL
The Diligence Stack analyzes how Compute Express Link (CXL) technology breaks the persistent memory wall in high-density AI infrastructure. By utilizing PCIe physical interfaces to provide memory-semantic access, CXL enables memory expansion and multi-node pooling directly across server clusters. This architecture offloads Key-Value caches and massive vector databases from constrained GPU memory without severe latency penalties. Consequently, data centers can eliminate stranded DRAM resources, drastically reducing hardware costs, and scale memory capacity independently from fixed CPU channel limits.
2026.08.07
UNITREE EYES $7.4B VALUATION AHEAD OF SHANGHAI IPO
Chinese humanoid and quadruped robot maker Unitree is expected to be valued at 50.6-55.9 billion yuan ($7.4-8.2 billion) after its planned Shanghai IPO, a sharp jump from its 12 billion yuan valuation just last month, according to sponsor Citic Securities.
2026.08.07
MOOVE RAISES $250 MILLION TO SCALE ROBOTAXI FLEET INFRASTRUCTURE
Moove raised a Series C at a $2.1 billion valuation to build out “Nests,” robotics-first depots that charge, maintain, service, and orchestrate autonomous vehicles continuously.
2026.08.07
A REQUIEM FOR OPTANE: INTEL'S KV-CACHE KILLER THAT COULD HAVE EASED THE RAM PRICE CRUNCH
Intel's discontinued Optane memory, which thanks to its unmatched write endurance and low latency for small random writes, would have been an ideal fit for today's write-intensive LLM KV-caching workloads.
2026.08.07
LUMILENS RAISES $700 MILLION TO REPLACE DATA-CENTER WIRES WITH LIGHT
Lumilens emerged from stealth after securing over $700 million in funding, reaching a $5.51 billion valuation.
2026.08.07
TSMC SITTING ON $1B WORTH OF APPLE A20 PRO SOCS DUE TO DRAM DEFICIT
TSMC is reportedly holding nearly $1 billion worth of Apple A20 Pro processors in inventory that cannot ship due to severe global mobile DRAM shortages.
2026.08.07
FRORE CLAIMS LIQUIDJET DROPS NVIDIA RUBIN GPU TEMPERATURES BY 10°C
Frore Systems claims its LiquidJet coldplate technology can lower Nvidia Rubin GPU junction temperatures by up to 12°C, by etching targeted microchannels near thermal hotspots.
2026.08.07
WALDEN ROBOTICS SKIPS THE LEGS, BETTING WHEELS AND SIMPLE GRIPPERS WIN IN FACTORIES
Walden Robotics, a Toyota Research Institute spinout, emerged from stealth with $300 million in funding at a $1.1 billion valuation.
2026.08.07
ELON MUSK SAYS SPACEX WILL BUILD EXCLUSIVELY WITH NVIDIA, TOUTS SPACE-BASED AI SERVERS
Elon Musk announced that SpaceX will standardize exclusively on Nvidia hardware to power its expanding terrestrial data centers and upcoming orbital AI constellation. During SpaceX's Q2 conference call, Musk revealed a partnership with Nvidia to equip "Starmind" AI satellites with Vera Rubin architectures launching in 2027. Designed to operate in space and on Earth, SpaceX's custom rack designs prioritize power and cooling systems, targeting up to 20 gigawatts of infrastructure capacity to support Grok model scaling while bypassing terrestrial land constraints.
2026.08.05
HP, ASUS, AND ACER BEGIN USING CXMT MEMORY CHIPS
Major computer makers HP, Asus, and Acer have quietly begun sourcing DRAM memory chips from China's CXMT for select budget laptops. Triggered by a global memory deficit as leading suppliers Samsung, SK Hynix, and Micron shift manufacturing capacity to advanced High Bandwidth Memory, PC vendors are turning to alternative suppliers to keep production lines moving.
2026.08.05
SAMSUNG REVEALS NEW 3D-MEMORY ROADMAP, BETTING ON STACKING MEMORY DIRECTLY ATOP AI CHIPS
At FMS 2026, Samsung unveiled zHBM, a concept architecture that vertically stacks HBM directly on top of an AI accelerator instead of placing it alongside the chip, shortening the distance data travels and targeting roughly 8x the performance of next-gen HBM5 alongside more than 10x the memory density, using advanced wafer-bonding technology. Samsung also showed zNAND-O, an edge-AI-focused NAND design combining high space efficiency with low latency, and V10 BV-NAND, a bonding-based V-NAND architecture breaking past 400 layers for the first time, a roughly 58% density jump over its prior generation.
2026.08.05
INTEL FOUNDRY ENGINEERS BLUEPRINT FOR HYPER-LARGE FORM FACTOR AI CHIPS
At the IEEE ECTC conference, Intel Foundry unveiled novel packaging techniques designed to manufacture hyper-large form factor (HLFF) silicon measuring 240mm by 240mm. Addressing the physical limits of single-reticle chips, Intel's design blueprint combines compute die arrays with High-Bandwidth Memory (HBM) using EMIB-T embedded silicon bridges. To maintain reliability across massive package footprints, multi-point underfill material is introduced for dispensing material alongside thick stiffener rings and glass-core substrates that eliminate substrate warping and structural voids during thermal curing.
2026.08.05
WHEN EXPRESSIVE HUMANOID ROBOTS MESS UP, PEOPLE'S BRAINS TRIGGER A TRUST-WITHDRAWAL RESPONSE
Researchers found that when an expressive, animated humanoid robot made a social error, participants' brains showed coordinated activity linked to rising oxytocin, which predicted falling trust and less influence over their choices, an effect absent when the robot was stationary and expressionless. It undercuts the common design assumption that making robots more lifelike and expressive automatically earns more trust, suggesting expressiveness can backfire when the robot doesn't reliably deliver on the social behavior it signals.
2026.08.05
PRISMML SHRINKS A 27B-PARAMETER MODEL TO 3.9GB, SMALL ENOUGH TO RUN FREE ON AN IPHONE
Caltech-rooted startup PrismML compressed a 27-billion-parameter model down to just 3.9GB, small enough to run locally on an iPhone 17 Pro's neural accelerators, or on iPad, Mac, and PC, while retaining roughly 90% of the original's performance. A model of this size normally can't fit in a phone's usable memory at all, making this a real test of how far Apple's on-device silicon can be pushed. Apple has confirmed early talks with PrismML about the technology.
2026.08.05
OLIX RAISES $312M AT $3.3B VALUATION, APPOINTS NICK MCKEOWN TO BOARD
Optical networking startup OLIX has secured a $312 million Series B funding round at a $3.3 billion valuation to accelerate its specialized AI inference silicon. OLIX's X-1 platform uses custom DX-1 decode accelerators with high-speed SRAM and rack-scale optical interconnects instead of scarce High-Bandwidth Memory (HBM), drastically lowering latency for ten-trillion-parameter AI models.
2026.08.05
HUAWEI WARNS OF PHYSICAL SCALING LIMIT FACING NVIDIA
Huawei chief semiconductor scientist Liao Heng warned that Western chipmaker Nvidia is rapidly approaching an inescapable physical transistor scaling limit. Speaking in a rare interview, Liao argued that continually stacking massive compute dies with high-bandwidth memory creates severe reliability constraints. In response, Huawei is championing its proprietary "Tau Scaling Law" and LogicFolding architecture. The design approach prioritizes ultra-fast inter-component transmission speeds over brute-force transistor density, aiming to narrow the hardware performance gap without relying on advanced EUV lithography tools.
2026.08.05
GOVERNOR HALTS APPROVALS FOR TEXAS DATA CENTERS
Texas Governor Greg Abbott issued an executive directive freezing state grid connection approvals for new hyperscale data center projects, citing growing public concerns regarding electrical grid stability, water consumption, and opaque local tax incentives.
2026.08.05
KOENIGSEGG'S 86-POUND "DARK MATTER" MOTOR PUTS 800 HP IN A BACKPACK-SIZED PACKAGE
Koenigsegg built an 800-horsepower, 922 lb-ft electric motor weighing just 86 pounds, using a hybrid "raxial" design that merges radial and axial flux motor principles to pack in torque density without the usual bulk.
2026.08.05
EON EMERGES FROM STEALTH WITH $10.75M TO BEAM DATA CENTER TRAFFIC BETWEEN CONTINENTS VIA SPACE LASERS
Endeavor Optical Networks (EON) is targeting undersea fiber's biggest weakness, cables that are brittle, hard to access, and slow to repair, by building a constellation of roughly 20 laser-equipped satellites, each acting as a dedicated link between two continents.
2026.08.05
BASE POWER RAISES $1B TO PUT BATTERIES IN BACKYARDS
Home battery startup Base Power aims to stabilize fragile power grids strained by severe summer heat waves and rapidly expanding AI data center electricity demand.
2026.08.05
BLACKSTONE PITCHES $36B DEBT PACKAGE FOR ANTHROPIC CHIP LEASE
Private credit giants Blackstone and Apollo are arranging a record-breaking $36 billion debt package to finance Anthropic's massive AI accelerator expansion.
2026.08.05
AMD SALES OUTLOOK DISAPPOINTS INVESTORS AFTER AI-FUELED RALLY
AMD shares dropped in extended trading despite posting strong second-quarter earnings, as its forward revenue outlook failed to satisfy Wall Street's lofty expectations.
2026.08.05
US BANS FOREIGN-MADE ROBOTS, CHINA THREATENS RETALIATION
The FCC issued a ban this week on new, foreign-made humanoid robots, quadrupeds, robovacs, and sidewalk delivery robots, citing privacy risks from devices that map homes and public spaces plus supply-chain security concerns. China threatened retaliation, calling it a threat to "economic and trade stability." While the ban doesn't name China outright, the country controls roughly 85% of the global humanoid robot market. The rule doesn't target specific companies or actually curb data collection, and many US robotics firms that rely on Chinese components are worried it could gut their own supply chains instead.
2026.08.03