Topic
98 articles
China built Pacific internet infrastructure across 12 island nations. The US countered with rival cable funding. How cable geopolitics decides sovereignty in Tonga, Solomon Islands, Kiribati, Fiji and beyond.
Alibaba's T-Head shipped 560,000 Zhenwu AI chips to 400+ customers. The M890 outperforms Nvidia's H20. Here is what US export controls actually built in China.
The Pentagon's 1260H list grew to 188 entities June 8, adding Alibaba, Baidu, BYD, Nio, Unitree, and WuXi AppTec. DoD contracting bans start this month; third-party procurement bans hit June 2027.
Nvidia released DSX OS as open source — the AI factory software behind DGX Cloud. MaxLPS recovers stranded power for up to 40% more GPUs per megawatt; NVSentinel automates GPU fault recovery.
China's NMPA approved Neuracle's coin-sized NEO implant for commercial use — epidural design, 36 trials, spinal-cord patients — while Neuralink stays in clinical trials with 21 users.
BIS confirmed AI chip export controls apply to any Chinese-parented company globally, closing a loophole that let hundreds of thousands of Nvidia Blackwell servers reach Chinese subsidiaries in Malaysia and Singapore.
Nvidia fell harder than Nasdaq on June 5 as Broadcom dropped 15% on AI demand concerns. Here is why NVDA amplifies AI corrections and the investor outlook.
Goldman Sachs told IPO investors SpaceX AI revenue will surge from $3.2B in 2025 to $322B by 2030 — 68% of $474B total — as the $75B SPCX roadshow prices a $1.75T debut.
SpaceX fixed its IPO at $135 per share for a $1.75T valuation and $75B raise. Roadshow starts June 4, 2026; SPCX Nasdaq debut June 12. All-primary, 30% retail tranche.
On June 3, 2026 the EU launched its Tech Sovereignty Package: Cloud and AI Development Act to triple data centre capacity, Chips Act 2.0, and an EU Open Source Strategy.
Intel unveiled Xeon 6+ on 18A at Computex June 3, 2026 for agentic AI orchestration, 200GbE E835 adapters, and Crescent Island GPU details up to 480GB memory.
China State Council outbound investment rules effective July 1, 2026 tighten tech, data, and talent exports. Beijing can unwind completed deals after Meta Manus and Nexperia.
Shanghai Lingang undersea data center switched on May 2026: 24MW planned, offshore wind direct feed, seawater cooling, 192 racks. World-first model for AI power and latency.
IAEA chief Rafael Grossi met Saudi energy and foreign ministers in Riyadh June 3, 2026 after Barakah. Likely talks: SNAEP, Gulf nuclear safety, US 123 safeguards.
NVIDIA unveiled FOX factory AI at GTC Taipei June 4, 2026. Foxconn MoMClaw runs hundreds of agents on DGX Station GB300. 80% faster root-cause analysis claimed.
Broadcom Q2 FY2026 AI chips hit $10.8B (+143% YoY) on $22.2B revenue, but Q3 AI guide of $16B missed the $17.2B Street view. AVGO fell ~14% June 4.
SpaceX lowered its IPO target to at least $1.8T from $2T+ ahead of the June 4 roadshow. Pricing as early as June 11, SPCX debut ~June 12. $18.7B 2025 revenue, $4.94B loss after xAI.
Senators Warren and Kim accused the Trump admin June 2, 2026 of letting advanced US AI chips reach China via export gaps. BIS May 31 guidance closed China-HQ subsidiary loophole.
At COMPUTEX June 1, 2026 Jensen Huang unveiled N1X (RTX Spark): MediaTek SoC, 20 Grace cores, Blackwell GPU, 128GB memory, full CUDA on Windows fall 2026.
Commerce BIS weekend guidance (May 31, 2026): advanced AI chips to China-headquartered buyers need export licenses even outside China. Covers Blackwell, Rubin, AMD MI350x.
At COMPUTEX June 2026, Nvidia launched open Cosmos 3 world models (20T tokens, super/nano) for robots and AVs, plus RTX Spark N1X Windows chips with Microsoft, Dell, Lenovo.
June 1, 2026: Nvidia sells Isaac GR00T research humanoid on Unitree H2 Plus with Jetson Thor Blackwell. Unitree STAR IPO approved — 4.2B yuan raise, ~$6.2B valuation, 73-day review.
Nvidia Vera CPU launched May 31, 2026 at GTC Taipei: 88 Olympus Arm cores, 1.8x faster than x86 on agent workloads. Anthropic, OpenAI, NYSE, OCI, CoreWeave adopt. Fall 2026 availability.
China's cheap electricity and SMIC's record $9.3B revenue in 2025 reveal how US export controls inadvertently accelerated China's domestic chip industry. Data center racks grew 30% annually since 2016.
China imposed export controls on dual-use items and critical materials to Japanese military end-users in January 2026, escalating a chip-war front that now includes East Asia's two largest economies.
China launched a national humanoid robot ID system in May 2026: 29-digit codes, 28,000 units registered, 100+ firms enrolled. No code, no market access rule for domestic deployment.
Samsung DS workers ratified a 10-year bonus deal May 27, 2026: 10.5% of operating profit as stock, ~$340K average per chip employee, $26.6B pool tied to AI memory demand.
Supermicro co-founder Wally Liaw was arrested in 2026 for conspiring to smuggle $2.5B in restricted Nvidia H100 GPUs to China, with $510M already delivered via Southeast Asia shell companies.
Dell Q1 FY2027 earnings: AI server revenue hit $16.1B (+757% YoY), orders booked $24.4B, backlog reached a record $51.3B. FY27 AI revenue target raised to $60B. Stock surged 32%.
Huawei unveiled the Tau Scaling Law and LogicFolding architecture at IEEE ISCAS 2026, claiming 55% higher transistor density and 41% power efficiency without EUV lithography.
Span and Nvidia are deploying XFRA nodes — liquid-cooled boxes with 16 RTX Pro 6000 Blackwell GPUs — on residential homes, tapping unused grid capacity. 100 pilot units launch Q3 2026 on PulteGroup builds.
Cerebras (CBRS) debuted May 14, 2026 at $185, opened near $350, and hit a $100B+ market cap intraday on 68% first-day gains. WSE-3 targets inference vs Nvidia.
What Computex 2026 signals from Jensen Huang and TSMC imply for Vera Rubin timelines, packaging bottlenecks, and developer infra planning.
Export enforcement pressure around Taiwan supply chains and server routes is rising. A compliance and procurement playbook for platform teams.
Samsung and its 48,000-worker union reached a tentative deal on May 20, one day before the strike start. The $43B AI memory production risk is off the table for now.
Trump cleared H200 sales to Alibaba, Tencent, ByteDance. Beijing blocked its own companies from taking delivery. Jensen Huang says Nvidia has largely conceded China's AI market.
Samsung strike starts May 21. 45,000+ workers, 18 days, $700M per day in losses. HBM and DRAM production at risk. JPMorgan estimates $20.8B operating profit hit.
TSMC raised its 2030 forecast from $1T to $1.5T. AI and HPC account for 55%. Nine fab construction phases in 2026, three Arizona fabs in pipeline.
India Semiconductor Mission 2.0: Tata+PSMC first silicon late 2026 at Dholera. Micron ATMP open in Sanand. Rs 8,000 crore Budget 2026 outlay. Qualcomm 2nm tape-out from Indian design centres.
Huawei Ascend 950PR is CUDA-compatible. ByteDance commits $5.6B, Alibaba and Tencent ordering. 2.8x H20 FP4 performance. 750,000 units planned 2026. China's CUDA moat with Nvidia is broken.
Chinese startup Dishan Technology completed 2nm AI GPU prototype verification. FinFET/GAA hybrid, 40% efficiency gain. The problem: no foundry in China can manufacture it. TSMC blocked, SMIC not ready.
First Taiwan 2022 chip-security convictions: 10 years for ex-TSMC engineer, Tokyo Electron Taiwan fined $5M over 2nm PDK theft. Apr 27, 2026.
US and Taiwan finalize a chip-linked tariff framework scaling TSMC Arizona to about $465B across eleven coordinated fab phases. Apr 27, 2026.
Rapidus begins 2nm-class test production in Hokkaido, ships PDKs to first customers; Japan cumulative public support reaches ~$16.3B. Apr 27, 2026.
Qualcomm CEO in Seoul April 27: Samsung Foundry 2nm returns with TSMC dual-sourcing for Snapdragon 8 Elite 2 flagship Android SoCs and modems.
DeepSeek launched April 26 2026 a V4 model adapted for Huawei chips — trained partly on Huawei hardware. Fastest model to top Hugging Face. SMIC +10%, Huahong +15% on the news. Nvidia dependency broken.
Stanford HAI 2026 AI Index: US-China Arena gap shrank from 1,300 points (2023) to 39 points (2.7%) by March 2026. AI talent to US down 89% since 2017. China leads AI citations worldwide.
AEI 2026 report: China runs SMIC 7nm AI chips using DUV immersion lithography despite the EUV export ban. 5nm at 20% yield. 1.6M dies targeted in 2026.
Tesla Q1 2026 earnings April 22: capex raised to $25B — triple last year. Optimus volume production starts Fremont July 2026. Cybercab on schedule. AI compute doubling in 6 months. Negative FCF rest of 2026.
Intel Q1 2026 results: Data Center and AI division hit $5.1B revenue, up 22% YoY, beating estimates by $700M. Stock surged 25%. Q2 guidance $13.8-14.8B. Agentic CPU demand driving growth.
US intel: China delivered 2,000 tons of sodium perchlorate to Iran — enough for 785 missiles. MANPADs imminent. Trump threatened 50% tariffs. China denied everything.
TSMC posted $35.7B in Q1 2026 revenue — up 35% YoY, a new record. N2 2nm chips entering volume production. AI accelerator CAGR revised to 54%. What it means for GPU pricing and developers.
Trump paused tariffs 90 days for most countries at 10% but raised China to 145% on April 9. What it means for GPU prices, TSMC, iPhone, and developer budgets.
Hub for Nvidia, AMD, TSMC, ASML, HBM, export controls, and China fabs: how silicon constraints hit AI training, inference, and developer hardware.
US tariffs of 20-32% on semiconductor manufacturing equipment kick in April 9. Equipment cost drives chip prices — GPU inference and API rates will follow upward.
TSMC raising wafer prices 5-30% on advanced nodes due to US tariffs. Apple absorbed $3.3 billion so far. Here is what it means for GPU inference and developer API costs.
Meta deployed MTIA 400 chips — 6 petaflops FP8, 288GB HBM, 72 per rack — for GenAI inference. What it means for Llama API costs, Nvidia GPU demand, and open source AI economics.
Trump's Liberation Day tariffs hit April 2, 2025. One year on: laptops cost 18% more, cloud hardware expanded in US, and a second chip tariff arrived in January 2026.
April 2, 2025 was Liberation Day. One year later: Supreme Court ruled them unconstitutional in Feb 2026, $166B refund underway for 330,000 businesses. Tech impact explained.
Trump-Xi Beijing summit ended with an uneasy truce. China failed to get Blackwell export controls lifted. Taiwan never came up. The deal expires in 12 months.
Trump's January 2026 Section 232 tariff put 25% on Nvidia H200 and AMD MI325X exports. US data centers are exempt. Here's what it costs developers outside the US.
DRAM prices surged 171% YoY in 2026. A 32GB DDR5 kit hit $364, up from $80. OpenAI Stargate locked 40% of global wafer output. Here's the full developer impact breakdown.
Trump administration allows Nvidia H200 exports to China while banning Blackwell B100/B200/B300. What this split policy means for cloud compute access and developers.
TSMC 3nm node hit full capacity in March 2026 for the first time. Apple M5, Nvidia Blackwell, Amazon Trainium3, and Microsoft Maia2 all compete for the same fab slots.
Beihang and Harbin Institute bought Nvidia A100 servers through Super Micro shell firms in 2025-2026 despite U.S. export bans. Super Micro co-founder arrested. Here is how the smuggling worked.
DRAM prices rose 171% year-over-year and DDR5 spot prices quadrupled since September 2025. AI data centers are consuming memory supply faster than fabs can produce it.
SK Hynix ordered $8B of ASML EUV machines on March 24, the largest disclosed order in ASML history. The tools will produce HBM4 for Nvidia's Vera Rubin platform launching late 2026.
Qwen 3.5's 9B model outscores OpenAI's 120B model on GPQA Diamond, MMLU-Pro, and multilingual tests. Apache 2.0, multimodal, runs offline on 16GB RAM. Full developer guide.
Supermicro co-founder Wally Liaw was arrested March 19 for smuggling $2.5B in Nvidia B200 and H200 GPUs to China. The scheme used Southeast Asian front companies and staged fake servers to fool US auditors.
Hua Hong's Huali Fab 6 in Shanghai is readying 7nm chip production. US-blacklisted GPU maker Biren is already taping out there. What it means for China's AI chip stack.
Qatar expelled Iran's military attaches and halted all LNG production after Iranian missiles struck Ras Laffan. Gulf states say "price must be paid." Here is the full tech and semiconductor supply chain impact.
The 2026 Iran war triggered the largest oil supply disruption in history — 10 million barrels per day gone. Brent hit $110. Qatar helium is down one-third. Here is the full tech and semiconductor impact.
Super Micro co-founder Yih-Shyan Liaw was arrested for smuggling $2.5B in Nvidia AI servers to China using shell companies and dummy servers with serial numbers swapped using a hair dryer. Stock crashed 28%.
ShinyHunters stole 1PB of data from TELUS Digital by pivoting from a Salesloft breach via trufflehog. $65M ransom demanded, 24 BPO companies exposed.
Tesla's Terafab is a $25B bet on 2nm chip manufacturing targeting 100K wafer starts per month. If it works, Tesla cuts dependence on TSMC and Nvidia for AI hardware.
Iran attacked Gulf energy sites on March 19 including Qatar's Ras Laffan LNG hub, pushing Brent crude to $119 and cutting global air freight capacity 9%, delaying semiconductor shipments.
Jensen Huang confirmed Nvidia is restarting H200 manufacturing for China with US export licenses secured. ByteDance, Alibaba and Tencent approved for 400K+ units, capped at 75K per customer.
Jensen Huang's GTC 2026 keynote raised Nvidia's demand forecast to $1 trillion through 2027, confirmed Vera Rubin GPUs, NemoClaw agents, and an Uber autonomous driving deal.
ByteDance is deploying 36,000 Nvidia B200 Blackwell chips in Malaysia via a $2.5B deal, using Southeast Asian cloud firms to legally access hardware blocked under US export controls.
India is planning a $10.8 billion semiconductor fund covering chip design, manufacturing equipment and supply chain. Micron and Tata are already building fabs in Gujarat under the existing $10B program.
AWS and Cerebras announced a disaggregated inference architecture combining AWS Trainium and Cerebras CS-3 chips on Amazon Bedrock, claiming order-of-magnitude faster AI inference than GPUs.
Meta unveiled its MTIA chip roadmap in March 2026 — four generations of custom RISC-V inference chips made by TSMC and designed with Broadcom, with MTIA 300 already in production.
Nvidia is launching NemoClaw, an open-source AI agent platform for enterprise workforces. It's hardware-agnostic, not CUDA-locked, with Salesforce, Cisco, Google, and CrowdStrike already on board.
The Strait of Hormuz has been disrupted since Feb 28, 2026. Brent crude hit $126/barrel. European LNG rose 60%. AI data centers burning gas are now paying significantly more to run.
The IEA approved the largest emergency oil release in history after the Strait of Hormuz closed. Brent crude is still above $90. AWS data centers in UAE and Bahrain were hit by drones. Qatar's helium supply is offline, threatening chip fabs globally. Here is the full developer and infrastructure impact.
Intel's 18A process node is the most ambitious foundry bet in semiconductor history. With $8.5B in CHIPS Act funding, RibbonFET transistors, and Microsoft as a confirmed customer, can Intel actually challenge TSMC — and why does it matter for AI hardware and US chip independence?
China's SMEE has a working DUV prototype and active EUV program. If it ships by 2027, US export controls on China's chip industry effectively fail.
China is running the largest state-directed AI programme in history — often called its "AI Manhattan Project." But US and allied export controls have critical gaps. Here is how China is navigating restrictions, what the gaps are, and what this means for global AI competition.
Japan's Rapidus, TSMC Dresden, Intel Germany, and Samsung Texas are all racing to establish leading-edge semiconductor manufacturing outside Taiwan. Here is the state of each programme, the strategic stakes, and what it means for AI hardware supply chains.
Micron's Sanand ATMP facility in Gujarat has begun commercial production in early 2026, marking India's first modern semiconductor manufacturing output. This post unpacks what that actually means for AI ready chips, the India AI Mission, and global supply chains.
Qatar Energy force majeure after the Ras Laffan strike halted one third of global helium supply. Which chip fabs are most exposed and what it means for AI hardware timelines.
PLA military planning around a 2027 decisive capability date has put TSMC at the centre of every AI hardware risk discussion. This post explains the military context, TSMC's role, the Arizona hedge, and what developers and enterprises should plan for in 2026.
The United States now applies steep tariffs on Chinese electronics and is heading toward a June 2027 review cliff, while GPU exports are governed by case by case rules. This post explains what that mix means for AI hardware buyers and how to plan budgets in 2026.
China submitted its 15th Five-Year Plan to the National People's Congress on March 5, 2026. AI appears 50+ times. Semiconductors barely register. Washington's chip export strategy is targeting the wrong layer. Here is what developers and tech strategists need to understand.
India's first semiconductor assembly plant opened in Sanand, Gujarat on February 28, 2026. Micron is now shipping DRAM and NAND chips to Dell, Asus, and Qualcomm from Indian soil.
SK Hynix completed the first HBM4 chip in January 2026, targeting 70% of Nvidia Rubin GPU supply. Samsung is close behind. South Korea controls the memory that powers all frontier AI.
Huawei plans to produce 600,000 Ascend 910C AI chips in 2026, nearly doubling 2025 output. China's AI companies are training models on a hardware stack completely separate from Nvidia and CUDA.
Apple's first smart home hub launches Spring 2026 assembled in Vietnam by BYD. Samsung is investing $1.8B in OLED production there. Vietnam targets 50,000 semiconductor engineers by 2030.