7월, 2026의 게시물 표시

Samsung Q2 2026 Profit Hits $58B as AI Memory Drives a Record Quarter

이미지
▲ Samsung Q2 operating profit near $58B, a tech record Quarterly operating profit is the money a company earns from its core business over three months. Samsung just rewrote the record books with it. If you use AI services, the shift now unfolding in the semiconductor market is worth understanding. What Changed for Samsung in a Single Year Samsung disclosed preliminary Q2 2026 results on July 7. Revenue came in near 171 trillion won and operating profit around 89.4 trillion won (about $58B ). A year earlier, operating profit was just 4.68 trillion won , so profit jumped roughly 19x . Only twelve months ago Samsung was mired in a memory downturn; now the picture has completely flipped. ▲ DRAM up 44%, NAND up 53% quarter over quarter AI Memory Powered the $58B Quarter The engine behind these numbers is memory. AI servers pack in huge amounts of HBM (High Bandwidth Memory, a chip that moves data at ultra-high speed) and DRAM. As companies race to build AI data centers world...

Humanoid Robot Safety: The 2028 ISO Standard That Decides the Race

이미지
▲ Robots' enemy isn't AI - it's gravity A humanoid robot is a machine built on the human body plan - two arms, two legs - so it can work in spaces designed for people. Those machines have left the lab for factories and warehouses - and the hardest problem changed with them. A July 5 Wall Street Journal report says the top priority is no longer intelligence - it's making sure the robot never lands on you. Why Falling Is a Humanoid Robot's Biggest Risk Next-generation humanoids weigh 150-200 lbs - the mass of an adult, in metal. As one safety engineer told the WSJ, "the biggest risk is gravity": a machine that heavy can crush the person beside it. Unlike caged industrial arms repeating fixed motions, humanoids run on probabilistic AI judgment calls, so their next move is harder to predict. No serious injuries have been reported yet - but the machines are arriving faster than the rules. ▲ Nvidia, Neura, Dexmate fall-proof strategies Nvidia, Neura...

OpenAI Offers US Government a 5% Stake - $42.6B Alaska-Style AI Fund

이미지
▲ OpenAI 5% government stake card news A sovereign wealth fund is a state-owned investment fund that grows public assets and shares the returns with citizens. That once-obscure term just landed in Silicon Valley: the Financial Times reports OpenAI has floated handing the US government a 5% stake. The question of who owns AI's wealth is now officially on the table. Why OpenAI Offered Washington Equity The timing explains the move. The proposal surfaced six days after Washington froze the full launch of GPT-5.6, which in late June went only to users vetted by the government. Rival Anthropic spent most of June offline under US export controls (government limits on tech distribution), returning July 1 . Add last year's 10% Intel purchase, and the pattern is clear: Washington now gatekeeps AI launches, and OpenAI chose to negotiate. ▲ Stake math: $42.6 billion, Alaska model The Alaska Model and the $42.6 Billion Math OpenAI's March funding round set its valuation ...

GM Micron Memory Supply Deal 2026 - The Chip Crunch Reaches Cars

이미지
▲ GM Micron long-term auto memory deal A Strategic Customer Agreement (SCA) is a long-term contract in which a chipmaker guarantees supply volumes to one major customer. On July 1, 2026 , General Motors signed exactly that kind of memory deal with Micron - a sign that memory chips now shape the production schedules of finished-goods industries like autos. Why GM Went Straight to a Memory Maker The starting point is the AI-driven memory crunch. AI data centers are buying up memory at unprecedented scale, leaving every other industry scrambling. Meanwhile, today's cars are rolling computers: driver-assist (ADAS) and AI cockpit systems keep raising the amount of memory each vehicle needs. Automakers also remember idling production lines when chips ran short - so this time, the carmaker went to the chipmaker first and locked in volumes. ▲ GM Micron deal key numbers Inside the GM Micron Deal: LPDRAM to a $2B Virginia Fab The agreement has three pillars. First, GM secures lon...

UBTech U1 Home Robot Launch: 13,361 Day-One Orders (2026)

이미지
▲ UBTech U1: 13,361 orders on launch day A home companion robot is a human-shaped machine built to live with people at home rather than work on a factory floor. On July 1, 2026, China's UBTech launched one called U1, and if you follow consumer tech, it signals that robots may soon appear in living rooms, not just warehouses. Why the UBTech U1 launch matters UBTech is the world's first publicly traded humanoid robot company, listed in Hong Kong. Until now, humanoids from rivals like Tesla and Figure were aimed at factories and warehouses. U1 breaks from that. It is designed as a companion that lives alongside people at home. A listed robot maker betting on the living room marks a notable shift for the whole industry. ▲ 183cm, 88 joints, from about $17,650 UBTech U1 specs and price U1 comes in a 183cm male build and a 168cm female build, each with 88 joints and silicone skin that lets it talk and hold eye contact. Its emotion AI runs on-device, meaning it processes...

Meta's Operation Cannes - 45,000 Secret Prompts Probing Rival Chatbots

이미지
▲ Meta Operation Cannes rival chatbot probe Operation Cannes is Meta's covert project that hired hundreds of contractors to pose as children and secretly probe rival AI chatbots' safety guardrails. If a kid in your life talks to chatbots, this is your story: the experiment exposed the actual safety level of the AI tools children use every day, per a WIRED investigation out this week. How Meta's Operation Cannes Worked The operation was run by Covalen, a Meta contractor, using hundreds of workers based largely in Kenya. They created fake accounts with under-18 birthdates and approached ChatGPT, Gemini, and Character.AI. Posing as children in crisis, they asked about suicide, self-harm, eating disorders, sex, and drugs, then logged the answers in spreadsheets. The targets had no idea. ▲ 45,000 crisis prompts key facts 45,000 Crisis Prompts: The Key Facts A single testing round in August 2025 sent more than 45,000 prompts , and Operation Cannes stayed active unt...

Claude Fable 5 Returns as US Lifts Export Controls, Sonnet 5 Launches

이미지
▲ Claude Fable 5 returns worldwide after US lifts controls Claude Fable 5 is Anthropic's newest AI model, and for two weeks it disappeared for users around the world. On July 1, 2026, it returned. If you use AI tools day to day, this episode matters, because the cause was a US government export control, not a technical outage. Why Claude Fable 5 Went Dark for Two Weeks The story began on June 12 , when the US government placed export controls on Fable 5 and its sibling Mythos 5, barring foreign nationals from using them. With no way to verify a user's nationality in real time, Anthropic suspended access for every user at once. The trigger was a report from Amazon researchers, who found a way to slip past Fable 5's safeguards (a "jailbreak") and surface software vulnerabilities. ▲ June 30 controls lifted, July 1 global restore Export Controls Lifted and Sonnet 5 Arrives On June 30 , the US Commerce Department lifted the controls after Anthropic agreed t...