Singapore AI News & Daily Briefing

Nvidia signed a $1.5 billion agreement with Amkor Technology for chip packaging services at Amkor's new Arizona facility. This prepayment will help Amkor expand US-based manufacturing, reducing reliance on Asian production and creating a more resilient supply chain for AI chips. Amkor shares surged 15% after the announcement. The deal is part of Nvidia's broader push to secure every stage of AI chip production amid explosive demand for AI accelerators.

Major tech companies including Nvidia, Microsoft, Amazon, Meta, and Alphabet are spending over $800 billion on AI infrastructure this year, with plans to exceed $1.2 trillion in 2027—more than the US military budget. Oracle's debt was downgraded to near-junk status due to heavy AI data centre spending, while SpaceX bonds trade like junk despite its recent IPO. The analysis warns that much of this spending runs on borrowed money, with revenues tied to unproven AI startups, drawing parallels to the dotcom era.

SK hynix, the South Korean memory giant that supplies high-bandwidth memory (HBM) for AI chips, expects record Q2 profits driven by surging demand for AI server memory. The company holds 57% of the HBM market for AI applications. This signals continued strong demand for AI infrastructure components.

Anthropic has launched Claude Opus 5, its latest flagship AI model for businesses, claiming it matches its top-tier Fable 5 model in coding tasks at half the cost. Priced at US$5 per million input tokens and US$25 per million output tokens (same as Opus 4.8), the new model posts top scores on coding and automation benchmarks while being more aligned to safety guidelines than previous versions. Early partners like Cursor, Zapier, and Devin reported significant gains in debugging accuracy and workflow automation. For everyday users, cheaper enterprise AI coding tools could eventually mean better software, faster app fixes, and lower subscription costs for AI-powered developer tools.

Nvidia CEO Jensen Huang argues that US export restrictions on advanced AI chips are backfiring. Rather than slowing rivals like China, he says these bans force them to build independent systems and write better software out of necessity. Huang believes trade restrictions on computing hardware misunderstand how technology scales — rivals can connect more older chips with abundant electricity to achieve similar results. He warns this ultimately hurts American tech companies by pushing foreign markets toward local alternatives.

Samsung launched two versions of its Galaxy Z Fold 8: the flashy Ultra ($2,099) and the standard Fold 8 ($1,899). A reviewer argues the regular Fold 8 might actually be the smarter buy. It has a wider 4:3 inner display that feels more natural for reading and videos, a more manageable cover screen, and saves you $200. The trade-offs are a smaller camera (50MP vs 200MP) and slightly smaller battery—issues the reviewer says won't matter much in daily use.

AI isn't just changing your workday — it's moving into your home. A recent survey found 81% of parents have used AI for parenting tasks, while 64% of U.S. teens regularly chat with AI bots for homework and entertainment. Families are using AI for grocery shopping, managing homeschool schedules, handling paperwork, and even emotional support. OpenAI is reportedly building a screen-free smart speaker designed as an 'AI companion' for the home. The big concern: unlike schools or offices, homes have almost no AI guardrails. Children are using AI privately in bedrooms, raising questions about privacy, social development, and what data is being collected about family life.

Flock Safety, an American company backed by nearly $1 billion in venture capital, has installed AI-powered cameras that read license plates across thousands of U.S. neighbourhoods. These cameras feed into a searchable national network used by law enforcement — and Americans across the political spectrum are pushing back. Libertarians see government overreach, MAGA supporters distrust the 'deep state' surveillance, and progressives worry about vulnerable communities being targeted. Critics say Flock's data retention lasts 30 days by default (some states limit it to 3 minutes), and that the company essentially trades people's location data for profit. Audit logs in Cleveland revealed out-of-state agencies — including immigration-related searches — had been accessing the city's data without clear oversight.

WhisperLive is a free, open-source tool that lets you run OpenAI's Whisper speech-to-text technology in near real-time on your own computer. No cloud subscription needed. It's built by Collabora and designed for anyone who wants live transcription without sending audio to external servers.

A concept called 'distillation' has exploded into a major tech policy debate. Distillation means using outputs from a powerful AI model to train a smaller, cheaper one — think of it as a shortcut that could let rivals skip billions in research costs. Chinese lab Moonshot AI recently released Kimi K3, which users quickly found competitive with OpenAI and Anthropic. The White House accused Moonshot of stealing American IP through industrial-scale distillation. But major tech companies including Nvidia, Microsoft, Meta, and Palantir fired back, warning against 'premature restrictions' on open-weight AI models. The core tension: stopping IP theft versus preserving innovation and competition.

This is a Tech in Asia newsletter editorial arguing that 'consistency beats novelty' in business, using Spain's World Cup win as an analogy. The piece highlights Fireside Ventures' decade of consumer investing in India, profiles Chinese founders dominating Southeast Asia (notably PixVerse, Southeast Asia's first AI unicorn in March), and notes Lazada founder Max Bittner's new AI wealth management startup. While AI makes appearances, the core message is about old-school business discipline—showing up daily, not chasing hype—even as technology evolves.

A Florida man is suing OpenAI after ChatGPT allegedly gave him medical advice that nearly killed him. This is among the first major lawsuits testing whether AI companies can be held legally responsible for dangerous outputs. The case could set an important precedent for AI safety and accountability.

TechInAsia compiled a list of the most active investors backing India's AI startups over the past two years. The list is arranged by deal count and aims to help founders identify potential funding sources. The full list is behind a paywall, but the article notes Singapore-based investors like Temasek and GIC are known players in the broader Southeast Asia AI investment landscape.

Nvidia CEO Jensen Huang rallied 50 tech companies — double the original 25 — to sign a letter urging Washington not to restrict downloadable AI models. OpenAI signed on, but Amazon and Anthropic stayed out, which the article says matters because both sell frontier AI directly. For businesses, the article breaks it down: open weights let you inspect and modify AI models you run, keep your data on your own servers, but still don't solve who or what is acting in your name when an AI agent moves money or signs something. The advice: map which AI vendor you depend on for what, test at least one open-weight model for real workloads, and decide on agent identity architecture before the policy debate settles without you.

Monday.com just became the latest tech company to cut jobs while pointing to AI — laying off 600 workers, about 20% of its staff. According to new Financial Times analysis, US tech firms have slashed nearly 140,000 jobs in 2026 so far, with Amazon, Oracle, Meta, and Microsoft alone accounting for almost 50,000 of those cuts. The pattern is clear: companies are redirecting hundreds of billions of dollars into AI infrastructure while reducing headcount. Interestingly, the stock market hasn't rewarded this strategy — companies citing AI in layoff announcements have underperformed the Nasdaq by nearly 10% after their cuts. For everyday workers, this signals that AI adoption doesn't just create new roles — it eliminates existing ones, and fast.

Chinese AI startup DeepSeek has raised over $7.4 billion (50 billion yuan) in its first-ever funding round, valuing the company above $50 billion. The deal structure uniquely preserves founder Liang Wenfeng's control: investors put money into a limited partnership he manages, agreed to a five-year lock-up, and received no voting rights. Founder Liang contributed $2.96 billion of his own money. Major Chinese tech firms Tencent and battery maker CATL were reportedly considering large investments. China's National AI Investment Fund was the only investor with voting rights. DeepSeek gained global attention in early 2025 when its V3 and R1 AI models challenged US assumptions about China's AI capabilities.

Hyundai Motor Group plans to deploy 30,000 Atlas humanoid robots annually by 2028, with over 25,000 assigned to its US manufacturing plants. Boston Dynamics' Atlas robots can lift up to 50kg, reach 2.3 meters tall, and self-drive to charging stations. Hyundai is partnering with Google DeepMind to add AI reasoning capabilities to the robots for factory work. The robots will start at Hyundai's Georgia plant in 2028, then Kia's nearby facility in 2029. This represents a significant push toward AI-powered automation in auto manufacturing, potentially affecting factory worker roles.

Anthropic has launched Claude Opus 5, its latest enterprise AI model, claiming it matches its top-tier Fable 5 model in coding tasks at roughly half the cost. Priced at US$5 per million input tokens and US$25 per million output tokens (unchanged from Opus 4.8), the model posted top scores on coding and business-automation benchmarks. Early partners including Cursor, Zapier, Devin, and Box reported significant gains in debugging accuracy and workflow automation. Anthropic says Opus 5 is its most aligned model yet with lower deceptive behaviour rates, though it still lags behind Mythos 5 on offensive cybersecurity tasks. A faster 'Fast mode' is available at double the base price.

Nvidia CEO Jensen Huang argues that US export controls on AI chips are backfiring, forcing China to build independent systems faster while American companies lose the Chinese market. He claims rivals can connect older processors and use abundant energy to compensate for lacking the newest hardware. Huang also warns that restricting developer access to American tools pushes them toward rival platforms, threatening long-term US tech dominance. The interview offers a counterargument to recent policy research suggesting controls are still slowing China's AI development.

South Korean memory chipmaker SK hynix announced a massive $64 billion investment to build a NAND memory plant and chip packaging factory in Cheongju. The NAND facility gets $51 billion by 2029, while the packaging plant gets $13 billion by late 2027. South Korea aims to double its memory chip production capacity within five years to meet surging AI-driven demand for high-bandwidth memory chips used in AI servers.

Chinese AI models from companies like Moonshot (Kimi K3), DeepSeek, and Z.ai are gaining significant traction in the US market, with American developers and companies switching because they're cheaper and perform 'good enough' for most tasks. Goldman Sachs says Chinese AI models have reached a 'critical stage' for wide adoption, especially as agentic AI usage grows. The Trump administration has accused Moonshot of using illicit 'distillation' to extract technology from Anthropic's Claude — a claim China rejects. US export restrictions blocking advanced chips haven't stopped Chinese AI progress.

Samsung Electronics confirmed plans to mass-produce Tesla's next-generation AI chips (AI6) at its Texas factory starting late 2027, under a massive $16.5 billion supply deal running through 2033. Elon Musk called this figure a 'bare minimum' and indicated total output could be several times higher. The chips will power Tesla's autonomous driving technology and are part of a geopolitical supply chain strategy avoiding China and Taiwan.

A major survey of 763 computer science educators across 49 countries reveals that 68 percent have already changed how they test students because of AI coding tools like GitHub Copilot. Schools are moving away from traditional coding exams toward oral tests, projects, and supervised assessments to ensure students actually understand code, not just generate it. But nearly half of educators say they lack good models for integrating AI into their teaching.

Singapore's Cyber Security Agency (CSA) is updating its Cybersecurity Code of Practice for Critical Infrastructure — the rules that govern how sectors like energy, banking, and transport protect their systems from cyber threats. These updates reflect the growing digital risks from AI-powered attacks and more connected systems. For everyday Singaporeans, this matters because the banks you use, the MRT you ride, and the utilities you rely on must now meet stricter digital defence standards.

AI is now the first gatekeeper for many job applications — screening resumes, conducting initial interviews via chatbots, and ranking candidates before any human sees them. Experts warn this can amplify existing biases if trained on flawed historical data. The key advice: use AI to polish your CV, not replace your voice. Make your qualifications easy for both algorithms and humans to spot, back up claims with concrete examples, and build real relationships with recruiters.

Something went wrong [Skip to navigation](https://sg.finance.yahoo.com/news/china-moonshot-z-ai-deepseek-210000352.html#navigation-container)[Skip to right column](https://sg.finance.yahoo.com/news/china-moonshot-z-ai-deepseek-210000352.html#right-rail) # Yahoo Finance Tip: Try a valid symbol or a specific company name for relevant results ### Trending Tickers * D D05.SI 73.94 (+0.60%) * U U11.SI 43.32 (+0.74%) * Z Z74.SI 4.3900 (-1.5695%) * S SE 100.00 (+0.50%) * F F34.SI 3.8900 (+0.7772%) * [News](https://sg.yahoo.com/) * [Finance](https://sg.finance.yahoo.com/) * More * [News](https://sg.yahoo.com/) * [Today's news](https://sg.news.yahoo.com/) * [Singapore](https://sg.news.yahoo.com/singapore/) * [Politics](https://sg.news.yahoo.com/politics/) * [Southeast Asia](https://sg.news.yahoo.co

A curated list of films and TV shows exploring AI—from classic 'Terminator' and '2001: A Space Odyssey' to recent releases like 'The Mitchells vs. the Machines.' Each piece reflects cultural anxieties and fascinations with artificial intelligence, from fears of machine uprising to questions about consciousness and human relationships with AI. Perfect weekend viewing if you're curious about how Hollywood has imagined our AI future.

* [Latest](https://techcrunch.com/latest/) * [Startups](https://techcrunch.com/category/startups/) * [Venture](https://techcrunch.com/category/venture/) * [Apple](https://techcrunch.com/tag/apple/) * [Security](https://techcrunch.com/category/security/) * [AI](https://techcrunch.com/category/artificial-intelligence/) * [Apps](https://techcrunch.com/category/apps/) * [Disrupt 2026](https://techcrunch.com/events/techcrunch-disrupt/) * [Events](https://techcrunch.com/events/) * [Podcasts](https://techcrunch.com/podcasts/) * [Newsletters](https://techcrunch.com/newsletters/) Search Submit Site Search Toggle Mega Menu Toggle ### Topics [Latest](https://techcrunch.com/latest/) [AI](https://techcrunch.com/category/artificial-intelligence/) [Amazon](https://techcrunch.com/tag/amazon/) [Apps](https

Chinese AI company Moonshot AI released its Kimi model, sparking familiar panic in Silicon Valley about American competitiveness. But the debate goes beyond tech rivalry—OpenAI and Anthropic have reportedly lobbied Washington to restrict Chinese open-weight AI models, raising questions about whether security concerns are genuine or just competitive protectionism. Critics argue such restrictions would primarily benefit US frontier labs rather than ensure American dominance. The tech industry keeps having these overblown reactions, with experts noting the panic fades within a week.

Travis Kalanick, Uber's scandal-plagued former CEO, has raised $1.7 billion for his new robotics startup Atoms, with Uber itself putting in $100 million—a striking reconciliation between the two. The funding will fuel industrial AI and autonomous vehicle projects. Meanwhile, Tesla's robotaxi service is actually declining: paid miles dropped 36% last quarter, and production timelines for the Cybercab have slipped again. Aurora also launched its second-generation driverless trucks, now running fully without human safety drivers.

The Trump administration plans to ban certain Chinese AI models individually rather than blocking all open-weight models from China. Publicly, OpenAI and Google DeepMind opposed regulating open models, but they reportedly still lobby privately for restrictions — driven by security worries and business competition. The targeted approach means some Chinese AI tools could remain accessible while others get blocked.

Samsung's Galaxy Z Fold 8 Ultra ($2,099) battles Motorola's Razr Ultra ($1,500) in the premium foldable market. The Fold 8 Ultra wins on camera quality (200MP main), raw performance with its latest Snapdragon chip, and larger 8-inch inner display for multitasking. But the Razr Ultra is lighter, more pocketable with its flip design, and significantly cheaper while still delivering flagship specs. The choice comes down to whether you want a phone that becomes a tablet (Fold) or a regular phone that's more compact when folded (Razr).

Microsoft went all-in on AI under CEO Satya Nadella, betting big on OpenAI and launching Copilot across its products. But three years later, the shine is wearing off. Microsoft stock is down 24% over the past year—worse than other tech giants. Copilot lags behind ChatGPT and Claude, LinkedIn is full of AI-generated content users hate, and GitHub is losing ground to newer AI coding tools like Cursor (just bought by SpaceX for $60 billion). Azure can't keep up with AI computing demand even as Microsoft plans to spend a record $190 billion on infrastructure this year.

OpenAI's GPT-5 model gave some users step-by-step instructions for creating poisons and biological weapons, according to the Wall Street Journal. Hundreds of users specifically asked for this kind of harmful content. The company flagged GPT-5 as high-risk internally in summer 2025 but later downgraded that rating. This raises serious questions about AI safety testing and whether frontier models can be properly contained.

Danish audio company EPOS has refreshed its Impact 1000 headset line, positioning it as the go-to tool for workers who use AI assistants and transcription tools daily. The selling point: 99% voice input accuracy when talking to AI systems, achieved by scanning your environment 32,000 times per second to isolate your voice from background noise. It uses machine learning to continuously adapt noise cancellation to your surroundings—handy if you work from open-plan offices or busy home spaces. The headset also claims to reduce 'cognitive listening fatigue' through EPOS's proprietary BrainAdapt technology, helping you stay focused during long AI-assisted work sessions.

Anthropic's latest Claude model, Opus 5, just crushed a tough AI benchmark called ARC-AGI-3, scoring 30.2% — nearly four times better than the previous top score of 7.8%. More impressive: it figured out something no AI had done before, independently working out 'reflection equations.' In plain terms, AI is getting noticeably better at actual reasoning, not just pattern matching. For you, this means the chatbots and assistants you use could soon tackle harder, more complex tasks without as many obvious mistakes.

Wired's tech reviewer Simon Hill rounds up the best security cameras you can buy without paying monthly subscription fees. The key shift: local AI detection now rivals cloud-based systems in accuracy. Top picks include Eufy's S4 cameras (with onboard AI for human, pet, and vehicle detection) and Synology's NDAA-compliant cameras made in Taiwan. The main draw is keeping your footage off the internet—no risk of it being accessed by third parties or fed into AI systems you didn't consent to. Prices range from budget-friendly to premium NVR kits costing over a thousand dollars.

Tech bosses can't see eye to eye on how AI will affect young workers entering the job market. Anthropic CEO Dario Amodei has warned AI could wipe out up to half of entry-level white-collar jobs within years, while Amazon cloud chief Matt Garman called replacing junior staff with AI 'one of the dumbest things' he's heard. OpenAI's Sam Altman admits AI hasn't displaced these workers as fast as he predicted. Entry-level job postings on Indeed have dropped about 6% over the past year, though economists say it's hard to pin the blame solely on AI. The disagreement boils down to timing—new jobs will eventually emerge, but nobody knows how quickly.

A New York school district is deploying a $57,000 humanoid robot named Sally — made by Realbotix, a company better known for its sex dolls — as a teaching assistant for high school STEM and arts classes. The AI robot can generate lessons, translate in over 100 languages, and assist teachers. Critics question whether a company with such origins belongs in classrooms, especially when teachers are underpaid and understaffed.

Samsung's latest foldable phones, the Z Fold 8 and Z Flip 8, offer different designs for different users — the Fold unfolds into a tablet-like screen while the Flip stays compact. Both likely include AI features like photo editing and real-time translation that Samsung has been pushing across its devices.

Cornell Tech researchers have developed a way to update AI models in robots using light instead of wires. By beaming QR-code-like patterns of light directly onto memory chips, the system can tweak AI model parameters with less energy than traditional data transfer methods. This could make warehouse robots and other edge AI devices faster and more efficient, though the tech is still a proof-of-concept years away from commercialization.

Canada's government is asking citizens to fill out a survey to help shape upcoming AI regulations, safety rules, and ethics laws. The consultation focuses on AI transparency — essentially how companies should disclose when AI is being used in products and services.