Singapore AI News & Daily Briefing

Elastic, the company behind the popular Elasticsearch search tool, is buying Deductive AI for up to $85 million. Deductive uses AI to automatically find and fix bugs in software—potentially cutting the time engineers spend on system outages by up to 90%. The startup was only founded in 2023 and had about $1 million in annual revenue. This fits a wider trend of established tech companies buying AI startups to add automation features to their existing products.

The US government is raising alarms that ASML—the Dutch company with a monopoly on extreme ultraviolet (EUV) chip-making machines—may have illegally shipped one to China. These $200 million machines are the only tools on Earth capable of printing the most advanced chip patterns, and every cutting-edge processor (Nvidia GPUs, Apple chips) depends on them. ASML denies any EUV machine exists in China, but senior US officials claim they have evidence. If true, it would be one of the biggest breaches of US-led export controls designed to limit China's AI capabilities. No evidence has been made public yet.

Allbirds, the shoe company once beloved for its comfy sneakers, has pivoted entirely to AI infrastructure under new CEO Nadia Carlsten. Now renamed Smartbird, the company sold its shoe business for $43 million and raised $100 million more from investors hoping to ride the AI wave. Carlsten, a former AWS executive, plans to target companies that want direct control over their AI servers rather than using public cloud giants. It's a classic Silicon Valley pivot story, though whether it will work remains unclear.

Allbirds, the struggling shoe company, has officially pivoted to AI infrastructure and rebranded as Smartbird. The new CEO Nadia Carlsten, a former AWS executive, is building a team from scratch after Allbirds sold its shoe business for $43 million and raised $100 million from stock investors. Smartbird aims to serve companies that need direct control over their AI compute infrastructure—typically for data sovereignty or political reasons—rather than competing directly with hyperscalers. The company targets pharma, energy, financial, and public sector clients. Carlsten expects compute clusters deployed for customers by year-end.

A UN report warns that AI data centers guzzle far more than electricity — by 2030, they could consume 945 terawatt-hours of power AND 9.3 trillion liters of water annually. Generating a single AI image uses about 1,450 times more energy than a basic text query. The catch? As AI gets more efficient, people use it more — so overall impact keeps rising despite better technology. Singapore has felt this pressure firsthand: data centers already account for over 7% of the country's electricity use, prompting a government moratorium on new builds between 2019 and 2022. Today, new data center capacity is awarded selectively to operators meeting strict efficiency and green energy standards.

India's data center pipeline has reached 8.3 gigawatts of capacity, signaling massive growth in digital infrastructure across the region. This expansion is being driven by increasing cloud adoption, AI workloads, and digital services demand. The growth has implications for power infrastructure, real estate, and regional tech competitiveness.

A new survey finds 85% of US desk workers are learning AI tools on their own time outside of work. Tech employees report spending 10 to 20 hours per week experimenting with AI tools like coding assistants and AI agents, often paying out of pocket for subscriptions. Workers say the pressure is real: companies are cutting jobs while investing in AI, and those who don't keep up risk being left behind. One worker even built an AI agent to negotiate hotel rates for them on evenings and weekends.

South Korea's top search engine Naver says its new AI-powered search tab attracted over 3 million users in just one month, helping the company gain ground against competitors. This shows traditional search engines are fighting back against AI-native rivals by rolling out smart features. For everyday users, it means search engines are getting smarter and more competitive — which could mean better results for you.

TechInAsia's premium landscape report maps out Israel's major AI companies, top investors, and funding trends. It aims to give professionals a clearer picture of who dominates Israel's AI scene and where money is flowing.

Meta and Reliance Industries are building a massive AI data center in Jamnagar, Gujarat—phase one alone is 168 megawatts, far larger than India's typical 60 MW data centers. Meta will lease the facility and cover energy and water costs. It's being powered by nearly 1 gigawatt of new clean energy contracts in India and uses desalinated seawater for cooling to address water scarcity concerns. The deal extends a partnership that already includes Meta's $5.7 billion investment in Jio Platforms in 2020. This signals the massive power and water demands that come with building AI infrastructure.

This is a personal Father's Day essay about the real-world need for AI tools in family caregiving. The author, who works in AI and robotics, argues that the most useful home AI isn't flashy humanoid robots but quiet, practical tools like passive sensors, fall detection, routine monitoring, and care coordination software. He notes that 63 million Americans are now caregivers, and that unpaid family caregiving is worth over $1 trillion annually. The piece is more reflective than news-driven—it advocates for AI that feels like "being watched over, not watched."

Nvidia has officially qualified all three major memory chipmakers — Samsung, SK hynix, and Micron — to supply the next-generation HBM4 memory for its upcoming Vera Rubin AI platform. Jensen Huang announced this in Seoul, noting all three are now in production and competing for Nvidia's orders amid ongoing AI memory shortages. SK hynix currently dominates the HBM market with a 62% share. HBM4 doubles data transfer speeds compared to the current HBM3E standard, which is critical for running large AI models. Nvidia is also opening a new R&D center in Korea.

A top AI researcher from Bridgewater—one of the world's biggest hedge funds—is joining Google DeepMind as chief strategy officer. Jasjeet Sekhon previously built Bridgewater's AI research lab and has academic roots at Harvard, Yale, and UC Berkeley. The move highlights how the massive $650 billion AI investment wave is pulling top quantitative talent from finance into AI labs, potentially reshaping how complex financial models inform real-world AI development.

LG CNS, a South Korean IT firm, is carving out a niche as the middleman between robot makers and factories that need automation. They don't build robots or AI models themselves—instead, they advise manufacturers on which robots to buy, train those robots for specific factory tasks, and manage day-to-day operations. The company is developing specialized "Robot Foundation Models" that act like an AI brain for robots, helping factories deploy automation faster. As robot prices fall, this integration role—getting robots actually working on factory floors—is becoming valuable. The humanoid robot market could reach $38 billion by 2035.

Amazon is investigating three engineers who publicly criticized the company's massive AI data center expansion at Seattle City Council meetings. The engineers called for government regulation of data centers and called out tech's "all-costs-justified AI build out." Amazon could potentially terminate them. This raises questions about whether tech workers can speak up about AI's environmental and social impacts without risking their jobs.

OpenAI CEO Sam Altman told Stanford students that many AI researchers held the field back for years by doubting that simply scaling up AI models would work. He pointed to OpenAI's recent breakthrough proving a mathematical conjecture as proof the skeptics were wrong. Altman's comments are a defense of the 'bigger is better' approach that shaped the current AI boom — and a jab at researchers who dismissed it.

The Trump administration forced AI company Anthropic to take its newest models offline, citing unexplained "national security concerns." Amazon reportedly alerted the White House after its researchers found ways around the models' safety guardrails. Cybersecurity experts are pushing back, warning that pulling these tools from US network defenders is dangerous. Meanwhile, some observers say Anthropic might actually benefit from the "bad boy" image this creates. This shows how geopolitics increasingly shapes which AI tools you can access.

One of AI's founding figures is calling out the industry's sustainability crisis — xAI has lost $2.5 billion in a single quarter and is now renting out its data centers just to stay afloat. Yann LeCun argues this isn't just xAI's problem: the entire model of pouring billions into AI without clear profits may be heading for a collapse. He believes the current approach of building systems that just generate text has limits, and that companies like OpenAI and Anthropic will eventually be forced to raise prices or face a reckoning.

A US data company has released over half a million hours of labeled Ukraine war drone footage for AI training — the largest real-world combat dataset ever assembled. Military buyers in the US, Ukraine, and NATO countries can use this footage to train autonomous drones and targeting systems without needing expensive simulations. The footage captures real weather, terrain, and combat scenarios that algorithms struggle to replicate artificially.

Major companies like Amazon, Walmart and Uber are now limiting how much they use AI because running it costs far more than expected. Despite years of aggressive adoption, firms are discovering that AI bills add up fast and are now urging employees to cut back — a sign that the technology's promises may be harder to profit from than hoped.

A UC Berkeley study analyzing over 500,000 student grades found that writing and coding courses saw grade improvements after ChatGPT launched—but mainly in homework, not exams. This suggests students are outsourcing work to AI rather than actually learning. The pattern mirrors what happened with calculators in math: grades went up, but real understanding may have declined.

A major new study from the Brookings Institution covering 50 countries finds that AI's risks in schools outweigh its benefits for children. The report warns that students are offloading their thinking to AI tools, potentially harming their cognitive development — essentially getting answers without learning how to think. While AI can help with reading and writing practice, researchers say it cannot replace the guidance of a real teacher.

Apple's iOS 27 brings practical AI features woven into everyday apps — from splitting restaurant bills by scanning receipts to auto-updating compromised passwords, organizing Safari tabs, and reducing smart home notification clutter. These aren't flashy AI chatbots; they're small conveniences that make your iPhone feel smarter without you noticing. Features roll out in beta now, public release later this year.

AI company Anthropic is rolling out mandatory identity verification using government-issued photo IDs starting July 8, 2026. Users accessing certain features will need to submit ID through a third-party partner called Persona Identities. Accounts can be banned if verification fails. Anthropic says the move is needed to prevent abuse and comply with legal obligations, but it raises privacy concerns about handing over personal documents to AI companies.

Brands are secretly using fake AI-generated people to pretend they're real customers raving about products on Instagram and social media. An investigation found companies hiring creators to make hyperrealistic digital 'influencers' — sometimes under NDAs so they can't admit it's fake. UK regulators have no rules forcing disclosure, though the EU's new AI Act will require labeling from August. Research shows 70% of people can't tell real from fake AI videos anymore. The practical impact: you might be watching a fake 'bride' or 'customer' pushing products, and there's currently no law requiring brands to admit it.

Amazon Web Services (AWS) just launched two new services to fix common problems with AI coding assistants. The first, Continuum, automatically finds and fixes security vulnerabilities in code. The second, Context, gives AI agents a better understanding of your company's specific data and rules. The problem AWS is solving: AI tools that write code fast but make costly mistakes because they don't understand your business. These services target companies that want to deploy AI agents for software development without the usual headaches.

Americans are increasingly negative about AI despite using it more — only 16% believe it will benefit society while 49% already use chatbots daily. Gen Z leads in both usage (66%) and skepticism (48%), suggesting widespread adoption hasn't changed minds. Many feel forced to use AI at work even as they distrust it.

Wired shares 28 practical tips to get better results from ChatGPT — from asking the AI to critique your ideas like a curious 10-year-old would, to using the '80-20 rule' to learn topics faster. These prompting tricks can make everyday tasks like writing emails, learning new subjects, or planning trips much smoother.

I didn't think that this would work, but the EverExceed 5V solar panel has done wonders for my security cameras.

Your phone holds a lot of personal information, from where you go to who you talk to every day. Here's how to limit who can access your data and when.

Unveiled at CES, the 14th-gen Lenovo ThinkPad X1 Carbon Aura Edition features a redesigned double-sided motherboard and modular components.

The Verge traces how iRobot's Roomba, once a笨拙的碰撞式吸尘器, sparked the consumer robotics wave. The podcast explores how early Roombas used simple bump-and-navigation algorithms that seemed crude but laid groundwork for today's smart home robots. For everyday readers, it's a reminder that the smart gadgets in your home likely trace their roots to experiments that looked silly at first.

A Reddit discussion video exploring data collection practices in AI systems — examining how user data flows into AI training pipelines and the transparency gaps around what companies actually do with that information.

Nobel laureate John Jumper is leaving DeepMind for rival Anthropic. Jumper isn't the only big name leaving Google DeepMind.

Picture this: a team of camera-clad cleaners and a private chef to boot, all wired with high-tech recording apparatus show up at your home. You are not part of a reality TV show, and have not woken up in an Aldous Huxley or Margaret Atwood novel. Instead, you are a resident of New York City, where AI companies are sending free cooking and cleaning staff straight to people's doors. But, there is a catch: this AI company is gathering data to train the next generation of cooking and cleaning robots, and every inch of your apartment is now being recorded. The initiative, dubbed Shift by AI firm Micro AGI, is part of a growing number of companies developing the next generation of autonomous robots, which tech bosses hope will be able to do everything from the washing up to serving as live-in pe

A top AI researcher from Bridgewater—one of the world's biggest hedge funds—is joining Google DeepMind as chief strategy officer. Jasjeet Sekhon previously built Bridgewater's AI research lab and has academic roots at Harvard, Yale, and UC Berkeley. The move highlights how the massive $650 billion AI investment wave is pulling top quantitative talent from finance into AI labs, potentially reshaping how complex financial models inform real-world AI development.

India's data center pipeline has reached 8.3 gigawatts of capacity, signaling massive growth in digital infrastructure across the region. This expansion is being driven by increasing cloud adoption, AI workloads, and digital services demand. The growth has implications for power infrastructure, real estate, and regional tech competitiveness.

South Korea's top search engine Naver says its new AI-powered search tab attracted over 3 million users in just one month, helping the company gain ground against competitors. This shows traditional search engines are fighting back against AI-native rivals by rolling out smart features. For everyday users, it means search engines are getting smarter and more competitive — which could mean better results for you.

Meta and Reliance Industries are building a massive AI data center in Jamnagar, Gujarat—phase one alone is 168 megawatts, far larger than India's typical 60 MW data centers. Meta will lease the facility and cover energy and water costs. It's being powered by nearly 1 gigawatt of new clean energy contracts in India and uses desalinated seawater for cooling to address water scarcity concerns. The deal extends a partnership that already includes Meta's $5.7 billion investment in Jio Platforms in 2020. This signals the massive power and water demands that come with building AI infrastructure.

LG CNS, a South Korean IT firm, is carving out a niche as the middleman between robot makers and factories that need automation. They don't build robots or AI models themselves—instead, they advise manufacturers on which robots to buy, train those robots for specific factory tasks, and manage day-to-day operations. The company is developing specialized "Robot Foundation Models" that act like an AI brain for robots, helping factories deploy automation faster. As robot prices fall, this integration role—getting robots actually working on factory floors—is becoming valuable. The humanoid robot market could reach $38 billion by 2035.

Nvidia has officially qualified all three major memory chipmakers — Samsung, SK hynix, and Micron — to supply the next-generation HBM4 memory for its upcoming Vera Rubin AI platform. Jensen Huang announced this in Seoul, noting all three are now in production and competing for Nvidia's orders amid ongoing AI memory shortages. SK hynix currently dominates the HBM market with a 62% share. HBM4 doubles data transfer speeds compared to the current HBM3E standard, which is critical for running large AI models. Nvidia is also opening a new R&D center in Korea.

Singapore's Health Promotion Board tested giving pre-diabetic and overweight people continuous glucose monitors and digital coaching via an app. The 12-week pilot with 1,700 people showed promising results—participants kept their blood sugar in the healthy range an extra 43 minutes daily, with heavier cases improving by 2.6 hours. Nearly 80% saw their BMI drop. HPB now uses AI to personalize health nudges for the nearly 900,000 monthly users of its Healthy 365 app, analyzing user habits to send reminders at the right time.

Hong Kong now has a convenience store run entirely by one humanoid robot named 'Xiao Gai' built by Beijing-based Galbot. The 24-hour pop-up at Hung Hom uses the robot to stock shelves, serve customers, and handle checkouts. The operator expects foot traffic to jump 40% and plans 100 more robot-staffed stores across ten cities. It's another sign robots are moving from novelty into real retail jobs.

A new Chinese AI model called GLM-5.2 is impressing Silicon Valley tech leaders with its coding capabilities. Made by z.AI, it runs on a 1 million token context window, letting it handle massive coding tasks. Open-source like DeepSeek, anyone can download and run it themselves—unlike closed US models from OpenAI or Anthropic. Vercel CEO called it "genuinely impressed," while a former Meta/Google/Microsoft VP said it's the first open model he'd use daily. The US-China AI race is heating up as Chinese companies push out cheaper, capable open-source models that could grab market share from American giants.

Such clients can now get ‘something close to private-banking quality’ from the technology Published Mon, Jun 22, 2026 · 07:00 AM WEALTH managers keen to stay relevant in the age of artificial intelligence may soon find that clients with a mere US$1 million in liquid assets are no longer worth spending human hours on. “The mass-affluent client now gets something close to private-banking quality from AI,” said Debasish Patnaik, senior partner at McKinsey. That not only “strips the value from the adviser whose role was standardised advice”, but it also means “the kind of person [hired into wealth management](https://www.businesstimes.com.sg/companies-markets/stanchart-singapore-hire-more-wealth-bankers-amid-plans-cut-over-7000-jobs-globally) changes fundamentally”, he said in an interview. Th

German startup Eternal.ag is deploying autonomous robots in greenhouses to solve a critical problem: labor shortages in harsh farming conditions. The company's harvesting robot works 22 hours a day (charging only 3 hours), replacing six human operators that would cost $250,000 annually in developed economies. The robot navigates GPS-denied environments using cameras, LiDAR, and ultrasonic sensors, cuts plant stems, and places produce in baskets without damaging fruit. It uses AI training through digital twins and physical navigation. Business model is Robots-as-a-Service, earning revenue per cut produce. Currently optimized for tomatoes, with plans to expand to cucumbers and horticulture. CEA (Controlled Environment Agriculture) is a $103 billion market expected to double by 2030.

A slightly longer version of Model Y appears on track for the U.S. market. The Model Y L (long wheelbase) 6-seater will go on sale as soon as Tesla readies U.S. production, as [reported](https://autos.yahoo.com/new-vehicles-and-reviews/articles/tesla-bigger-longer-model-y-170013327.html) here in April. More recent [reports](https://www.notateslaapp.com/news/4326/exclusive-tesla-model-y-l-launching-this-fall-in-north-america) and reviews of the Model Y L are setting the stage for production in the U.S. “China, Australia, and India are supplied by the factory in China, which will not supply vehicles to the U.S.,” Sam Fiorani, VP, Global Vehicle Forecasting, at AutoForecast Solutions, told me at the end of April. “Production of the Model Y L is expected to begin in the U.S. in September, whic

A new survey finds 85% of US desk workers are learning AI tools on their own time outside of work. Tech employees report spending 10 to 20 hours per week experimenting with AI tools like coding assistants and AI agents, often paying out of pocket for subscriptions. Workers say the pressure is real: companies are cutting jobs while investing in AI, and those who don't keep up risk being left behind. One worker even built an AI agent to negotiate hotel rates for them on evenings and weekends.

TechInAsia's premium landscape report maps out Israel's major AI companies, top investors, and funding trends. It aims to give professionals a clearer picture of who dominates Israel's AI scene and where money is flowing.

This is a personal Father's Day essay about the real-world need for AI tools in family caregiving. The author, who works in AI and robotics, argues that the most useful home AI isn't flashy humanoid robots but quiet, practical tools like passive sensors, fall detection, routine monitoring, and care coordination software. He notes that 63 million Americans are now caregivers, and that unpaid family caregiving is worth over $1 trillion annually. The piece is more reflective than news-driven—it advocates for AI that feels like "being watched over, not watched."

Amazon is investigating three engineers who publicly criticized the company's massive AI data center expansion at Seattle City Council meetings. The engineers called for government regulation of data centers and called out tech's "all-costs-justified AI build out." Amazon could potentially terminate them. This raises questions about whether tech workers can speak up about AI's environmental and social impacts without risking their jobs.

A UN report warns that AI data centers guzzle far more than electricity — by 2030, they could consume 945 terawatt-hours of power AND 9.3 trillion liters of water annually. Generating a single AI image uses about 1,450 times more energy than a basic text query. The catch? As AI gets more efficient, people use it more — so overall impact keeps rising despite better technology. Singapore has felt this pressure firsthand: data centers already account for over 7% of the country's electricity use, prompting a government moratorium on new builds between 2019 and 2022. Today, new data center capacity is awarded selectively to operators meeting strict efficiency and green energy standards.

Hong Kong now has a convenience store run entirely by one humanoid robot named 'Xiao Gai' built by Beijing-based Galbot. The 24-hour pop-up at Hung Hom uses the robot to stock shelves, serve customers, and handle checkouts. The operator expects foot traffic to jump 40% and plans 100 more robot-staffed stores across ten cities. It's another sign robots are moving from novelty into real retail jobs.

A new Chinese AI model called GLM-5.2 is impressing Silicon Valley tech leaders with its coding capabilities. Made by z.AI, it runs on a 1 million token context window, letting it handle massive coding tasks. Open-source like DeepSeek, anyone can download and run it themselves—unlike closed US models from OpenAI or Anthropic. Vercel CEO called it "genuinely impressed," while a former Meta/Google/Microsoft VP said it's the first open model he'd use daily. The US-China AI race is heating up as Chinese companies push out cheaper, capable open-source models that could grab market share from American giants.

Singapore's Health Promotion Board tested giving pre-diabetic and overweight people continuous glucose monitors and digital coaching via an app. The 12-week pilot with 1,700 people showed promising results—participants kept their blood sugar in the healthy range an extra 43 minutes daily, with heavier cases improving by 2.6 hours. Nearly 80% saw their BMI drop. HPB now uses AI to personalize health nudges for the nearly 900,000 monthly users of its Healthy 365 app, analyzing user habits to send reminders at the right time.

German startup Eternal.ag is deploying autonomous robots in greenhouses to solve a critical problem: labor shortages in harsh farming conditions. The company's harvesting robot works 22 hours a day (charging only 3 hours), replacing six human operators that would cost $250,000 annually in developed economies. The robot navigates GPS-denied environments using cameras, LiDAR, and ultrasonic sensors, cuts plant stems, and places produce in baskets without damaging fruit. It uses AI training through digital twins and physical navigation. Business model is Robots-as-a-Service, earning revenue per cut produce. Currently optimized for tomatoes, with plans to expand to cucumbers and horticulture. CEA (Controlled Environment Agriculture) is a $103 billion market expected to double by 2030.