Singapore AI News & Daily Briefing

A brief report says Southeast Asia is well positioned to grow AI through better digital infrastructure, workforce skills, governance trust, data flows, and regional teamwork. Singapore appears as a key example of this approach.

Micron Technology, the biggest US memory chip maker, saw its shares jump 16% after forecasting Q4 revenue of $50 billion—blowing past analyst estimates by over $7 billion. AI data centers are gobbling up memory faster than companies can make it. Shortages could last beyond 2027, keeping prices high for computers, phones, and cars. Micron is expanding its Singapore factory to help meet demand.

Google DeepMind and partners are pouring $10 million into research on how millions of AI agents can safely interact without causing unexpected problems. They're asking researchers worldwide to study risks that pop up when independent AI systems talk to each another across networks—things like unpredictable economic activity or security issues. The goal is to build safety frameworks before the multi-agent world fully arrives.

OpenAI shared data showing its Codex coding tool has transformed how its own employees work. Non-technical staff like lawyers and recruiters now use it for 85% of their AI tasks. Among all users, non-developer adoption grew 137 times faster than developers. Some workers now use AI to do jobs outside their actual job descriptions. At the extreme end, power users run over 60 hours of AI work per day.

Google just released Gemma 4 12B, a slimmed-down AI model that handles text, images, and audio without bulky extra processors. The kicker: it runs on a regular laptop with 16GB RAM—no cloud needed. It's open-source under Apache 2.0, so developers can download and use it freely. The model has already been downloaded 150 million times.

Google DeepMind released its AI Control Roadmap—a security framework treating AI agents as potential 'insider threats' even when aligned. The system monitors agent reasoning, blocks harmful actions before execution, and uses other AI as supervisors. DeepMind analyzed 1 million coding agent tasks and found most flagged issues stem from agent overeagerness rather than malicious intent. The company also published a policy framework for industry-wide adoption.

Notion is shutting down its AI-powered email client on September 22, 2026, admitting that most users skip their inbox entirely and let AI agents handle email instead. The app was originally built using technology from Skiff, an encrypted email startup Notion acquired in 2024. Users need to export drafts and scheduled emails before the shutdown date, though Gmail history will remain intact.

Researchers at the University of Arizona built an acoustic synapse using sound waves that can mimic brain neurons more efficiently than electronic neuromorphic chips. The device uses 'phi-bits' to encode multiple values simultaneously, consuming at most one-tenth the power of current state-of-the-art neuromorphic hardware. It could enable smaller, more energy-efficient AI chips for pattern recognition and sensory processing.

Capital One hired a Chief Scientist to build original AI research capabilities in financial services, treating AI as a scientific discipline rather than just technology to deploy. The bank has launched what it calls the first fully agentic AI customer service experience built in-house—a car buying tool that takes actions on behalf of customers. Capital One leads all financial institutions in AI patents.

The US government has lifted export restrictions on Anthropic's most powerful AI models, Mythos and Fable, which had blocked foreign access since June. Anthropic agreed to work with the US government on security protocols for future releases. The restriction reversal came as Asian AI companies started releasing competing models of similar capability, creating pressure on American AI competitiveness. For everyday users and businesses, this means access to these cutting-edge AI tools is restored.

Renewable energy firm Ib vogt and data center operator AirTrunk have launched Malaysia's first corporate green power agreement specifically for data centers. The 30 MW solar project will supply clean energy to AirTrunk's Malaysian operations, marking a step toward sustainable AI infrastructure in Southeast Asia.

Your next gadget will cost more, and AI is partly to blame. The worldwide rush to build AI data centers has created a severe memory chip shortage—companies like Apple, Samsung, Microsoft, Sony, and Nintendo are all raising prices on laptops, tablets, smartphones, and gaming consoles. The reason: the same memory chips (DRAM) used in your devices are also needed for AI servers. Data center operators are outbidding consumers for these components, squeezing supply. Micron says new factories will take years to come online, so expect higher prices to stick around.

AI Singapore, the national agency driving the country's AI strategy, has new leadership in place to push forward its ambitions. The timing aligns with Singapore's broader Smart Nation goals, where AI plays a central role in digital government, economy, and society. For Singaporeans, this leadership change could shape how quickly and effectively AI is adopted across public services, jobs, and daily life.

US industrial giant Wesco International has bought Singapore's Newark Engineering for $136 million. Newark designs and installs cooling systems for data centers—the massive warehouses that house servers powering AI and cloud services. With AI demand surging, data center operators are scrambling to manage heat output. This acquisition lets Wesco tap into Southeast Asia's fast-growing data center market, where Singapore, Malaysia, and Indonesia are all building new facilities. For everyday users, this means more data center capacity could eventually mean faster, more reliable AI services.

Banks are saying AI applications that interact with the real world—think robots in factories, self-driving vehicles, and industrial automation—could reshape how telecom companies earn money. Instead of just selling data plans to phone users, telecoms could charge businesses premium rates for reliable, fast connections needed to run these machines. Maybank predicts Physical AI traffic could make up 32% of all mobile data by 2031, potentially lifting telecom revenue growth from low single digits to high single digits. This is a big shift from previous AI booms that mostly benefited data centers and chipmakers.

A new survey of over 5,000 workers across six Southeast Asian countries—including Singapore—finds that 73 percent feel their bosses are out of touch with actual digital needs at work. Companies are investing heavily in IT and finance systems but neglecting employee experience and training. Sixty-two percent worry AI will make their jobs obsolete, yet 88 percent still want AI to handle routine tasks. The report warns that poor adoption strategies are turning AI from an opportunity into a threat.

Singapore's Ministry of Social and Family Development is spending $15 million over three years to bring AI into the social service sector. The ministry is partnering with NCS and ST Engineering to develop tools that identify families at risk earlier and help social workers manage heavy caseloads. One tool called Scribe cuts documentation time by at least half by transcribing conversations — even in Singlish and Cantonese — into case notes automatically. Another flags blind spots in assessments. Social workers say this frees them to actually connect with clients instead of drowning in paperwork. Minister Masagos Zulkifli stressed AI won't replace human judgment — it's meant to help professionals reach families faster and more effectively.

LinqAlpha, a startup building AI tools for institutional investors, has raised $22 million in Series A funding from investors including East Ventures and other SEA players. The company helps Wall Street firms and hedge funds deploy AI agents that analyze market signals before they hit consensus. Its clients manage over $5 trillion in assets. The funding will fuel expansion across Asia.

NTU Singapore is launching a new part-time graduate program to train chip designers, developed with the Singapore Semiconductor Industry Association and the Economic Development Board. Students will learn to design digital chips using industry-standard tools from companies like Synopsys and Cadence, getting hands-on experience from concept to fabrication-ready designs. The program targets working engineers and graduates looking to enter the semiconductor field, with courses that can stack toward a full Master's degree. It's designed to feed talent into fabless companies like AMD, Broadcom, and MediaTek that already run design operations in Singapore.

Taiwanese prosecutors have detained two Super Micro Computer employees and placed four on administrative leave, following raids on the company's local offices as part of an investigation into illegal exports of Nvidia AI chips to China. The suspects allegedly falsified documents to ship servers containing coveted AI processors through intermediaries like Albatron Technology and Chief Telecom. Taiwan is now considering criminalizing AI chip exports to China outright, since current law doesn't explicitly prohibit them — even as the US pressures allies to cut off Beijing's access to advanced AI technology.

Singapore's government is rolling out new tools to help retail small businesses adopt AI. Enterprise Singapore and the Ministry of Trade and Industry launched a refreshed Retail Industry Digital Plan and a new two-year incubator programme called Retail Accelerator at L^IFE. The updated plan focuses on practical uses like demand forecasting, AI-powered marketing, and staff training tools. A local florist cut cashier workloads and marketing time using AI, while a jeweller built a voice-activated inventory system. Sixteen retailers are joining the accelerator, getting subsidised space at SCAPE, live-streaming studios, and mentorship.

Southeast Asia's startup scene is bouncing back but still trails global funding growth. HSBC has launched a new Innovation Banking arm in Singapore, committing $1.5 billion to support venture-backed startups in the region. The highlight is LionsBot, a Singapore-made cleaning robot maker that raised funds and is now prepping for a potential IPO, showcasing how homegrown robotics firms can scale globally with the right financial backing.

Singapore-based startup K25.ai is positioning itself as Asia's homegrown prediction markets platform, filling a gap left by US giants Polymarket and Kalshi who face bans across many Asian markets. The platform combines AI with live-streamed events and K-pop entertainment, including a deal with BTS. It has secured Curacao and EU gaming licenses while preparing to navigate regulations market-by-market across Asia.

A Chinese AI startup called Z.ai just released a model called GLM-5.2 that developers say performs nearly as well as top offerings from OpenAI and Anthropic—but costs about one-sixth the price. It's now ranked fifth globally on AI performance leaderboards and has drawn praise from figures like Marc Andreessen. The model is open-source, meaning anyone can use or modify it. Experts call it a "mini DeepSeek moment" that could reshape the AI industry, especially as U.S. regulation of AI remains unpredictable. However, major enterprises in banking and security remain skeptical of Chinese AI models due to data concerns.

Kunlunxin, Baidu's chip design arm, just won a massive $139 million deal to supply AI chips to China Mobile. These chips are designed to work with Nvidia's popular CUDA software, making it easier for Chinese companies to switch away from American chips. A Huawei-affiliated company also got part of the order, and both are competing against Nvidia in China's massive domestic market.

Nvidia has started taking orders for its Vera processors from Chinese customers, with chips potentially shipping by August. This comes after months of stalled H200 exports due to US restrictions. Chinese firms are reportedly testing hundreds of servers overseas before committing, part of a broader pattern of using offshore data centers to access advanced chips.

Britain's biggest electronics retailer Currys is warning that smartphones and laptops will get more expensive later this year. AI data centers are hogging the world's memory chip supply, leaving gadget makers fighting over scraps. Apple and Xbox have already raised prices, and analysts expect things to worsen before they improve. Currys CEO says his company secured supplies ahead of time, but smaller retailers won't be so lucky.

Microsoft has launched a new company to help businesses adopt AI, backed by a $2.5 billion commitment. Details remain thin on what specifically this firm will do or how it differs from Microsoft's existing AI consulting services.

The AI boom is straining communities around data centers. In Lowell, Massachusetts, residents are fighting back after a facility installed cooling tanks behind homes, ran diesel generators that worsened air quality, and consumed 118,000 gallons of water daily. The city council voted 10-0 to freeze new data center construction for a year. Experts say heat waves make data centers harder to operate, requiring more power and water to stay cool. While researchers say the surge is 'very much a local phenomenon' rather than a national crisis, communities bearing the environmental costs are demanding accountability.

Y Combinator-backed Weave Robotics has launched Isaac 1, a home robot priced at $7,999 (or $449 monthly) that promises to do your laundry, make your bed, and tidy up your home. The launch post racked up over 13 million views on X, fueling speculation that household robots may finally be becoming mainstream—though critics call it 'slow' and 'clunky.' The company uses personal data to improve services, though it's unclear if home data trains the robot.

Chinese robotics firm UBTech has launched ultra-realistic humanoid companion robots with AI that reads human emotions. Top models cost up to $145,000 and feature lifelike details like visible pores and fingerprints. The company already has over 13,000 orders. UBTech predicts China's market for these robots will explode from billions to trillions of yuan over the next decade, as AI "companions" move from sci-fi to reality.

Rising inflation and uncertainty over Federal Reserve interest rate decisions are making AI-focused stock investments more volatile. When job numbers and inflation data miss expectations, markets react sharply — the Nasdaq dropped 4% in one day. Crypto is suffering even more, behaving like a high-risk tech stock rather than a safe haven. Meanwhile, upcoming IPOs from OpenAI and Anthropic are expected to draw big investor interest.

Standard Form, a US healthcare software startup, argues that cheaper AI tools do not automatically solve the hardest problems in healthcare adoption. Its CTO says the real barriers are trust, hospital procurement, and proving value to cautious buyers—not coding costs. He also questions whether Indonesia's 300 million people translate to US-style software budgets, suggesting the best Southeast Asian software targets daily needs rather than enterprise software.

Kuaishou's AI video platform Kling is reportedly close to raising US$3 billion in funding, having already generated RMB 340 million in Q4 revenue. This positions Kling as a major contender in the AI video space, competing directly with OpenAI's Sora (which has since been discontinued).

Eli Lilly has signed a deal worth up to US$2.75 billion with Hong Kong-based biotech Insilico Medicine to develop AI-discovered drug candidates. Insilico will get US$115 million upfront, with more tied to future milestones. The company has used generative AI to develop at least 28 drugs, with nearly half already in clinical trials. This is one of the biggest licensing deals proving AI can genuinely accelerate drug discovery—Insilico says it can nominate preclinical candidates in 12-18 months versus the traditional 4.5-year industry average.

OpenAI has reportedly proposed giving the Trump administration a 5% equity stake, with other US AI companies asked to do the same. The arrangement would give the government a financial interest in AI firms, though it's unclear if competitors like Google or Anthropic would agree. This marks a potential turning point in how the US government engages with AI companies beyond just regulating them.

Nvidia claims its new GB200 NVL72 server system can speed up certain open-source AI models by up to 10 times compared to older hardware. The system links 72 Blackwell GPUs to handle AI tasks more efficiently. However, analysts have cut 2025 shipment forecasts for these cabinets by over 50%, and pricing remains undisclosed. Major cloud providers like AWS and Google Cloud are deploying the technology. For users, this could eventually mean faster, cheaper AI services, though widespread availability remains uncertain.

This is a feature story on how FIFA and sports organizations plan to use AI and data analytics to win over fans in Asia-Pacific ahead of the 2026 World Cup. It covers fan engagement tactics, data-driven marketing, and technology trends in sports entertainment.

Nvidia is rolling out a new way for small AI startups to access its powerful computing hardware—instead of paying huge upfront costs, they can get cloud credits now and share a slice of their future revenue with Nvidia later. Nvidia will also connect these startups with data center operators and cloud providers, making it easier to get AI projects running without the usual delays of building infrastructure. It's a bid to expand Nvidia's customer base beyond big tech companies and into emerging AI businesses.

The US government is negotiating voluntary safety standards with major AI companies like OpenAI, Anthropic, and Google before releasing their most powerful models. Think of it as a pre-flight check for AI — companies would need to meet certain benchmarks and timelines before launching advanced AI to the public. The move comes after concerns that cutting-edge AI could be weaponised by adversarial nations like China or Russia. For everyday users, this could mean new AI features (like smarter chatbots or coding tools) might take longer to reach you, as companies navigate these new requirements.

SK hynix, the world's second-largest memory chipmaker, plans a massive $64 billion expansion at its Cheongju facility focused on advanced chip packaging technology. The new plant will combine multiple chips into single high-density units, which is crucial for AI processors. This expansion targets the growing AI chip market where advanced packaging is increasingly important.

Samsung spent a record $25.7 billion on research and development in 2025, a 7.8% jump from the previous year, all aimed at winning back ground in the AI memory chip race. The company lost its profit leadership to SK hynix in 2025, which now dominates the high-bandwidth memory (HBM) market with 57% share versus Samsung's 22%. This spending surge signals that memory prices—and potentially your next smartphone or laptop—will keep climbing as AI demand outstrips supply.

Vinton Cerf, who helped invent the internet's core plumbing, says AI is at a crossroads similar to where the web was in its early days. He argues AI agents will need standardized, open communication rules—not just English chitchat—to work together reliably. His takeaway: the AI companies that succeed will be the ones that build platforms others can build on, not those trying to control everything themselves.

Meta has agreed to purchase 1.6 gigawatts of AI computing capacity from Crusoe, a US data center developer, across facilities in Texas and Missouri. This massive power deal shows how tech giants are locking up AI infrastructure to support their expanding AI ambitions, similar to how cloud providers secured data center capacity in earlier eras.

AI models are getting commoditized fast — their cost has dropped 1,000x in three years, making the real money opportunity shift to the infrastructure layer that delivers AI answers (inference). This market is now worth US$118 billion and growing, accounting for two-thirds of all AI compute spending. For everyday users, this shift could mean cheaper AI tools over time as companies compete on delivery rather than raw model smarts.

SK hynix said the packaging plant will focus on advanced packaging that combines chips into a single high-density unit.

South Koreans are increasingly turning to AI startups to create video messages from deceased loved ones. Companies like Seoul-based Vaice charge around $390 to generate AI videos using photos and voice samples, producing clips where grandparents or parents appear to deliver personal messages. Families play these videos during memorial rituals or holidays. While users say the technology brings comfort—some burst into tears watching their late relatives speak—experts warn it could trap grieving people in denial. There's growing calls for laws to protect the dignity of the deceased and prevent misuse of their images and voices.

A chip startup called OXMIQ, founded by a former Intel executive, has raised $35 million in Series A funding to develop licensable GPU technology. The investment signals continued confidence in AI chip innovation beyond established players like Nvidia. For everyday users, this could mean more competition in the AI hardware space over time, potentially leading to more choices and possibly lower costs for AI-powered services.

This article discusses a Chinese AI startup called Liqing that secured backing from venture firms Shunwei and Sequoia China. The piece explores funding dynamics in China's competitive AI landscape and what investor confidence in new AI players means for the broader market.

Israeli AI startups collectively raised US$3 billion in June alone, with AppsFlyer leading at over US$1 billion from major tech investors including Google, Meta, Unity, and Moloco. This shows big tech companies continue betting heavily on AI-focused assets even in the current funding climate.

Amazon is reportedly joining a planned $100 million funding round for Twelve Labs, an AI startup focused on video understanding technology. This adds to a broader trend of big tech companies investing in AI startups to secure capabilities and partnerships.

Vibe coding startup Base44 (acquired by Wix for $80 million) has launched its own AI model called Base 1 to solve a common problem: AI-coded websites all look the same — beige backgrounds, generic sans-serif fonts, cookie-cutter layouts. The company argues that frontier AI models optimized for everything from poetry to coding produce generic user interfaces. Base 1 is trained specifically on design tasks using data from Wix's large design team. Users can now choose Base 1 alongside models like Claude Opus 4.8 and GPT-5.5 when building apps through Base44.

AIPath is a decision-intelligence startup that lets companies simulate growth strategies before committing resources. It helped one insurance client cut customer acquisition costs from $240 to just $43. The company is currently raising a $1.5 million seed round and targets businesses with $5 million to $50 million in revenue.

Google's electricity consumption jumped 37% in 2025—the biggest spike in the company's history—driven by its AI data center expansion. The tech giant's data centers now use enough power to rival entire countries like New Zealand. While Google claims it offset this with clean energy purchases and cut operational emissions by 2%, its supply chain emissions rose 25% due to manufacturing in Asia-Pacific grids still reliant on fossil fuels. This highlights the hidden carbon cost behind AI services like ChatGPT, search, and cloud computing.

Singapore authorities seized a $42 million mansion connected to an investigation into illegal Nvidia chip trading. The crackdown, which included raids on Super Micro's Taiwan offices, signals tighter enforcement of export controls on AI chips. Singapore's role as a major semiconductor trading hub makes this a significant development for regional supply chains.

Researchers trained GPT-4 Turbo to impersonate 112 UK public figures on a TV debate show. When participants were asked to rate AI-generated responses against real politicians, the AI was rated as MORE authentic, coherent, and relevant. More than half of participants thought the fake AI versions sounded more like the real person than the actual person. This is a serious red flag—if AI can make up convincing fake politicians that fool people, it could supercharge misinformation ahead of elections. The researchers are calling for public education and regulations to combat this threat.

AIPath, a Singapore-based startup, helps companies test business strategies using AI simulation before spending money on execution. Their platform can map thousands of growth options and ranked strategies in minutes. The company won first place in the 2025 AI Agents Global Challenge and helped an insurance client cut customer acquisition costs from $240 to just $43. AIPath is currently raising a $1.5 million seed round.

OpenAI's Sam Altman is proposing that AI companies like his own donate 5% of equity to a U.S. sovereign wealth fund, with other AI firms doing the same. The idea is to give everyday Americans a stake in the AI industry's growth and ease political pressure on big tech. A separate Senate bill from Bernie Sanders goes further, proposing a 50% tax on AI company stock to fund a $7 trillion public wealth fund. Both proposals face major hurdles, including needing congressional approval, and remain purely theoretical for now.

Nvidia is actively funding AI startups as a strategy to reduce Big Tech companies' control over its chip business. By supporting emerging AI companies, Nvidia is positioning itself as a key financial backer shaping the AI compute market, potentially loosening the grip that Google, Amazon, and Microsoft have on AI infrastructure.

OpenAI is reportedly offering the Trump administration a 5% stake in the company, according to The Decoder. The move signals the company's effort to strengthen ties with Washington as it faces mounting regulatory scrutiny. What the government would offer in return remains unclear.

Microsoft has launched a new division called Microsoft Frontier with a $2.5 billion commitment to help businesses actually deploy AI successfully. The venture will include 6,000 engineers and experts focused on getting AI working in real companies like Unilever and Accenture, not just building models. It's a bet that the real money in AI isn't just in building AI, but in making AI work for enterprises.

AI agents have quietly taken over more freelance work than ever. According to the Remote Labor Index, the share of paid freelance jobs AI can complete at professional quality has surged from 2.5% to 16% in just eight months—more than quadrupling. This means AI is increasingly capable of handling tasks like writing, design, and coding that freelancers previously dominated. For freelancers, this is a wake-up call: competition just got fiercer, and rates may come under pressure as clients realize AI can do the job faster and cheaper.

AI is getting too expensive for companies to use freely. Major firms including Atlassian, Adobe, Amazon, and Citi are now restricting how much AI their employees can use—or charging them based on token usage instead of flat fees. Atlassian's AI spending tripled from $5 million to over $15 million in a single month, hitting $120 million annually. Citi has even blocked access to top AI models like Claude Opus 4.7 and GPT-5.5. The message to workers: use cheaper, less powerful AI models, or cut back entirely. One developer reportedly consumed nearly half an entire company's AI budget in a month.

Netflix's new reality show 'The Golden Ticket' will use AI to recreate the voice of the late Gene Wilder for a Willy Wonka series—with his estate's blessing. While Wilder's wife called it a celebration of his imagination, fans on social media called it 'disrespectful' and 'a plastic substitute,' with some jokingly comparing it to the disastrous 'Willy's Chocolate Experience' event. Studios appear to be testing what audiences will accept, following examples like Disney's digital recreation of James Earl Jones as Darth Vader. Experts say the more beloved the voice, the more scrutiny the AI version faces.

A new Los Angeles Times investigation reveals that Big Tech's climate pledges are unraveling as AI workloads surge. The article examines how Microsoft's carbon neutrality claim, Google's clean energy matching, and Amazon's climate commitments are being strained by the massive electricity demands of AI training and inference. The piece ties into Google's 37% energy spike and broader industry trends.

A survey of 1,100 Dutch young people aged 16-34 found that 51% discuss personal matters—dating, mental health, social situations—with AI chatbots. A quarter share more with bots than with friends or family. One respondent used ChatGPT as a 'talking diary' after a breakup. Experts warn this trend could hinder personal development, as AI provides calculated responses rather than genuine empathy. 7% consider chatbots a friend, with lonelier users more likely to form these bonds.

Amazon's carbon emissions surged 16% in 2025, with the main cause being its massive data center expansion to power AI services. The tech giant is building more server farms than ever to handle AI workloads, and this infrastructure boom comes with a heavy environmental price tag. For consumers, this raises questions about the hidden ecological cost of every AI query, chatbot interaction, or AI-generated image they use.

While chatbots and image generators grab headlines, AI is quietly transforming heavy industries like wind energy. The article explores how AI is being deployed in operational technology—monitoring turbines, predicting equipment failures, and optimizing industrial processes that keep power flowing. For everyday readers, this matters because the same AI driving efficiency in data centers can also help keep electricity affordable and renewable energy reliable.

Microsoft has released Azure Linux 4.0, a significant update to its in-house Linux distribution designed to handle AI workloads and cloud infrastructure. The new version offers improved performance for running AI models and cloud services, positioning it as a serious contender to replace Windows Server in enterprise environments. Azure Linux already powers Microsoft's AI infrastructure including Copilot.

OpenAI has floated giving the US government a 5 percent ownership stake as a way of easing tensions with the Trump administration and blunting mounting public backlash against AI, according to the Financial Times. CEO Sam Altman argued that giving the public a financial interest in the company would be the best way to share […]

Jeff Bezos' family office backed five AI startups in June, accounting for 10% of all family office dealmaking that month. The biggest bet was a $12 billion Series B for Prometheus, the physical AI startup where Bezos serves as co-CEO. Prometheus aims to build an 'artificial engineer' to accelerate product design from jet engines to pharmaceuticals.

SpaceX has filed for an orbital data center constellation of up to 1 million satellites and recently revealed design specs for an AI-1 satellite data center. However, experts argue the economics don't add up yet. Cooling a single Nvidia H100 GPU in space requires 1.4 square meters of radiator, and a 100-megawatt data center would need 2,500 massive radiators. While Musk claims space-based AI will be cheaper within 2-3 years, analysts estimate cost parity is 5-10 years away at best. Launch capacity and satellite manufacturing would need to scale astronomically to meet Musk's vision.

The CEO of Digitas, one of the world's largest advertising agencies, is calling out AI hype at Cannes. Amy Lanzi says most AI promises in marketing are 'false promises'—similar to the hype around programmatic advertising a decade ago. While her parent company Publicis has invested in AI tools, she emphasizes that advertising still needs human creativity and judgment, not just automation.

ZDNet walks through setting up OpenAI API spending limits to prevent runaway bills from AI agents. Covers usage tiers that unlock higher monthly spending caps, spending alerts, rate limits, and hard limits that reject API calls once budget is hit. Essential reading for developers running AI agents at scale.

Popular TV-tracking app TV Time, with over 26 million lifetime installs, is shutting down on July 15, 2026. Its owner Whip Media—acquired by Blue Torch Capital in 2025—is pivoting to AI-focused products like Helix, an AI-powered workflow tool. The company will delete all user data. This follows a trend where AI priorities are causing companies to shut down beloved consumer apps, similar to Mozilla discontinuing Pocket.

A sponsored MIT Technology Review Insights report explores how companies are merging AI tools with traditional process management methods like Lean Six Sigma. The AI-powered process optimization market could exceed $113 billion in the next decade, with 88% of business leaders planning to increase investments in AI-infused process intelligence. The report argues that companies with strong existing process frameworks are best positioned to benefit from AI integration, while those without solid foundations risk wasting their investments.

A peer-reviewed study published in PLOS One has found a critical gap: there's little to no legal or ethical guidance for using AI in citizen science projects. Citizen science involves everyday people helping conduct research, like tracking wildlife or monitoring air quality. The study says AI could help scale these efforts, but almost no rules exist around AI training data transparency or ethical safeguards. Researchers made 10 recommendations to fix this, emphasizing that AI adoption in public science needs proper guardrails first.

People are getting creative with OpenClaw AI agents for dating. One user built a system to auto-generate Instagram reels about sports losses to attract heartbroken fans, racking up a million views and 200 DMs in days. Others use it to plan dates, though security experts warn about privacy risks when AI agents get access to personal accounts and messages.

Tesla just released its second-quarter delivery and production report, showing that the automaker is starting to recover after a particularly brutal sales year in 2025. The company said that it produced a total of 451,758 vehicles between April and June of this year, including 442,936 Model 3 and Model Y vehicles, as well as 8,822 […]

UK electrical industry groups warn that plug-in solar kits pose 6 risks: fire from old wiring, shock from two-way power flow, cheap hardware, grid instability, insurance liability gaps, and poor DIY installation. They urge consumers to hire licensed electricians and caution that a 'rushed rollout could weaken confidence in the energy transition.'

HP's OmniBook Ultra 14 is a premium thin-and-light laptop built for professionals, featuring either Qualcomm Snapdragon or Intel processors, a sharp 3K OLED touchscreen, and a keyboard reviewers call excellent. The standout feature is battery life — it runs nearly 24 hours on a single charge. Pricing starts around $1,200 for the base Intel model.

ExpressVPN's password manager now supports secure password sharing between users and passkeys — a more secure login method that cannot be phished like traditional passwords. Passkeys replace the need to remember complex strings of characters, using your device's biometric security instead. This upgrade signals the broader shift toward passwordless authentication that could make your accounts harder to hack.

X Trending * [I wore the world's first HDR10 smart glasses](https://www.zdnet.com/article/rayneo-air-4-pro-hdr10-smart-glasses-ces/) * [TCL's new E Ink tablet beats the Remarkable and Kindle](https://www.zdnet.com/article/tcl-note-a1-nxtpaper-ces/) * [Anker's new charger is one of the most unique I've ever seen](https://www.zdnet.com/article/anker-wireless-chargers-ces-2026/) * [Best laptop cooling pads](https://www.zdnet.com/home-and-office/best-laptop-cooling-pad/) * [Best flip phones](https://www.zdnet.com/article/best-flip-phone/) * [Roku TV vs Fire Stick](https://www.zdnet.com/home-and-office/home-entertainment/roku-tv-vs-fire-stick/) * [Galaxy Buds 3 Pro vs Apple AirPods Pro 3](https://www.zdnet.com/article/apple-airpods-pro-3-vs-samsung-galaxy-buds-3-pro/) * [M5 MacBook Pro vs M4 Ma

Taiwanese prosecutors have detained two Super Micro Computer employees and placed four on administrative leave, following raids on the company's local offices as part of an investigation into illegal exports of Nvidia AI chips to China. The suspects allegedly falsified documents to ship servers containing coveted AI processors through intermediaries like Albatron Technology and Chief Telecom. Taiwan is now considering criminalizing AI chip exports to China outright, since current law doesn't explicitly prohibit them — even as the US pressures allies to cut off Beijing's access to advanced AI technology.

Singapore-based startup K25.ai is positioning itself as Asia's homegrown prediction markets platform, filling a gap left by US giants Polymarket and Kalshi who face bans across many Asian markets. The platform combines AI with live-streamed events and K-pop entertainment, including a deal with BTS. It has secured Curacao and EU gaming licenses while preparing to navigate regulations market-by-market across Asia.

Kuaishou's AI video platform Kling is reportedly close to raising US$3 billion in funding, having already generated RMB 340 million in Q4 revenue. This positions Kling as a major contender in the AI video space, competing directly with OpenAI's Sora (which has since been discontinued).

The US government is negotiating voluntary safety standards with major AI companies like OpenAI, Anthropic, and Google before releasing their most powerful models. Think of it as a pre-flight check for AI — companies would need to meet certain benchmarks and timelines before launching advanced AI to the public. The move comes after concerns that cutting-edge AI could be weaponised by adversarial nations like China or Russia. For everyday users, this could mean new AI features (like smarter chatbots or coding tools) might take longer to reach you, as companies navigate these new requirements.

Meta has agreed to purchase 1.6 gigawatts of AI computing capacity from Crusoe, a US data center developer, across facilities in Texas and Missouri. This massive power deal shows how tech giants are locking up AI infrastructure to support their expanding AI ambitions, similar to how cloud providers secured data center capacity in earlier eras.

AI models are getting commoditized fast — their cost has dropped 1,000x in three years, making the real money opportunity shift to the infrastructure layer that delivers AI answers (inference). This market is now worth US$118 billion and growing, accounting for two-thirds of all AI compute spending. For everyday users, this shift could mean cheaper AI tools over time as companies compete on delivery rather than raw model smarts.

Israeli AI startups collectively raised US$3 billion in June alone, with AppsFlyer leading at over US$1 billion from major tech investors including Google, Meta, Unity, and Moloco. This shows big tech companies continue betting heavily on AI-focused assets even in the current funding climate.

AI Singapore, the national agency driving the country's AI strategy, has new leadership in place to push forward its ambitions. The timing aligns with Singapore's broader Smart Nation goals, where AI plays a central role in digital government, economy, and society. For Singaporeans, this leadership change could shape how quickly and effectively AI is adopted across public services, jobs, and daily life.

Nvidia claims its new GB200 NVL72 server system can speed up certain open-source AI models by up to 10 times compared to older hardware. The system links 72 Blackwell GPUs to handle AI tasks more efficiently. However, analysts have cut 2025 shipment forecasts for these cabinets by over 50%, and pricing remains undisclosed. Major cloud providers like AWS and Google Cloud are deploying the technology. For users, this could eventually mean faster, cheaper AI services, though widespread availability remains uncertain.

South Koreans are increasingly turning to AI startups to create video messages from deceased loved ones. Companies like Seoul-based Vaice charge around $390 to generate AI videos using photos and voice samples, producing clips where grandparents or parents appear to deliver personal messages. Families play these videos during memorial rituals or holidays. While users say the technology brings comfort—some burst into tears watching their late relatives speak—experts warn it could trap grieving people in denial. There's growing calls for laws to protect the dignity of the deceased and prevent misuse of their images and voices.

A chip startup called OXMIQ, founded by a former Intel executive, has raised $35 million in Series A funding to develop licensable GPU technology. The investment signals continued confidence in AI chip innovation beyond established players like Nvidia. For everyday users, this could mean more competition in the AI hardware space over time, potentially leading to more choices and possibly lower costs for AI-powered services.

This article discusses a Chinese AI startup called Liqing that secured backing from venture firms Shunwei and Sequoia China. The piece explores funding dynamics in China's competitive AI landscape and what investor confidence in new AI players means for the broader market.

Amazon is reportedly joining a planned $100 million funding round for Twelve Labs, an AI startup focused on video understanding technology. This adds to a broader trend of big tech companies investing in AI startups to secure capabilities and partnerships.

Singapore's government is rolling out new tools to help retail small businesses adopt AI. Enterprise Singapore and the Ministry of Trade and Industry launched a refreshed Retail Industry Digital Plan and a new two-year incubator programme called Retail Accelerator at L^IFE. The updated plan focuses on practical uses like demand forecasting, AI-powered marketing, and staff training tools. A local florist cut cashier workloads and marketing time using AI, while a jeweller built a voice-activated inventory system. Sixteen retailers are joining the accelerator, getting subsidised space at SCAPE, live-streaming studios, and mentorship.

Eli Lilly has signed a deal worth up to US$2.75 billion with Hong Kong-based biotech Insilico Medicine to develop AI-discovered drug candidates. Insilico will get US$115 million upfront, with more tied to future milestones. The company has used generative AI to develop at least 28 drugs, with nearly half already in clinical trials. This is one of the biggest licensing deals proving AI can genuinely accelerate drug discovery—Insilico says it can nominate preclinical candidates in 12-18 months versus the traditional 4.5-year industry average.

Southeast Asia's startup scene is bouncing back but still trails global funding growth. HSBC has launched a new Innovation Banking arm in Singapore, committing $1.5 billion to support venture-backed startups in the region. The highlight is LionsBot, a Singapore-made cleaning robot maker that raised funds and is now prepping for a potential IPO, showcasing how homegrown robotics firms can scale globally with the right financial backing.

OpenAI has reportedly proposed giving the Trump administration a 5% equity stake, with other US AI companies asked to do the same. The arrangement would give the government a financial interest in AI firms, though it's unclear if competitors like Google or Anthropic would agree. This marks a potential turning point in how the US government engages with AI companies beyond just regulating them.

Samsung spent a record $25.7 billion on research and development in 2025, a 7.8% jump from the previous year, all aimed at winning back ground in the AI memory chip race. The company lost its profit leadership to SK hynix in 2025, which now dominates the high-bandwidth memory (HBM) market with 57% share versus Samsung's 22%. This spending surge signals that memory prices—and potentially your next smartphone or laptop—will keep climbing as AI demand outstrips supply.

Vinton Cerf, who helped invent the internet's core plumbing, says AI is at a crossroads similar to where the web was in its early days. He argues AI agents will need standardized, open communication rules—not just English chitchat—to work together reliably. His takeaway: the AI companies that succeed will be the ones that build platforms others can build on, not those trying to control everything themselves.

Vibe coding startup Base44 (acquired by Wix for $80 million) has launched its own AI model called Base 1 to solve a common problem: AI-coded websites all look the same — beige backgrounds, generic sans-serif fonts, cookie-cutter layouts. The company argues that frontier AI models optimized for everything from poetry to coding produce generic user interfaces. Base 1 is trained specifically on design tasks using data from Wix's large design team. Users can now choose Base 1 alongside models like Claude Opus 4.8 and GPT-5.5 when building apps through Base44.

AIPath, a Singapore-based startup, helps companies test business strategies using AI simulation before spending money on execution. Their platform can map thousands of growth options and ranked strategies in minutes. The company won first place in the 2025 AI Agents Global Challenge and helped an insurance client cut customer acquisition costs from $240 to just $43. AIPath is currently raising a $1.5 million seed round.

US industrial giant Wesco International has bought Singapore's Newark Engineering for $136 million. Newark designs and installs cooling systems for data centers—the massive warehouses that house servers powering AI and cloud services. With AI demand surging, data center operators are scrambling to manage heat output. This acquisition lets Wesco tap into Southeast Asia's fast-growing data center market, where Singapore, Malaysia, and Indonesia are all building new facilities. For everyday users, this means more data center capacity could eventually mean faster, more reliable AI services.

Banks are saying AI applications that interact with the real world—think robots in factories, self-driving vehicles, and industrial automation—could reshape how telecom companies earn money. Instead of just selling data plans to phone users, telecoms could charge businesses premium rates for reliable, fast connections needed to run these machines. Maybank predicts Physical AI traffic could make up 32% of all mobile data by 2031, potentially lifting telecom revenue growth from low single digits to high single digits. This is a big shift from previous AI booms that mostly benefited data centers and chipmakers.

NTU Singapore is launching a new part-time graduate program to train chip designers, developed with the Singapore Semiconductor Industry Association and the Economic Development Board. Students will learn to design digital chips using industry-standard tools from companies like Synopsys and Cadence, getting hands-on experience from concept to fabrication-ready designs. The program targets working engineers and graduates looking to enter the semiconductor field, with courses that can stack toward a full Master's degree. It's designed to feed talent into fabless companies like AMD, Broadcom, and MediaTek that already run design operations in Singapore.

Nvidia is rolling out a new way for small AI startups to access its powerful computing hardware—instead of paying huge upfront costs, they can get cloud credits now and share a slice of their future revenue with Nvidia later. Nvidia will also connect these startups with data center operators and cloud providers, making it easier to get AI projects running without the usual delays of building infrastructure. It's a bid to expand Nvidia's customer base beyond big tech companies and into emerging AI businesses.

SK hynix, the world's second-largest memory chipmaker, plans a massive $64 billion expansion at its Cheongju facility focused on advanced chip packaging technology. The new plant will combine multiple chips into single high-density units, which is crucial for AI processors. This expansion targets the growing AI chip market where advanced packaging is increasingly important.

SK hynix said the packaging plant will focus on advanced packaging that combines chips into a single high-density unit.

AIPath is a decision-intelligence startup that lets companies simulate growth strategies before committing resources. It helped one insurance client cut customer acquisition costs from $240 to just $43. The company is currently raising a $1.5 million seed round and targets businesses with $5 million to $50 million in revenue.

Boston Dynamics' newest Atlas humanoid robot is 10 times simpler to build than its predecessor, using far fewer parts while matching or exceeding its capabilities. The company aims to manufacture 30,000 units annually, with parent company Hyundai keeping most for its own factories. Boston Dynamics emphasizes its reasoning AI layer lets the robot adapt to new tasks without lengthy reprogramming. Hyundai, which owns Kia and Genesis, brings automotive-scale manufacturing expertise that could finally make humanoids affordable.

Meta, the company behind Facebook and Instagram, spent roughly $145 billion on AI infrastructure in 2026 and now wants to rent out its excess computing power to other businesses. Think of it like a hotel that built too many rooms and is now opening bookings to strangers. The move has boosted Meta's stock by 9% as investors see a new revenue stream beyond ads. Smaller cloud players like CoreWeave and Nebius are watching nervously since Meta was once their customer and is now becoming their competitor.

Chinese AI startup Z.ai has launched ZCode, an AI coding tool that connects to multiple frontier models for writing, reviewing, and deploying code. It undercuts American competitors like Cursor and GitHub Copilot on price, with plans starting at $16.20 monthly compared to Cursor's $20 minimum. ZCode serves as the official development environment for Z.ai's GLM 5.2 open-source model.

Reports suggest the Trump administration and Anthropic have not discussed the government taking stakes in the AI company. This comes amid earlier reports that OpenAI was in talks about giving the US government a 5% stake, which Anthropic appears to not be pursuing.

OpenAI appears to be developing a small physical keyboard for coders to use with its Codex programming platform. The article suggests two reasons: branding (competing with Anthropic's momentum) and as a test run for larger AI hardware ambitions. The company is separately working with former Apple designer Jony Ive on more ambitious AI devices like pendants or desk accessories.

Tesla has launched a longer Model Y in the U.S. starting at $61,990 with a six-seat configuration. The stretched version adds six inches to the wheelbase, giving much more room in the third row compared to the cramped seven-seater option. It comes with a year of Full Self-Driving capability included. Deliveries began in Malaysia last month and the vehicle is also available in China, Australia, and India.

South African retailer Pick n Pay has launched an AI shopping assistant called Penny that lets customers build grocery orders using voice notes, text, or photos of handwritten lists and recipes. Powered by Google's Gemini AI, the assistant suggests recipes, meal plans, and substitutions while supporting multiple languages. It's the retailer's push to catch up with rivals in the competitive on-demand grocery market.

Meta CEO Mark Zuckerberg admitted internally that AI agents are not developing as quickly as the company had hoped. Earlier this year, Meta cut 8,000 jobs and reassigned 7,000 more workers to AI teams, betting that AI would automate human tasks. Despite investing up to $145 billion in AI infrastructure this year, the expected productivity gains have not materialized. Zuckerberg said improvements should come in the next three to six months, but some engineers describe the AI unit as a difficult place to work.

Microsoft is spending $2.5 billion on a new unit called 'Frontier Company' that will embed 6,000 AI engineers directly inside enterprise clients' teams. Unlike typical software sales, these engineers will work on-site to integrate AI into the companies' actual workflows with measurable returns. Microsoft is pitching itself as platform-neutral, unlike competitors OpenAI and Anthropic who push their own models.

Anthropic, the AI company behind Claude, is reportedly in early talks with Samsung to manufacture custom AI chips. This follows OpenAI's similar move with its Jalapeño chip. The goal is to reduce reliance on Nvidia and cut the massive infrastructure costs of running AI systems. Anthropic still says Nvidia remains important to its operations.

Google and Amazon just released sustainability reports showing their carbon emissions jumped 25% and 16% respectively — and AI is the main culprit. Both companies are building massive data centers to power AI services, consuming huge amounts of energy and water. Despite their net-zero pledges, the AI boom is making those climate promises harder to keep, with both firms now investing in natural gas plants to meet surging power demand.

AI startup Anthropic is in talks with Samsung to develop custom AI chips, joining rivals like OpenAI and Google in reducing dependence on Nvidia. The move signals a broader industry shift toward purpose-built silicon for AI workloads. Samsung already manufactures chips for Nvidia and is building an AI chip factory in South Korea with Nvidia.

Anthropic's Fable 5 AI model has shattered records on the Remote Labor Index (RLI) benchmark, hitting a 16.1% automation rate for real freelance work tasks — double the previous leader. The Center for AI Safety tested Fable 5 alongside OpenAI's GPT-5.5 and Anthropic's Opus 4.8, with Fable 5 dramatically outperforming both. The benchmark measures how often AI can complete economically valuable freelance projects (like graphic design, video work, and data analysis) at a quality a paying client would accept. While 16% is far from replacing all freelancers, the speed of improvement is striking — the benchmark leader's score quadrupled in under eight months.

SpaceX's $60 billion acquisition of AI coding startup Cursor raises big questions: will it keep offering OpenAI and Anthropic models, or will it become a captive tool for Musk's ecosystem? Cursor says it plans to stay open, but the deal puts SpaceX in direct competition with its own AI suppliers.

A Tesla driver in Texas faces manslaughter charges after his car using Full Self-Driving (FSD) struck and killed a woman inside her home. The driver claimed he was relying on FSD at the time. This case is significant because it tests who bears legal responsibility when autonomous driving features fail — the person behind the wheel or the carmaker. As more vehicles gain self-driving capabilities, this precedent could shape how such technology gets regulated and used worldwide.

Anthropic reduced the system prompt for Claude Code by 80 percent with its Fable 5 models. The company found that newer models perform better with fewer instructions because they're more imaginative and can get confused by overly strict rules. Instead of detailed guidelines, Anthropic now steers behavior through context rather than explicit commands.

Update on Proposed data center would pull from lake that supplies most of Fort Worth’s water. Coverage via Reddit/r/technology.

Meta has launched a new app called Pocket that lets users create and share interactive AI-powered "gizmos" built from text prompts. This has nothing to do with the Mozilla read-it-later service that shut down last year. Instead, it's Meta's latest move to put AI creation tools directly in people's hands — users can build little interactive experiences and share them with friends.

X Trending * [I wore the world's first HDR10 smart glasses](https://www.zdnet.com/article/rayneo-air-4-pro-hdr10-smart-glasses-ces/) * [TCL's new E Ink tablet beats the Remarkable and Kindle](https://www.zdnet.com/article/tcl-note-a1-nxtpaper-ces/) * [Anker's new charger is one of the most unique I've ever seen](https://www.zdnet.com/article/anker-wireless-chargers-ces-2026/) * [Best laptop cooling pads](https://www.zdnet.com/home-and-office/best-laptop-cooling-pad/) * [Best flip phones](https://www.zdnet.com/article/best-flip-phone/) * [Roku TV vs Fire Stick](https://www.zdnet.com/home-and-office/home-entertainment/roku-tv-vs-fire-stick/) * [Galaxy Buds 3 Pro vs Apple AirPods Pro 3](https://www.zdnet.com/article/apple-airpods-pro-3-vs-samsung-galaxy-buds-3-pro/) * [M5 MacBook Pro vs M4 Ma

A TechCrunch opinion piece argues that AI hype has officially peaked — using Jersey Mike's IPO filing as evidence. The sandwich chain, which sells submarine sandwiches, somehow found a way to mention 'AI' 22 times in its investor documents, including vague risk warnings about 'using AI Technologies in our business.' The piece uses this to illustrate how companies now feel compelled to shoehorn AI into any pitch to attract investor funding, even when their actual business has nothing to do with AI.

Reddit threads are becoming goldmines for AI companies because they capture how humans actually think and make decisions—the messy arguments, workarounds, frustrations, and jokes that reveal real human behavior. As AI-generated content floods the web, authentic human conversations become scarcer and more valuable. Users posting on Reddit should know their discussions are being used to train AI systems.

Renewable energy firm Ib vogt and data center operator AirTrunk have launched Malaysia's first corporate green power agreement specifically for data centers. The 30 MW solar project will supply clean energy to AirTrunk's Malaysian operations, marking a step toward sustainable AI infrastructure in Southeast Asia.

Kunlunxin, Baidu's chip design arm, just won a massive $139 million deal to supply AI chips to China Mobile. These chips are designed to work with Nvidia's popular CUDA software, making it easier for Chinese companies to switch away from American chips. A Huawei-affiliated company also got part of the order, and both are competing against Nvidia in China's massive domestic market.

A Chinese AI startup called Z.ai just released a model called GLM-5.2 that developers say performs nearly as well as top offerings from OpenAI and Anthropic—but costs about one-sixth the price. It's now ranked fifth globally on AI performance leaderboards and has drawn praise from figures like Marc Andreessen. The model is open-source, meaning anyone can use or modify it. Experts call it a "mini DeepSeek moment" that could reshape the AI industry, especially as U.S. regulation of AI remains unpredictable. However, major enterprises in banking and security remain skeptical of Chinese AI models due to data concerns.

Standard Form, a US healthcare software startup, argues that cheaper AI tools do not automatically solve the hardest problems in healthcare adoption. Its CTO says the real barriers are trust, hospital procurement, and proving value to cautious buyers—not coding costs. He also questions whether Indonesia's 300 million people translate to US-style software budgets, suggesting the best Southeast Asian software targets daily needs rather than enterprise software.

This is a feature story on how FIFA and sports organizations plan to use AI and data analytics to win over fans in Asia-Pacific ahead of the 2026 World Cup. It covers fan engagement tactics, data-driven marketing, and technology trends in sports entertainment.

A new survey of over 5,000 workers across six Southeast Asian countries—including Singapore—finds that 73 percent feel their bosses are out of touch with actual digital needs at work. Companies are investing heavily in IT and finance systems but neglecting employee experience and training. Sixty-two percent worry AI will make their jobs obsolete, yet 88 percent still want AI to handle routine tasks. The report warns that poor adoption strategies are turning AI from an opportunity into a threat.

LinqAlpha, a startup building AI tools for institutional investors, has raised $22 million in Series A funding from investors including East Ventures and other SEA players. The company helps Wall Street firms and hedge funds deploy AI agents that analyze market signals before they hit consensus. Its clients manage over $5 trillion in assets. The funding will fuel expansion across Asia.

Nvidia has started taking orders for its Vera processors from Chinese customers, with chips potentially shipping by August. This comes after months of stalled H200 exports due to US restrictions. Chinese firms are reportedly testing hundreds of servers overseas before committing, part of a broader pattern of using offshore data centers to access advanced chips.

Chinese robotics firm UBTech has launched ultra-realistic humanoid companion robots with AI that reads human emotions. Top models cost up to $145,000 and feature lifelike details like visible pores and fingerprints. The company already has over 13,000 orders. UBTech predicts China's market for these robots will explode from billions to trillions of yuan over the next decade, as AI "companions" move from sci-fi to reality.

Researchers trained GPT-4 Turbo to impersonate 112 UK public figures on a TV debate show. When participants were asked to rate AI-generated responses against real politicians, the AI was rated as MORE authentic, coherent, and relevant. More than half of participants thought the fake AI versions sounded more like the real person than the actual person. This is a serious red flag—if AI can make up convincing fake politicians that fool people, it could supercharge misinformation ahead of elections. The researchers are calling for public education and regulations to combat this threat.

Singapore's Ministry of Social and Family Development is spending $15 million over three years to bring AI into the social service sector. The ministry is partnering with NCS and ST Engineering to develop tools that identify families at risk earlier and help social workers manage heavy caseloads. One tool called Scribe cuts documentation time by at least half by transcribing conversations — even in Singlish and Cantonese — into case notes automatically. Another flags blind spots in assessments. Social workers say this frees them to actually connect with clients instead of drowning in paperwork. Minister Masagos Zulkifli stressed AI won't replace human judgment — it's meant to help professionals reach families faster and more effectively.

Rising inflation and uncertainty over Federal Reserve interest rate decisions are making AI-focused stock investments more volatile. When job numbers and inflation data miss expectations, markets react sharply — the Nasdaq dropped 4% in one day. Crypto is suffering even more, behaving like a high-risk tech stock rather than a safe haven. Meanwhile, upcoming IPOs from OpenAI and Anthropic are expected to draw big investor interest.

Your next gadget will cost more, and AI is partly to blame. The worldwide rush to build AI data centers has created a severe memory chip shortage—companies like Apple, Samsung, Microsoft, Sony, and Nintendo are all raising prices on laptops, tablets, smartphones, and gaming consoles. The reason: the same memory chips (DRAM) used in your devices are also needed for AI servers. Data center operators are outbidding consumers for these components, squeezing supply. Micron says new factories will take years to come online, so expect higher prices to stick around.

Microsoft has launched a new company to help businesses adopt AI, backed by a $2.5 billion commitment. Details remain thin on what specifically this firm will do or how it differs from Microsoft's existing AI consulting services.

Google's electricity consumption jumped 37% in 2025—the biggest spike in the company's history—driven by its AI data center expansion. The tech giant's data centers now use enough power to rival entire countries like New Zealand. While Google claims it offset this with clean energy purchases and cut operational emissions by 2%, its supply chain emissions rose 25% due to manufacturing in Asia-Pacific grids still reliant on fossil fuels. This highlights the hidden carbon cost behind AI services like ChatGPT, search, and cloud computing.

Singapore authorities seized a $42 million mansion connected to an investigation into illegal Nvidia chip trading. The crackdown, which included raids on Super Micro's Taiwan offices, signals tighter enforcement of export controls on AI chips. Singapore's role as a major semiconductor trading hub makes this a significant development for regional supply chains.

Nvidia is actively funding AI startups as a strategy to reduce Big Tech companies' control over its chip business. By supporting emerging AI companies, Nvidia is positioning itself as a key financial backer shaping the AI compute market, potentially loosening the grip that Google, Amazon, and Microsoft have on AI infrastructure.

AI agents have quietly taken over more freelance work than ever. According to the Remote Labor Index, the share of paid freelance jobs AI can complete at professional quality has surged from 2.5% to 16% in just eight months—more than quadrupling. This means AI is increasingly capable of handling tasks like writing, design, and coding that freelancers previously dominated. For freelancers, this is a wake-up call: competition just got fiercer, and rates may come under pressure as clients realize AI can do the job faster and cheaper.

While chatbots and image generators grab headlines, AI is quietly transforming heavy industries like wind energy. The article explores how AI is being deployed in operational technology—monitoring turbines, predicting equipment failures, and optimizing industrial processes that keep power flowing. For everyday readers, this matters because the same AI driving efficiency in data centers can also help keep electricity affordable and renewable energy reliable.

Microsoft has released Azure Linux 4.0, a significant update to its in-house Linux distribution designed to handle AI workloads and cloud infrastructure. The new version offers improved performance for running AI models and cloud services, positioning it as a serious contender to replace Windows Server in enterprise environments. Azure Linux already powers Microsoft's AI infrastructure including Copilot.

Microsoft has launched a new division called Microsoft Frontier with a $2.5 billion commitment to help businesses actually deploy AI successfully. The venture will include 6,000 engineers and experts focused on getting AI working in real companies like Unilever and Accenture, not just building models. It's a bet that the real money in AI isn't just in building AI, but in making AI work for enterprises.

A new Los Angeles Times investigation reveals that Big Tech's climate pledges are unraveling as AI workloads surge. The article examines how Microsoft's carbon neutrality claim, Google's clean energy matching, and Amazon's climate commitments are being strained by the massive electricity demands of AI training and inference. The piece ties into Google's 37% energy spike and broader industry trends.

Britain's biggest electronics retailer Currys is warning that smartphones and laptops will get more expensive later this year. AI data centers are hogging the world's memory chip supply, leaving gadget makers fighting over scraps. Apple and Xbox have already raised prices, and analysts expect things to worsen before they improve. Currys CEO says his company secured supplies ahead of time, but smaller retailers won't be so lucky.

Netflix's new reality show 'The Golden Ticket' will use AI to recreate the voice of the late Gene Wilder for a Willy Wonka series—with his estate's blessing. While Wilder's wife called it a celebration of his imagination, fans on social media called it 'disrespectful' and 'a plastic substitute,' with some jokingly comparing it to the disastrous 'Willy's Chocolate Experience' event. Studios appear to be testing what audiences will accept, following examples like Disney's digital recreation of James Earl Jones as Darth Vader. Experts say the more beloved the voice, the more scrutiny the AI version faces.

The AI boom is straining communities around data centers. In Lowell, Massachusetts, residents are fighting back after a facility installed cooling tanks behind homes, ran diesel generators that worsened air quality, and consumed 118,000 gallons of water daily. The city council voted 10-0 to freeze new data center construction for a year. Experts say heat waves make data centers harder to operate, requiring more power and water to stay cool. While researchers say the surge is 'very much a local phenomenon' rather than a national crisis, communities bearing the environmental costs are demanding accountability.

Y Combinator-backed Weave Robotics has launched Isaac 1, a home robot priced at $7,999 (or $449 monthly) that promises to do your laundry, make your bed, and tidy up your home. The launch post racked up over 13 million views on X, fueling speculation that household robots may finally be becoming mainstream—though critics call it 'slow' and 'clunky.' The company uses personal data to improve services, though it's unclear if home data trains the robot.

OpenAI's Sam Altman is proposing that AI companies like his own donate 5% of equity to a U.S. sovereign wealth fund, with other AI firms doing the same. The idea is to give everyday Americans a stake in the AI industry's growth and ease political pressure on big tech. A separate Senate bill from Bernie Sanders goes further, proposing a 50% tax on AI company stock to fund a $7 trillion public wealth fund. Both proposals face major hurdles, including needing congressional approval, and remain purely theoretical for now.

Popular TV-tracking app TV Time, with over 26 million lifetime installs, is shutting down on July 15, 2026. Its owner Whip Media—acquired by Blue Torch Capital in 2025—is pivoting to AI-focused products like Helix, an AI-powered workflow tool. The company will delete all user data. This follows a trend where AI priorities are causing companies to shut down beloved consumer apps, similar to Mozilla discontinuing Pocket.

A sponsored MIT Technology Review Insights report explores how companies are merging AI tools with traditional process management methods like Lean Six Sigma. The AI-powered process optimization market could exceed $113 billion in the next decade, with 88% of business leaders planning to increase investments in AI-infused process intelligence. The report argues that companies with strong existing process frameworks are best positioned to benefit from AI integration, while those without solid foundations risk wasting their investments.

Meta, the company behind Facebook and Instagram, spent roughly $145 billion on AI infrastructure in 2026 and now wants to rent out its excess computing power to other businesses. Think of it like a hotel that built too many rooms and is now opening bookings to strangers. The move has boosted Meta's stock by 9% as investors see a new revenue stream beyond ads. Smaller cloud players like CoreWeave and Nebius are watching nervously since Meta was once their customer and is now becoming their competitor.

Chinese AI startup Z.ai has launched ZCode, an AI coding tool that connects to multiple frontier models for writing, reviewing, and deploying code. It undercuts American competitors like Cursor and GitHub Copilot on price, with plans starting at $16.20 monthly compared to Cursor's $20 minimum. ZCode serves as the official development environment for Z.ai's GLM 5.2 open-source model.

Reports suggest the Trump administration and Anthropic have not discussed the government taking stakes in the AI company. This comes amid earlier reports that OpenAI was in talks about giving the US government a 5% stake, which Anthropic appears to not be pursuing.

OpenAI appears to be developing a small physical keyboard for coders to use with its Codex programming platform. The article suggests two reasons: branding (competing with Anthropic's momentum) and as a test run for larger AI hardware ambitions. The company is separately working with former Apple designer Jony Ive on more ambitious AI devices like pendants or desk accessories.

South African retailer Pick n Pay has launched an AI shopping assistant called Penny that lets customers build grocery orders using voice notes, text, or photos of handwritten lists and recipes. Powered by Google's Gemini AI, the assistant suggests recipes, meal plans, and substitutions while supporting multiple languages. It's the retailer's push to catch up with rivals in the competitive on-demand grocery market.

Anthropic reduced the system prompt for Claude Code by 80 percent with its Fable 5 models. The company found that newer models perform better with fewer instructions because they're more imaginative and can get confused by overly strict rules. Instead of detailed guidelines, Anthropic now steers behavior through context rather than explicit commands.

Anthropic, the AI company behind Claude, is reportedly in early talks with Samsung to manufacture custom AI chips. This follows OpenAI's similar move with its Jalapeño chip. The goal is to reduce reliance on Nvidia and cut the massive infrastructure costs of running AI systems. Anthropic still says Nvidia remains important to its operations.

SpaceX's $60 billion acquisition of AI coding startup Cursor raises big questions: will it keep offering OpenAI and Anthropic models, or will it become a captive tool for Musk's ecosystem? Cursor says it plans to stay open, but the deal puts SpaceX in direct competition with its own AI suppliers.

X Trending * [I wore the world's first HDR10 smart glasses](https://www.zdnet.com/article/rayneo-air-4-pro-hdr10-smart-glasses-ces/) * [TCL's new E Ink tablet beats the Remarkable and Kindle](https://www.zdnet.com/article/tcl-note-a1-nxtpaper-ces/) * [Anker's new charger is one of the most unique I've ever seen](https://www.zdnet.com/article/anker-wireless-chargers-ces-2026/) * [Best laptop cooling pads](https://www.zdnet.com/home-and-office/best-laptop-cooling-pad/) * [Best flip phones](https://www.zdnet.com/article/best-flip-phone/) * [Roku TV vs Fire Stick](https://www.zdnet.com/home-and-office/home-entertainment/roku-tv-vs-fire-stick/) * [Galaxy Buds 3 Pro vs Apple AirPods Pro 3](https://www.zdnet.com/article/apple-airpods-pro-3-vs-samsung-galaxy-buds-3-pro/) * [M5 MacBook Pro vs M4 Ma

Tesla has launched a longer Model Y in the U.S. starting at $61,990 with a six-seat configuration. The stretched version adds six inches to the wheelbase, giving much more room in the third row compared to the cramped seven-seater option. It comes with a year of Full Self-Driving capability included. Deliveries began in Malaysia last month and the vehicle is also available in China, Australia, and India.

AI startup Anthropic is in talks with Samsung to develop custom AI chips, joining rivals like OpenAI and Google in reducing dependence on Nvidia. The move signals a broader industry shift toward purpose-built silicon for AI workloads. Samsung already manufactures chips for Nvidia and is building an AI chip factory in South Korea with Nvidia.

Microsoft is spending $2.5 billion on a new unit called 'Frontier Company' that will embed 6,000 AI engineers directly inside enterprise clients' teams. Unlike typical software sales, these engineers will work on-site to integrate AI into the companies' actual workflows with measurable returns. Microsoft is pitching itself as platform-neutral, unlike competitors OpenAI and Anthropic who push their own models.

Google and Amazon just released sustainability reports showing their carbon emissions jumped 25% and 16% respectively — and AI is the main culprit. Both companies are building massive data centers to power AI services, consuming huge amounts of energy and water. Despite their net-zero pledges, the AI boom is making those climate promises harder to keep, with both firms now investing in natural gas plants to meet surging power demand.

Anthropic's Fable 5 AI model has shattered records on the Remote Labor Index (RLI) benchmark, hitting a 16.1% automation rate for real freelance work tasks — double the previous leader. The Center for AI Safety tested Fable 5 alongside OpenAI's GPT-5.5 and Anthropic's Opus 4.8, with Fable 5 dramatically outperforming both. The benchmark measures how often AI can complete economically valuable freelance projects (like graphic design, video work, and data analysis) at a quality a paying client would accept. While 16% is far from replacing all freelancers, the speed of improvement is striking — the benchmark leader's score quadrupled in under eight months.

A TechCrunch opinion piece argues that AI hype has officially peaked — using Jersey Mike's IPO filing as evidence. The sandwich chain, which sells submarine sandwiches, somehow found a way to mention 'AI' 22 times in its investor documents, including vague risk warnings about 'using AI Technologies in our business.' The piece uses this to illustrate how companies now feel compelled to shoehorn AI into any pitch to attract investor funding, even when their actual business has nothing to do with AI.

Meta has launched a new app called Pocket that lets users create and share interactive AI-powered "gizmos" built from text prompts. This has nothing to do with the Mozilla read-it-later service that shut down last year. Instead, it's Meta's latest move to put AI creation tools directly in people's hands — users can build little interactive experiences and share them with friends.

Boston Dynamics' newest Atlas humanoid robot is 10 times simpler to build than its predecessor, using far fewer parts while matching or exceeding its capabilities. The company aims to manufacture 30,000 units annually, with parent company Hyundai keeping most for its own factories. Boston Dynamics emphasizes its reasoning AI layer lets the robot adapt to new tasks without lengthy reprogramming. Hyundai, which owns Kia and Genesis, brings automotive-scale manufacturing expertise that could finally make humanoids affordable.

A Tesla driver in Texas faces manslaughter charges after his car using Full Self-Driving (FSD) struck and killed a woman inside her home. The driver claimed he was relying on FSD at the time. This case is significant because it tests who bears legal responsibility when autonomous driving features fail — the person behind the wheel or the carmaker. As more vehicles gain self-driving capabilities, this precedent could shape how such technology gets regulated and used worldwide.