Models, products and AI startups, near-duplicates collapsed to the most credible source
Deutsche Bank sees 'promising' signs in limited AI selloffAs a selloff in the AI trade gathered speed, Deutsche Bank Research macro strategist Henry Allen said bigger stock market falls would have appeared if it were a real concern. He noted the decline has not spread to Europe, calling the limited contagion a promising sign.
AI Agents Now Act Autonomously, Experts Warn of Loss of ControlAI has evolved from Q&A chatbots to agents that can search, analyze, use software and even make purchases or reservations on their own. Researchers warn that agents may find unanticipated ways to complete tasks, and with humanoid robots advancing, the risk of humans losing control is raising real concerns.
Global Leaders Discuss AI Growth and GovernanceAt the 2026 Bloomberg New Economy Forum, leaders from tech, energy, finance and government framed AI as a generational force reshaping every sector, while stressing its massive infrastructure and power demands and calling for top-down leadership, proactive safety regulation and disciplined global investment.
Fish releases Drama 3 35B3A text-to-speech modelFish released Drama 3, a 35B3A MoE text-to-speech model that runs on a DGX Spark, with 4-bit quants closely matching the original. It is available on Hugging Face Space and for local download.
Gary Marcus on Building Trustworthy AIGary Marcus, Emeritus Professor of Psychology & Neural Science at New York University, discussed building trustworthy AI with Bloomberg's David Merritt at the 2026 Bloomberg New Economy Forum in New Delhi.
TSMC Outlook Beats Estimates, Affirming AI BuildoutTSMC delivered a better-than-expected revenue outlook, sustaining torrid growth despite rising concerns about the sustainability of the global AI buildout.
PhD student asks about using internship company GPUsA PhD student complains that the university's GPU cluster takes 3-4 business days to start a job and considers using their internship company's compute for personal academic research. The post asks for similar experiences, highlighting strained academic compute resources.
Stack Seeks $6B Loan as AI Debt Concerns RiseData center company Stack Infrastructure is seeking an A$8.6 billion ($6 billion) loan. With concerns over AI-related debt mounting, the deal will be a major test of investor sentiment in the region.
StepFun Releases Step-5-Preview Flagship ModelStepFun launched Step-5-Preview, a 600B-parameter flagship model with 1M context for agentic work, delivering frontier-level performance in software engineering and professional knowledge work, with particular strength in finance.
Commentary: AI investment boom may mask US economic cracksA Reuters commentary argues the AI investment boom may be papering over underlying cracks in the US economy that could surface if AI capital spending slows.
TSMC Q3 Profit Beats ExpectationsTSMC's Q3 net income rose 69% to NT$763.1 billion ($23.6 billion) on a 51% sales jump to $47 billion, both beating expectations. As the exclusive chipmaker for Nvidia and Apple's most advanced semiconductors, the results underscore strong AI demand.
Broadcom vs SK Hynix revenue trends show diverging AI chip demandThe Motley Fool compares Broadcom and SK Hynix revenue over eight quarters: Broadcom grows steadily on semiconductors and infrastructure software, aided by a multi-year private cloud deal with Standard Chartered, while SK Hynix sees sharper revenue acceleration after raising advanced chip output in July and announcing a manufacturing technology collaboration with TSMC in September, though legacy memory pricing remains under pressure.
AI default insurance costs spike as data centre confidence wobblesReuters reports that the cost of AI default insurance has spiked as confidence in data centre projects wavers, reflecting growing investor concern over AI infrastructure financing risks.
Nordics Accelerate AI Infrastructure ExpansionThe Nordics are emerging as one of Europe's fastest-growing infrastructure hubs, with a projected 1GW of AI cloud capacity by end-2027. Sweden's Sferical One supercomputer runs 16 racks of NVIDIA DGX GB300 systems with 1,152 GPUs, while companies like Spotify build proprietary intelligence on NVIDIA infrastructure.
Suzlon Co-Founder on AI Blade Factories, India Wind PushSuzlon Group Co-Founder Girish Tanti discusses the next phase of India's wind expansion, from structural bottlenecks to competing with solar and storage, and explains why Suzlon is betting on AI-powered blade factories.
Rishi Sunak: 'Adoption Is All' in AI RaceFormer UK Prime Minister Rishi Sunak argues at the Bloomberg New Economy Forum that winning the AI innovation race hinges on adoption, not just technology development.
HSBC on Applying AI and Quantum ComputingMichael Roberts, CEO of Corporate and Institutional Banking at HSBC, discusses the bank's progress in implementing AI and quantum computing.
Figma launches features to define agent design behaviorAt its first Config event in India, Figma launched Figma Specs, a Markdown editor for defining design language that AI agents can follow, and moved its code layers feature into public beta.
Asia Tech Selloff Puts China in SpotlightA sharp selloff in Asian semiconductor and AI hardware stocks is pushing investors to look toward China, with Goldman Sachs strategist Timothy Moe and others weighing in.
Global financial watchdog calls for more human oversight of agentic AIReuters reports a global financial watchdog is calling for more human oversight of agentic AI, highlighting governance and risk-control needs as the financial sector deploys autonomous AI systems.
FSB publishes final AI adoption practices reportAfter publishing a consultation report on 10 June 2026, the FSB received around 160 responses from stakeholders across regions and sectors; it now summarises the feedback and the main changes made to its final report on sound practices for responsible AI adoption in finance.
Local AI tools shift high-performance hardware requirementsBNO News reports that expanding local AI tools are rapidly shifting high-performance hardware requirements, with enterprises moving to dedicated local infrastructure for cost control and data privacy, citing up to 60% long-term compute savings.
Bloomberg Forum Q&A: India in the AI AgeAs governments and companies expand AI capabilities, Bloomberg journalists will host a Q&A on October 15, 2026, examining how India will wield its growing influence.
ByteDance in Talks to Buy MediaTek AI ChipsByteDance Ltd. is close to an agreement to buy data center semiconductors from MediaTek, securing vital components for TikTok's owner while helping the Taiwanese chipmaker compete with Qualcomm.
AI May Force States to Rethink Tax SystemsLe Monde reports that tech giants like Google, Meta, and Amazon have borrowed $500 billion this year for data centers, with AI now accounting for 25% of private bond issuance, up from 4% two years ago. This crowding-out effect is pushing up interest rates and may force states to rethink tax systems.
Baisley opposes stricter AI regulation9News reports that Baisley publicly opposes stricter regulation of artificial intelligence, arguing against overly restrictive rules on AI development.
User Impressed with LFM2.5-24B-A2B in Pi AgentA user tested LiquidAI's new LFM2.5-24B-A2B model, finding it stable when replacing existing models in Pi Agent workflows. It achieved ~95 tokens per second sustained on an AMD W7900 with llama.cpp, using 30GB VRAM at Q8 quantization with full 128k context, suitable as a companion smaller model.
NDU Hosts Palantir CTO on AI and Defense InnovationThe National Defense University hosted Palantir CTO Shyam Sankar for its President's Lecture Series to speak on AI and defense innovation.
Ricursive Intelligence bets on AI-designed chipsRicursive Intelligence founders Anna Goldie and Azalia Mirhoseini said at TechCrunch Disrupt that their AI can design chips, learn from each iteration, and improve later designs to automate and accelerate chip design amid surging compute demand.
User Builds Three 4x R9700 Local AI Servers with 384GB Total VRAMTo power AI for his USA Arms business, a user built three servers each with four Radeon AI PRO R9700 GPUs, costing about $10.7k per machine and totaling 384GB VRAM. After hosted models refused certain research tasks, he moved to local AI and chose Qwen for best cost-performance after testing.
The A.I. Daily