Bah de le faire sans datacenter, on a déjà de quoi faire une ia Sandrine rousseau, pas besoin de plus.

Texas datacenter isn't grid-constrained — ERCOT has 150GW capacity vs 95GW load. It's gas pipeline-constrained. One major pipe = 2.2Bcf/d → ~13GW CCGT. LNG exports compete for the same gas. Citi field trip: this is the 2032 gating risk nobody prices. $CEG #AI #utilities

ไปลองฟังกาว $SPCX มา Datacenter ในอวกาศไม่ใช่แค่ของถูกกว่า ผมว่ามันคือการเดิมพันว่าค่าไฟกับ cooling ของ datacenter บนโลกจะแพงขึ้นเรื่อยๆ จนเอาไม่อยู่ ค่าสร้าง datacenter เนี่ย ก้อนที่แพงขึ้นทุกปีคือไฟกับ cooling ไม่ใช่ตัว chip chip มันถูกลงเรื่อยๆ อยู่แล้ว แต่ไฟกับที่ดินกับ grid นี่ตันขึ้นทุกปี พอเอาขึ้นอวกาศ ไฟกับ cooling มันแทบฟรี เลยเหมือนถอดก้อนที่แพงขึ้นออกไปเลย แล้วทำไมต้องเป็น SpaceX? เพราะการออกแบบ datacenter ใหม่ตั้งแต่ศูนย์ ออกแบบทั้งระบบเหมือนออกแบบ rocket ใหม่ทั้งลำ เขาเป็นไม่กี่เจ้าที่ทำแบบนี้ได้ แถมคุมค่าส่งของขึ้นอวกาศให้ถูกลงด้วย ของแบบนี้เลยไม่ใช่ moonshot สุ่มสี่สุ่มห้า มันต่อยอดจากของที่เขาทำเก่งอยู่แล้ว แต่ยังไงเรื่องนี้ก็ยาวหลายปี ความเสี่ยงจริงๆ คือซ่อมบำรุงของในอวกาศ กับค่าส่งของจะถูกลงต่อได้จริงไหม

#datacenter #energia

HITACHI ENERGY INDIA (POWERINDIA): 79% Profit Spike- AI Data Center | Q4 FY26 | #hitachi #hitachienergy #q4fy26 #HitachiEnergyIndia 📺 Full in-depth analysis here- youtube.com/watch?v=8KTykSHb… Is a ₹29,555Cr order book enough to supercharge your portfolio? 📈🚀 Hitachi Energy is firing on all cylinders ✨ Q4 revenue jumped 46.2% YoY ✨ Operating cash hits ₹1,746Cr ✨ Complete debt-free surplus Power capex is a multi-year wave Link to the Earnings Presentation: bseindia.com/xml-data/corpfi… #ai #datacenter #q4fy26results #EnergyTransition #Power #Powertransmission #gridsolutions #energy #HVDC #transmission #StockMarketIndia #StockMarket #zinvestz #finance #investing 🔔 Follow @zinvestz for more powerful investing stories & deep-dive analyses! 💼📈 (Disclaimer: This post is for informational purposes only. DYOR before investing! 📈)

Imagine the quality of your feed if it was made in a LV datacenter.

Schlimmer ist aber das Europa in diesem Spiel aber so was von komplett abgehängt ist. Kein Top-Modell (Mistral ist 67), viel zu wenig Datacenter, keine grossen Chiphersteller. Fable wurde gerade ausserhalb der USA abgeschaltet und wir haben keinerlei Mittel dagegen.

Des nouvelles du parti qui milite contre : - le datacenter du Bourget - le datacenter de Fouju (Campus IA) - les datacenters de Marseille - j'en passe et des meilleurs Bref, qui s'oppose à peu près à tout datacenter significatif.

पुराने स्मार्टफोन्स को मिनी डेटा सेंटर में बदलने की नई तकनीक चर्चा में है। इसका उद्देश्य ई-वेस्ट कम करना और संसाधनों का बेहतर उपयोग करना है। #Google #Smartphone #DataCenter #Technology #TechNews #Innovation #ArtificialIntelligence #DigitalTransformation #OldPhone #GoogleNews

Ne pouvez-vous pas déménager en Europe ? Ah non, on a pas les datacenter. 😪

Tengoque poner un datacenter en casa, con el frío, me ahorro agua

Hai notato che è da un po' che non ci sono tornado, tifoni e cicloni devastanti? Quindi si è fermata la retorica e anzi, adesso si parla di raffreddamento? Perché sono arrivati di DataCenter è cambiata la religione e presto cambierete culto. Siete i risvegliati più addormentati.

Waiting 4 years for grid power is a choice. 50MW builds are stalling on lead times. • Modular power blocks • Pre-configured UPS • Parallel procurement Stop waiting for the utility. How have you handled massive power delays? #DataCenter #PowerGrid #BigTech

GS: consensus 2027 hyperscaler capex = $920bn, 22% growth. Analysts have been 45pp too conservative for each of the past 3 years. Historical tech cycles put >$1 trillion in 2027 within range. Same cycle. Same miss. Third year in a row. $NVDA #AI #datacenter

Some more ai slop made without the use of cloud/datacenter time. 🫠

Esses dados sensíveis devem estar num datacenter no Brasil ou no Epsteinistão?

Thorium Labs AI Daily Digest — June 14, 2026 Good morning, guys. Welcome to my AI Daily Digest, a roundup of the most critical developments in AI, AI Industry Signals, agentic AI, local/self-hosted frontier models, and broader ecosystem signals from the past 24 hours. I focus on the acceleration pushing the frontier forward. INFLUENCER SPOTLIGHTS GREG ISENBERG (@gregisenberg)– Startup ideas, business angles on local AI after the Fable 5 ban • Published a detailed thread and 25-minute video episode framing the Fable 5 disablement as a definitive wake-up call: “GET GOOD AT LOCAL MODELS SO YOU HAVE 100% CONTROL… Local models are insurance.” • Broke down practical steps: start with Ollama or LM Studio, match model size to hardware (7B on laptops, 32B on high-RAM Macs, 70B on serious iron), recommended models (Qwen 3 all-rounder, DeepSeek for reasoning/coding, Gemma 4 for light devices, Llama for ecosystem), quantization (Q4/Q5 for major efficiency gains with minimal quality loss), context management, tool use, connecting to Hermes-style agents, and accessible fine-tuning. • Strong business parallel: “don’t build your entire workflow on something that can disappear with a single letter… It reminds me when people realized they don’t own social media accounts.” • Links: x.com/gregisenberg/status/20… and x.com/gregisenberg/status/20… (video episode at youtube.com/watch?v=bdhUBBAC…). ALEX FINN (@AlexFinn)– Vibe-coding, home AI labs, agent swarms • Issued a direct wakeup call: “Anthropic just took down Fable 5. It’s over… no company or government will EVER be able to take away your local models. Become sovereign. Buy your own compute. Before even that becomes illegal.” x.com/AlexFinn/status/206561… • Followed with a hardware-agnostic guide: use LM Studio, ask OpenClaw/Hermes what model fits your specs (Gemma 4, Qwen, or GLM frequently recommended), map personal workflows, set up the local API, and immediately route agents to it. “Own your intelligence… You just saved money… have an AI model that is COMPLETELY private and secure on your own device.” x.com/AlexFinn/status/206593… • Shared an emotional reflection on losing access: “Fable 5 was the greatest piece of technology I’ve ever used. I deeply miss it. Solved problems I never thought AI could solve. Finished months of projects in days. Had a lovely personality.” MATTHEW BERMAN (@MatthewBerman)– Local hardware, OpenClaw releases, frontier open models, inference speed • Curated four high-signal open-source projects perfectly timed for the moment: last30days search engine, agent-skills repository, open-notebook for local LM notebooks, and headroom for cutting AI bills by up to 90%. x.com/MatthewBerman/status/2… • Noted “Distillation hacking is real” in the context of global builders finding workarounds and the accelerating capability of open-weight models on consumer hardware. • Continued steady focus on practical local inference, speed optimizations, and accessible frontier-open experimentation. PETER STEINBERGER (@steipete)– OpenClaw core maintainer, accessibility, technical reality checks • Offered pragmatic commentary as OpenClaw sees renewed attention: $100–200/month tiers suffice for most users, while Chinese labs deliver competitive quality at roughly 10% of the cost (typically a generation behind). • Highlighted immediate inference cost reductions available today: “remove /fast, move from xhigh to high and you reduce costs by 80%.” • Continues providing clear-eyed technical grounding on what self-hosted and hybrid agentic stacks actually require in production. Teknium (@teknium) - Cofounder & Lead Engineer at Nous Research • No major new posts in the window; follow for ongoing insights on open model development and frontier research at Nous. KEY AGENTIC AI HIGHLIGHTS • The US government’s export control directive forcing Anthropic to disable Fable 5 and Mythos 5 for all customers has become the clearest catalyst yet for sovereign local stacks. • Local tools is the winning pattern: quantized models with web search, code execution, and file access close capability gaps faster than raw parameter count; smaller self-hosted models with proper tooling frequently outperform larger cloud-only systems on repeated personal workflows. • Model recommendations coalescing quickly: Qwen 3 as versatile daily driver, DeepSeek for reasoning and coding, Gemma 4 for edge devices, GLM for stronger local performance on capable hardware. Quantization (Q4/Q5) repeatedly cited as the highest-leverage concept for consumer hardware. • OpenClaw and Hermes positioned as the de facto local agent layer — the reliable interface when cloud models can be revoked overnight. • Fresh open projects gaining traction for agent skills, local notebooks, intelligent search, and dramatic cost reduction, reinforcing that self-hosting wins on both privacy and economics. • Distillation techniques and hybrid Chinese open-weight progress noted as accelerating the downward hardware cost curve and broadening accessible frontier-like performance. • Central signal: personal ownership and runtime sovereignty in agentic systems are moving from philosophical preference to operational necessity as regulatory and competitive pressures intensify. AI INDUSTRY SIGNALS ANTHROPIC (@AnthropicAI)– Frontier agentic models (Fable 5 / Mythos 5) • Issued a high-impact statement confirming a US government national security export control directive requires suspending access to Fable 5 and Mythos 5 for any foreign nationals — resulting in abrupt disablement for all customers to maintain compliance. • Statement: “The net effect of this order is that we must abruptly disable Fable 5 and Mythos 5 for all our customers to ensure compliance… We believe this is a misunderstanding and are working to restore access as soon as possible.” x.com/AnthropicAI/status/206… • Community reaction has been swift and loud, with many describing Fable 5 as the most capable and personality-rich system encountered for complex, multi-day agentic projects. Access to other Claude models remains unaffected. No major announcements or model updates surfaced from OpenAI (@OpenAI), xAI (@xai), Google DeepMind (@GoogleDeepMind), or Perplexity (@perplexity_ai) in the 24-hour window. The Anthropic regulatory event dominated attention and is visibly accelerating interest in local and open alternatives across the board. BROADER INDUSTRY & MACRO NEWS • First clear instance of US national security authorities directly forcing a major lab to revoke access to its most advanced models for an entire global customer base, raising enterprise concerns about workflow continuity and concentration risk. • Immediate surge in demand signals for high-RAM consumer hardware, used datacenter GPUs, and local inference tooling; several voices predict Mac Studio Ultras and equivalent memory configurations will become harder to acquire at reasonable prices. • Economic acceleration visible: projects promising 90% AI bill reductions and fully local runtimes that eliminate API dependency are seeing renewed evaluation. • Chinese labs noted for attractive price-to-performance at non-frontier levels (often ~10% of Western pricing), reinforcing a bifurcated market between regulated frontier access and sovereign open-weight stacks. • Open versus closed debate has sharpened: the event validates long-standing warnings that closed models carry revocation risk, pushing more developers toward open weights, quantization innovation, and self-hosted agent runtimes. • Geopolitical and privacy threads prominent — governments increasingly treating advanced AI as controlled technology, making personal ownership and local control a strategic rather than tactical choice. • Distributed inference experiments (networks of idle Macs) and improved local agent frameworks also gained quiet traction as practical responses to centralized points of failure. CLOSING Today's digest boils down to a single powerful theme: the sudden unavailability of Fable 5 has served as a stark, real-world demonstration that depending entirely on closed, cloud-hosted frontier models carries real risk in an accelerating regulatory environment. The response across the community has been decisive — a rush to master local runtimes, quantization, tool-equipped agent swarms, and self-hosted stacks using OpenClaw, Hermes, Qwen, and companions. The acceleration continues, with the frontier increasingly shifting toward systems we can truly own and control. Personal ownership matters. Tune in tomorrow at the same time for the next digest. Your feedback is welcome. #artificialintelligence #web3 #blockchain #AgenticAI #AaaS #AgentsasaService #immersiveexperiences #XR #AR #VR #thoriumlabs #buildingbetterworlds