🎥In the cold north, power flows like water, as digital infrastructure begins to breathe with the landscape. Built not only for today’s compute, but for the intelligence of tomorrow — where energy, cooling, and AI converge as one. #AIInfrastructure# #DigitalInfrastructure# #DataCenter# #HPC# #AIDataCenter# #Norway#

🦾 Why Data, Not Models, Is the Real Moat in Embodied AI The timing of this question is hard to miss. This week, the World Humanoid Robot Games released a 2,500+ hour dataset covering 12 scenario categories, 44 operations, and more than 10,000 tasks. Crucially, it also includes failures and edge cases. But raw hours tell only part of the story. Zhihu contributor 于超, an assistant professor at Tsinghua Shenzhen International Graduate School, shares his team’s view on why data has become embodied AI’s hardest-to-replicate advantage. 1️⃣ Robot scaling is fundamentally asymmetric Like language models, robot policies appear to benefit from more data, larger models, and greater compute. But these three inputs do not scale equally. Model architectures can be studied and reproduced quickly. General-purpose compute can, in principle, be purchased. High-quality robot data is different. It must be accumulated through physical interaction, and competitors cannot recreate it overnight. That asymmetry is what turns data into a moat. 2️⃣ Robot data must be manufactured LLMs inherited decades of internet data. Robots did not. Every useful trajectory must be produced through physical interaction. Even a simple cup-grasping task changes with the object, lighting, environment, camera angle, and robot body. So raw hours are not enough. What matters is the diversity of embodiments, tasks, objects, failures, and recoveries. Open X-Embodiment needed more than 20 institutions and 22 robot platforms to collect over one million trajectories. DROID used 50 collectors for a year, producing only 350 hours of data. New methods such as UMI and egocentric recording make collection easier. But every hour still requires real people, equipment, and time. 3️⃣ A successful trajectory can still be bad data Robot data quality is more complicated than whether a task was completed. On the hardware side, camera accuracy, encoder readings, force sensors, calibration, communication latency, and synchronization across modalities can all corrupt a trajectory. The human operator adds another source of noise. Teleoperating a robot is not the same as performing the action directly. Operators hesitate, pause, readjust, and develop habits for compensating for the control system. A task may succeed even when parts of the demonstration should never be imitated. Success is therefore only the coarsest possible label. One trajectory can contain both excellent behavior and inefficient or misleading actions. Training on the entire trajectory without distinction effectively tells the robot to learn both. 4️⃣ The next challenge is information density Collecting more trajectories is only half the problem. Teams must also identify which parts are worth learning from. Yu Chao’s team developed STEAM to detect local progress within a trajectory without frame-by-frame annotation or manually designed rewards. It separates useful progress from hesitation, failure, and recovery. The key question is shifting from “How many trajectories do we have?” to “How much useful information does each trajectory contain?” 5️⃣ Embodied data is physically expensive Text can be copied. Videos can be downloaded. Robot data requires a physical production process. Collecting one hour may involve a robot, sensors, teleoperation equipment, an operator, a suitable environment, task materials, and engineers who maintain and calibrate the system. Real factories, stores, and homes add even more complexity. And pressing the record button is only the beginning. Transmission, cleaning, governance, and storage can cost more than collection itself. The author offers a rough calculation. If a company wants one million hours of real-world data and reduces the combined collection and management cost to RMB 200 per hour, the total still reaches RMB 200 million. 🔑 The real moat compounds over time Quantity, quality, and cost explain why embodied AI data cannot be replicated through a short burst of spending. Large, diverse, high-quality datasets require physical infrastructure, operational discipline, and years of accumulation. As robot policies continue to benefit from scaling, the durable advantage will belong to teams that can repeatedly: 🔹 Collect broader real-world experience 🔹 Identify the most informative behavior 🔹 Preserve failures and recovery signals 🔹 Turn noisy trajectories into useful learning data In embodied AI, having data and knowing how to use it are becoming two very different capabilities. And the second may ultimately matter even more than the first. 🔗 Full analysis: #EmbodiedAI# #Robotics# #PhysicalAI# #RobotLearning# #AIData# #ScalingLaw# #Tsinghua#

AI DATA CENTER SHORT SELLERS GAIN $6 BILLION IN OCTOBER Short sellers have made $6B in paper profits betting against AI infrastructure stocks amid concerns over debt-funded expansion, according to ORTEX. Bets against $NBIS generated $2.94B, $CRWV nearly $1.7B, and $IREN and $APLD over $1.2B combined. Around $17B in shares across five AI infrastructure companies remain sold short.

AI data centers are about to explode 📈 Bottom-up analysis shows ~212GW of incremental capacity needed from 2026–2030. Annual deployment: • 2025: 13.0 GW • 2030E: 55.7 GW Nvidia still dominates, but AMD, Google TPU, AWS Trainium, Microsoft Maia, Meta MTIA & OpenAI are all scaling fast.

AI CONFERENCES AND INVESTOR DAYS, THURSDAY, OCTOBER 15, 2026 YESTERDAY: INVESTOR DAY TAKEAWAYS ACCENTURE PLC INVESTOR DAY Wed Oct 14, New York Replay: • New FY27 to FY29 framework from the 8-K financial presentation: mid-single digit revenue growth in local currency in FY27 and FY28, stepping up to mid-to-high single digits in FY29, with 10 to 30 bps of annual operating margin expansion; FY27 guidance reaffirmed at 3% to 6% local currency revenue growth, with acquisitions adding about 2% to 2.5% • FY26 base: revenue $74.2B, adjusted EPS $13.97, adjusted operating margin 15.8%, free cash flow $11.6B, bookings $84.5B for a 1.1 book-to-bill, and a record $11.5B returned to shareholders; about $5.5B of buybacks expected in FY27 and at least $9.5B of shareholder returns guided for FY27 • AI answer to the bear case: a $3 trillion addressable market was framed, a joint forward deployed engineering program with Workday for agentic AI in HR and finance was launched, and the UK customer and data intelligence firm Sagacity was acquired for Accenture Song; the Anthropic partnership announced Sep 18, at least $1B from each side over 5 years, was the backdrop • Stock reaction: shares rose about 4% premarket and about 2% intraday, then faded to close up 0.7% at $214.96; the stock is still down about 20% year to date after the roughly 16% rally on the Oct 1 results • Analyst reaction: UBS reaffirmed Buy with a $275 target, citing doubled emerging alliance revenue and more than tripled Q4 bookings; the cautious read is that the targets are an ambition, not proof, and FY27 has to deliver the first leg SAILPOINT INC INVESTOR DAY Wed Oct 14, 12:00 PM ET, virtual Replay: • No guidance or financial update, by design: the Oct 14 analyst and investor webcast was a product recap of Navigate, Austin Oct 5 to 8, and management said at the outset it would not discuss the financial impact of the products; replay is on the IR site for 1 year • The recap centered on autonomous identity security for the agentic era: three classes of Autonomous Agents that monitor runtime actions, enforce policy and right-size privileges, plus Autonomous Identity Security Posture Management for continuous exposure detection and real-time remediation in Atlas • Expanded Agentic Fabric maps the full AI agent surface area, with shadow AI discovery, runtime authorization and an instant kill switch; Human Fabric advances human identity governance; Zero Standing Privilege via just-in-time provisioning now covers humans, service accounts and AI agents • The demand framing: 79% of enterprises are already running AI agents in production, yet only 2% have deployed identity tools purpose-built to govern them, per the Horizons of Identity Security report • Stock reaction: shares closed down 0.4% at $19.30 on the day; Piper Sandler maintained Buy with a $25 target on Oct 14, with consensus at Buy and an average target near $21.88 TODAY SEMTECH CORP INVESTOR DAY Thu Oct 15, 9:00 AM PT, San Jose Webcast: • Data Center Analyst Day and Teach-In with CEO Dr. Hong Hou, Imran Sherazi of Signal Integrity Products, and CFO Mark Lin, plus an industry expert panel, Q and A sessions and product demos; live webcast and materials under News and Events on the IR site, replay for 90 days • Setup: shares have pulled back about 17% in the last 5 sessions into the event after a very strong year, so the bar is a credible long-term model, not a product tour • The number: the long-term financial objectives presented today, against the near-term anchor of the $410M Q3 revenue guide • Why it matters: FiberEdge, CopperEdge and DirectEdge sit in the 800G and 1.6T AI data center interconnect buildout, where active copper, linear pluggable optics and co-packaged optics timing decides share • Debates: how fast active copper cable ramps at the lead cloud customer, and what reserved capacity costs if demand slips • Watch: 1.6T and 3.2T roadmap, active copper cable customer wins and the Q4 ramp, data center revenue mix, and any gross margin framework SEMTECH CORP CONFERENCE PRESENTATION Thu Oct 15, 12:00 PM, San Jose, time unconfirmed Webcast: • A separate data center business listing for the same San Jose teach-in covered above • The company release confirms a 9:00 AM PT start; a separate 12:00 PM session time is unconfirmed TOMORROW No tracked events tomorrow. REMAINING THIS WEEK Nothing remaining this week. LATER • Mon Oct 19: Qnity Electronics at ICPT, Gangseo, Korea • Tue Oct 20: Zoom Communications Investor Day, 4:00 PM, webcast on Zoom IR • Fri Oct 23: Tempus AI at ESMO 2026 • Mon Oct 26: DCD Connect MENA, Dubai, Oct 26 to 27 • Tue Oct 27: Celestica Investor Day, virtual, tied to Q3 • Tue Oct 27: DCD Investment Forum MENA, Dubai • Tue Oct 27: Guidewire Software Investor Day, 7:30 PM • Wed Oct 28: Oracle Investor Day • Tue Nov 3: DCD Connect Virginia, Leesburg, VA, Nov 3 to 4 • Wed Nov 4: DCD Connect Brasil, DCD Investment Forum Brasil, and Xcelerated Compute Show Brasil, Sao Paulo, Nov 4 to 5 • Thu Nov 5: Box Investor Day, San Francisco • Fri Nov 6: Global Unichip at TSMC 2026 Japan OIP Ecosystem Forum, Tokyo • Mon Nov 9: Entegris Investor Day, 1:00 PM • Tue Nov 10: Cohu Investor Day, New York • Tue Nov 10: Flex Investor Day, Austin • Wed Nov 11: Munters Group Capital Markets Day, Stockholm • Thu Nov 12: Global Unichip at TSMC 2026 Taiwan OIP Ecosystem Forum, Hsin-chu • Sun Nov 15: Penguin Solutions at SuperComputing (SC26), Chicago • Sun Nov 15: SuperComputing Conference 2026 (SC26), Chicago, Nov 15 to 20 • Sun Nov 15: Vertiv at SuperComputing (SC26), Chicago • Tue Nov 17: Global Unichip at TSMC 2026 China OIP Ecosystem Forum, Beijing • Tue Nov 17: XTI Aerospace at ElevateUAV Summit, Florida City • Thu Nov 19: KVH Industries at CSN Greece ICT Conference, Athens

AI data centers are installing new cooling technology that saves water locally, but their rising electricity use is increasing water consumption at the power plants that feed them

AI data centers are installing new cooling technology that saves water locally, but their growing electricity use is increasing water consumption at the power plants that feed them

AI Conferences and Investor Days, Friday, October 9, 2026 TODAY No tracked events today. TOMORROW No tracked events tomorrow. REMAINING THIS WEEK Nothing remaining this week. LATER OCP GLOBAL SUMMIT 2026 Mon Oct 12 to Thu Oct 15, San Jose McEnery Convention Center. Theme: "Scaling Innovation for the AI Era". Sold out; registration closed with no onsite signup. No public livestream found; keynote and session recordings expected on the Open Compute Project YouTube channel after the event: Keynotes Mon Oct 12 (PT): AMD Chair/CEO Lisa Su, 4:15pm, open full-stack infrastructure for agentic AI; Broadcom's Asad Khamisy (SVP, Core Switching), 5:30pm, "Scaling AI Infrastructure Through Open Innovation"; NVIDIA's Ian Buck (VP Hyperscale/HPC), 5:55pm, "AI Factories at Scale"; Meta keynote (speaker and time unconfirmed). Themes to watch: rack-scale AI factory architectures (Vera Rubin NVL72, MGX, DSX); Ethernet vs InfiniBand for AI scale-up (Broadcom Tomahawk 6, Thor Ultra 800G, OCP ESUN vs NVIDIA Spectrum-X); power delivery and thermals (800V DC racks, liquid cooling); optical interconnect (CPO, 448G signaling). 42 tracked company sessions (Oct 12-13; times ET): AMD, NVIDIA, Broadcom, Meta, Alphabet, Arm, Marvell, Astera Labs, Credo, Lattice, SK hynix, Samsung, MediaTek, Intel, Cisco, HPE, Dell, Microsoft, Oracle, Super Micro, Vertiv, Eaton, Nokia, TE Connectivity, Celestica, Flex, Jabil, Sanmina, Wiwynn, Pegatron, Mitac, Foxconn Industrial Internet, Inventec, Keysight. Also: Lattice/AMI keynote Tue Oct 13, 10:00am PT, "The Open Control Plane: From AI Racks to Autonomous Infrastructure" (Ford Tamer, Sanjoy Maity). APPLIED MATERIALS INVESTOR DAY (INVESTOR BREAKFAST AT SEMICON WEST) Tue Oct 13, 8:00-10:30am PT. Webcast: replay same day. Setup: executive leadership team presents during SEMICON West. The number: the 2027 WFE (wafer fab equipment) growth outlook. Why it matters: AMAT is the largest WFE vendor; its outlook frames the 2027 capex cycle and AI-driven fab capacity plans. Watch: a firm 2027 WFE growth figure; commentary on AI-driven capacity adds. WORKDAY FINANCIAL ANALYST DAY Tue Oct 13, 2:00pm PT, Las Vegas (during Workday Rising). Webcast: replay 90+ days. Setup: analyst day alongside the annual customer conference. The number: Agentic AI ARR ~$600M (+200% YoY). Why it matters: the multi-year AI monetization framework sets expectations for converting the installed base to AI products. Debates: seat-based vs consumption (FlexCredits) pricing durability. Watch: FlexCredits disclosure; FY28 subscription growth vs the ~11% framework. ACCENTURE INVESTOR AND ANALYST CONFERENCE Wed Oct 14, 8:30-11:30am ET, virtual. Setup: shares +23% on the Q4 print (~$84.5B FY bookings), faded ~6% since. The number: refreshed long-term growth and margin targets. Why it matters: bellwether for enterprise IT services demand and AI-led consulting. Debates: fixed-price mix; granularity of AI disclosure. Watch: long-term targets; AI bookings disclosure. SAILPOINT INVESTOR DAY Wed Oct 14, 12:00pm ET, virtual. Webcast: replay 1 year. Setup: recap of product introductions from Navigate (annual identity security conference, Austin, Oct 5-8). The number: not a guidance event; product-led. Why it matters: identity security is consolidating; SailPoint's AI positioning against Microsoft and Palo Alto Networks. Watch: genuinely new capability vs repackaging; AWS/NVIDIA pipeline commentary. SEMTECH DATA CENTER TEACH-IN (ANALYST DAY) Thu Oct 15, 9:00am PT, San Jose. CEO Hong Hou, Imran Sherazi (Signal Integrity GM), CFO Mark Lin. Webcast: Setup: stock at 52-week highs; past ACC timing wobbles have been punished. The number: 1.6T/800G and ACC (active copper cable) ramp timing. Why it matters: key read on optical DSP and copper interconnect demand in AI clusters. Watch: ACC ramp milestones; long-term financial objectives. THINKEQUITY CONFERENCE Thu Oct 15, Mandarin Oriental, New York. Single-day small-cap broker conference. AI-relevant presenters include Roadzen (RDZN, 9:30am ET) and Mobilicom (drone track, 1:30pm ET). WEEK OF OCT 19 AND BEYOND • Qnity Electronics (Mon Oct 19, ICPT planarization/CMP conference, Gangseo): semiconductor materials read. • Zoom Communications Investor Day (Tue Oct 20, 4:00pm, timezone unconfirmed). • Tempus AI (Fri Oct 23, presenting at ESMO 2026): AI healthcare name at a medical oncology conference. • Celestica Investor Day (Tue Oct 27): AI data center EMS; key read on AI server build demand. • Guidewire Investor Day (Tue Oct 27, 7:30pm, timezone unconfirmed): insurance software. • Oracle Investor Day (Wed Oct 28): OCI and AI infrastructure. Also tracked: semicap names at the 18th Annual CEO Investor Summit (Tue Oct 13, California): ACM Research, Cohu, Photronics; Kulicke & Soffa (CEO Summit, San Francisco); BE Semiconductor (SEMICON West); Zhen Ding (20th QIC CEO Week, Hong Kong). Wolfspeed at the Wolfe Research quant/macro conference (Wed Oct 14, New York). Unusual Machines at the Maxim Growth Summit (Mon Oct 12, New York).

AI data centers are fueling optical interconnect demand! #VCSEL# sales are projected to hit 650 million units by 2031, growing at a 38% CAGR from 2025 to 2031 as major players race to seize opportunities.💡#TrendForce# on optical communications and NPO: 🔗