מינוי Plus

התחברות ל-baba News

חשבון אחד לאתר, לאייפון ולאנדרואיד — המנוי נשאר איתכם.

או קוד במייל

חיפוש סיפורים

הקלידו לפחות שני תווים. התוצאות מגיעות מכל המערכות ש-baba קורא.

↑↓ למעבר · ↵ לפתיחה · esc לסגירה

התחברות ל-baba News

התחברו כדי להמשיך לשאול. News Plus מסיר את המגבלה היומית.

או קוד במייל

TheMarkerטכנולוגיה

Israel Electric Corporation Adapts Grid for AI Data Centers

תומצת מתוך TheMarker על ידי baba

עסקים

עברית · מערכת חדשות אחת

The Israel Electric Corporation is adapting its power grid to meet the surging electricity demands of artificial intelligence data centers. Global server farm energy consumption is expected to more than double by 2030, with AI queries consuming significantly more power than standard searches. The IEC faces challenges from rapid technological change and extreme, unpredictable load fluctuations from AI farms, which can destabilize the grid. The corporation is shifting to continuous risk management and real-time operational adjustments to ensure reliable power quality.

הידיעה ב-5 שורות · תקציר של baba

  • Global server farm electricity use to more than double by 2030, reaching 945 TWh.
  • AI queries consume ten times more energy than standard internet searches.
  • AI server farms cause rapid load surges from 80 to 100 MW in milliseconds.
  • IEC is shifting to continuous risk management and real-time grid adjustments.
  • Adapting the grid for AI, cloud, and crypto is a central IEC objective.

The burgeoning demand for artificial intelligence services is creating a new reality for Israel's electricity grid. Eli Elisha, Head of the Distribution Engineering Division at the Israel Electric Corporation (IEC), stated at a recent conference that global electricity consumption by server farms is projected to more than double from 415 terawatt-hours in 2024 to 945 terawatt-hours by 2030. An AI-based query consumes approximately ten times more energy than a standard search. Elisha explained that the IEC is preparing for this dramatic increase, recognizing the sector's importance to the economy and Israel's technological standing. He highlighted the challenge of long energy infrastructure development timelines, often spanning years or decades, contrasting with the rapid evolution of technology. This disparity necessitates a shift in planning and management approaches.

The challenge extends beyond sheer consumption volume to its nature. Unlike traditional cloud services or crypto mining with relatively predictable usage patterns, AI model training in server farms causes sharp, extreme fluctuations. Elisha illustrated this by noting that AI farm loads can surge from 80 to 100 megawatts in milliseconds, potentially impacting other connected consumers. Such rapid increases can destabilize the grid, affecting not only supply reliability and continuity but also power quality, with voltage fluctuations and waveform distortions potentially disrupting sensitive electronic equipment. The complexity of server farms, relying on a chain of components including transformers, UPS systems, rectifiers, and cooling systems, further complicates operations.

Elisha emphasized that these challenges require a conceptual change in grid management. The traditional approach of upfront planning and connection checks is insufficient. The new reality demands continuous risk management, involving data collection, active monitoring of consumption patterns, anomaly detection, and real-time operational adjustments. He stressed that this cannot be a "set it and forget it" process and underscored the importance of communication with farm operators. Proposed solutions include limiting load change rates, energy storage, interference filtering, and maintaining operational margins within the grid. Elisha concluded that adapting the electricity grid to the demands of cloud, crypto, and AI infrastructure is a central objective for the IEC, requiring existing measures to be tailored to server farm characteristics and standards to be updated, while recognizing that standards are a foundation, not a substitute for risk management.

TheMarkerאחרים · תל אביב

התחברות ל-baba News

התחברו כדי לעקוב אחרי מערכות, נושאים ואנשים.

או קוד במייל

התחברות ל-baba News

התחברו כדי לעקוב אחרי מערכות, נושאים ואנשים.

או קוד במייל

מוזכרים

פתיחת המבזקים