Artwork

תוכן מסופק על ידי Evan Kirstel. כל תוכן הפודקאסטים כולל פרקים, גרפיקה ותיאורי פודקאסטים מועלים ומסופקים ישירות על ידי Evan Kirstel או שותף פלטפורמת הפודקאסט שלהם. אם אתה מאמין שמישהו משתמש ביצירה שלך המוגנת בזכויות יוצרים ללא רשותך, אתה יכול לעקוב אחר התהליך המתואר כאן https://he.player.fm/legal.
Player FM - אפליקציית פודקאסט
התחל במצב לא מקוון עם האפליקציה Player FM !

Scaling the Digital Backbone: How Data Centers Are Evolving for AI

25:54
 
שתפו
 

Manage episode 508084566 series 3499431
תוכן מסופק על ידי Evan Kirstel. כל תוכן הפודקאסטים כולל פרקים, גרפיקה ותיאורי פודקאסטים מועלים ומסופקים ישירות על ידי Evan Kirstel או שותף פלטפורמת הפודקאסט שלהם. אם אתה מאמין שמישהו משתמש ביצירה שלך המוגנת בזכויות יוצרים ללא רשותך, אתה יכול לעקוב אחר התהליך המתואר כאן https://he.player.fm/legal.

Interested in being a guest? Email us at [email protected]

The digital infrastructure powering artificial intelligence is undergoing a radical transformation as companies race to build ever-larger models. Edgecore's Tom traugott reveals how this shift is forcing a complete reimagining of data center design, power requirements, and cooling technologies.
What began as traditional cloud facilities has evolved into massive AI-optimized campuses. Edgecore's Mesa, Arizona project dramatically shifted from six smaller buildings to three massive structures, each capable of housing tens of thousands of GPUs. The physical constraints of modern AI hardware - particularly NVLink's one-meter distance limitation - have accelerated the adoption of direct-to-chip liquid cooling from niche application to industry baseline. As Tom explains, "Infrastructure becomes a competitive advantage" in this new landscape.
Power availability has emerged as the critical chokepoint. The staggering scale of new AI investments (Oracle-OpenAI's deal alone could require 4.5 gigawatts - four times Denver's consumption) raises profound questions about grid capacity. The industry is exploring creative solutions, including "curtailment for capacity" programs where data centers leverage their backup generation capabilities during peak demand to stabilize the grid while unlocking additional capacity.
Networking infrastructure has transformed from a relatively minor consideration to a critical and expensive component. Deploying a 100,000 GPU cluster requires approximately 1.8 million cable terminations and a small army of technicians. Even optical transceivers now consume significant power - up to 40 megawatts for large clusters, equivalent to the computing power of 20,000 GPUs.
Join us for this fascinating exploration of how the backbone of our AI revolution is being built, the challenges facing the industry, and the innovative solutions emerging at this crucial technological frontier. Discover how these massive infrastructure investments are creating what Tom calls "electroalchemy" - the transformation of energy into intelligence through the medium of silicon.

Support the show

More at https://linktr.ee/EvanKirstel

  continue reading

פרקים

1. Scaling the Digital Backbone: How Data Centers Are Evolving for AI (00:00:00)

2. [Ad] Crossing Borders (00:17:31)

3. (Cont.) Scaling the Digital Backbone: How Data Centers Are Evolving for AI (00:18:07)

518 פרקים

Artwork
iconשתפו
 
Manage episode 508084566 series 3499431
תוכן מסופק על ידי Evan Kirstel. כל תוכן הפודקאסטים כולל פרקים, גרפיקה ותיאורי פודקאסטים מועלים ומסופקים ישירות על ידי Evan Kirstel או שותף פלטפורמת הפודקאסט שלהם. אם אתה מאמין שמישהו משתמש ביצירה שלך המוגנת בזכויות יוצרים ללא רשותך, אתה יכול לעקוב אחר התהליך המתואר כאן https://he.player.fm/legal.

Interested in being a guest? Email us at [email protected]

The digital infrastructure powering artificial intelligence is undergoing a radical transformation as companies race to build ever-larger models. Edgecore's Tom traugott reveals how this shift is forcing a complete reimagining of data center design, power requirements, and cooling technologies.
What began as traditional cloud facilities has evolved into massive AI-optimized campuses. Edgecore's Mesa, Arizona project dramatically shifted from six smaller buildings to three massive structures, each capable of housing tens of thousands of GPUs. The physical constraints of modern AI hardware - particularly NVLink's one-meter distance limitation - have accelerated the adoption of direct-to-chip liquid cooling from niche application to industry baseline. As Tom explains, "Infrastructure becomes a competitive advantage" in this new landscape.
Power availability has emerged as the critical chokepoint. The staggering scale of new AI investments (Oracle-OpenAI's deal alone could require 4.5 gigawatts - four times Denver's consumption) raises profound questions about grid capacity. The industry is exploring creative solutions, including "curtailment for capacity" programs where data centers leverage their backup generation capabilities during peak demand to stabilize the grid while unlocking additional capacity.
Networking infrastructure has transformed from a relatively minor consideration to a critical and expensive component. Deploying a 100,000 GPU cluster requires approximately 1.8 million cable terminations and a small army of technicians. Even optical transceivers now consume significant power - up to 40 megawatts for large clusters, equivalent to the computing power of 20,000 GPUs.
Join us for this fascinating exploration of how the backbone of our AI revolution is being built, the challenges facing the industry, and the innovative solutions emerging at this crucial technological frontier. Discover how these massive infrastructure investments are creating what Tom calls "electroalchemy" - the transformation of energy into intelligence through the medium of silicon.

Support the show

More at https://linktr.ee/EvanKirstel

  continue reading

פרקים

1. Scaling the Digital Backbone: How Data Centers Are Evolving for AI (00:00:00)

2. [Ad] Crossing Borders (00:17:31)

3. (Cont.) Scaling the Digital Backbone: How Data Centers Are Evolving for AI (00:18:07)

518 פרקים

כל הפרקים

×
 
Loading …

ברוכים הבאים אל Player FM!

Player FM סורק את האינטרנט עבור פודקאסטים באיכות גבוהה בשבילכם כדי שתהנו מהם כרגע. זה יישום הפודקאסט הטוב ביותר והוא עובד על אנדרואיד, iPhone ואינטרנט. הירשמו לסנכרון מנויים במכשירים שונים.

 

מדריך עזר מהיר

האזן לתוכנית הזו בזמן שאתה חוקר
הפעלה