

AI Infra @ Berkeley Speaker Series #1 From Grid to Hyperscaler, Powering the AI Revolution
AI Infra @ UC Berkeley: Panel Discussion with Alumni and Speakers from PG&E, Microsoft and NVIDIA
AI is now limited by power, not chips. Data centers are asking for gigawatts, utilities are trying to keep up with the new load, and cloud providers are rethinking where and how they build.
What we’ll cover:
What it really takes to power AI at scale
Grid constraints, interconnection queues and how long large loads take to come online
How data center design is changing for AI workloads
Where the opportunities are for engineers, founders and investors
Agenda of the day
6:45pm - 7:00pm, Club President open stage, introduction of AI Infra @ Berkeley: club leadership, background, member breakdown, upcoming events, etc.
7:00pm - 7:10pm, Shruti Sethi from Microsoft Azure, Berkeley Haas MBA '23 (LinkedIn)
7:15pm - 7:25pm, Mike Tu from NVIDIA, Berkeley Haas EMBA '28 (LinkedIn)
7:30pm - 7:40pm, Joby Bernstein from PG&E, Stanford GSB MBA/MS (LinkedIn)
7:45pm - 8:00pm, moderated speaker discussion.
8:00pm - 8:15pm, Open Q&A
8:15pm - Stage close, announcement of Winter Mixer and future panels.
Speakers:
Shruti: At Microsoft Azure, she focuses on cloud storage efficiency, scalability, and cost optimization, including HDD/SSD workload placement, fleet optimization, software efficiency, and capacity planning.
Joby: At PG&E, his work and interests include grid innovation, energy strategy, and emerging technologies supporting grid modernization, increasing data center power demand, and sustainability.
Mike: At NVIDIA, Mike is a Principal Electrical Engineer with 20 years of experience in data center infrastructure and power systems. His expertise includes next-generation power architecture and NVIDIA's 800VDC system for efficient, scalable AI infrastructure.
Topics:
Grid: How utilities are using AI and new technologies to balance growing power demand from data centers.
Hyperscalers: How hyperscalers are managing increasing AI workloads, optimizing infrastructure, and improving energy efficiency and sustainability.
AI Power Architecture: How NVIDIA's 800VDC architecture enables more efficient, scalable, and sustainable power delivery for AI data centers.
🗓 Thursday, November 19
🕕 6:45pm – 8:15pm PT
https://luma.com/event/manage/evt-uVtMbVlXLuA1AVD/overview