Uploaded May 2026 | Updated November 2026, 5 hours ago
AI data centers are consuming massive amounts of energy, and that demand is only growing.
In this student research project, electrical engineering student John “Ike” Ogbu explores how wide bandgap semiconductors like silicon carbide (SiC) and gallium nitride (GaN) could improve GPU power efficiency inside modern AI infrastructure. His work focuses on one of the most critical challenges in data center design: converting power efficiently from servers down to the low voltages required by GPUs.
As large language models scale, GPUs require more power, increasing energy loss during power conversion. By redesigning the voltage regulator module (VRM) using next-generation semiconductor materials, this project shows how even small efficiency gains could have a major impact at the scale of hyperscale data centers.
Chapters:
00:00 Meet Ike: Electrical engineering student project
00:30 Why AI data centers use so much energy
01:05 The challenge of powering GPUs
01:40 What are wide bandgap semiconductors?
02:15 Silicon vs. SiC and GaN
02:50 Improving GPU voltage regulation
03:20 Lessons from the project
04:00 What’s next
_________________
SEE MORE UNDERGRADUATE STUDENT VIDEOS:
youtube.com/playlist?list=PLtHrXQsUhv80vhJpQUkJSbNM7ZyIyC6jU&si=QxsYywibJ4US_Ljv
_________________
SUBSCRIBE TO OUR MONTHLY RESEARCH NEWSLETTER
harvard.us1.list-manage.com/subscribe?u=598d73b470a40f403bba4fb2c&id=318af7d2ea
_________________
VISIT US ON SOCIAL MEDIA
facebook.com/hseas
instagram.com/harvardseas
twitter.com/hseas
linkedin.com/school/1711270/admin/dashboard
_________________
Hashtags:
#AIInfrastructure #DataCenters #PowerElectronics #Semiconductors #GalliumNitride #SiliconCarbide #GPUs #EnergyEfficiency #ElectricalEngineering #SustainableAI #EngineeringResearch #TechInnovation
AI data centers are consuming massive amounts of energy, and that demand is only growing.
In this student research project, electrical engineering student John “Ike” Ogbu explores how wide bandgap semiconductors like silicon carbide (SiC) and gallium nitride (GaN) could improve GPU power efficiency inside modern AI infrastructure. His work focuses on one of the most critical challenges in data center design: converting power efficiently from servers down to the low voltages required by GPUs.
As large language models scale, GPUs require more power, increasing energy loss during power conversion. By redesigning the voltage regulator module (VRM) using next-generation semiconductor materials, this project shows how even small efficiency gains could have a major impact at the scale of hyperscale data centers.
Chapters:
00:00 Meet Ike: Electrical engineering student project
00:30 Why AI data centers use so much energy
01:05 The challenge of powering GPUs
01:40 What are wide bandgap semiconductors?
02:15 Silicon vs. SiC and GaN
02:50 Improving GPU voltage regulation
03:20 Lessons from the project
04:00 What’s next
_________________
SEE MORE UNDERGRADUATE STUDENT VIDEOS:
youtube.com/playlist?list=PLtHrXQsUhv80vhJpQUkJSbNM7ZyIyC6jU&si=QxsYywibJ4US_Ljv
_________________
SUBSCRIBE TO OUR MONTHLY RESEARCH NEWSLETTER
harvard.us1.list-manage.com/subscribe?u=598d73b470a40f403bba4fb2c&id=318af7d2ea
_________________
VISIT US ON SOCIAL MEDIA
facebook.com/hseas
instagram.com/harvardseas
twitter.com/hseas
linkedin.com/school/1711270/admin/dashboard
_________________
Hashtags:
#AIInfrastructure #DataCenters #PowerElectronics #Semiconductors #GalliumNitride #SiliconCarbide #GPUs #EnergyEfficiency #ElectricalEngineering #SustainableAI #EngineeringResearch #TechInnovation










