Nvidia GTC 2025: CEO Jensen Huang Unveils Major AI Chip Roadmap & Trillion-Dollar Revenue Forecast
Nvidia's annual GPU Technology Conference (GTC) commenced on March 16 in San Jose, California, and runs through March 19. CEO Jensen Huang delivered a keynote address on the first day, during which he projected significant demand for the company's AI chips and announced several new hardware and software products.
Financial Projections
During his keynote, Huang projected that purchase orders for Nvidia's Blackwell and Vera Rubin chip technologies will reach $1 trillion through 2027. This updated projection follows a previous company estimate from last year of a $500 billion revenue opportunity from these technologies.
"Purchase orders for Nvidia's Blackwell and Vera Rubin chip technologies will reach $1 trillion through 2027."
Nvidia Finance Chief Colette Kress indicated last month that current year growth is anticipated to exceed that earlier estimate. Nvidia shares experienced an approximate 2% increase on the day of the announcement. The company has a valuation of approximately $4.5 trillion.
In February, Nvidia projected its year-over-year revenue for the current quarter to increase by about 77%, reaching approximately $78 billion. The company has reported 11 consecutive quarters with revenue growth exceeding 55%.
Hardware Announcements
Vera Rubin System
Nvidia announced plans to launch its next-generation AI computing system, Vera Rubin, later this year. The company states the system, composed of approximately 1.3 million components, will offer up to 10 times more performance per watt compared to its predecessor, Grace Blackwell.
The Rubin computing chip architecture was initially announced in 2024. Nvidia commenced production of Rubin in January. According to the company:
- Rubin is projected to operate 3.5 times faster than the Blackwell architecture for model-training tasks
- It is projected to be 5 times faster for inference tasks, reaching speeds up to 50 petaflops
Nvidia anticipates increasing production of these chips in the second half of the current year.
Nvidia Groq 3 LPU
Huang unveiled the Nvidia Groq 3 Language Processing Unit (LPU) , which is the company's first chip stemming from the acquisition of startup Groq for $20 billion in December. The LPU is slated for shipment in the third quarter.
Software Developments
NemoClaw Platform
Nvidia introduced NemoClaw, a new software stack designed to support the development and deployment of AI agents on the OpenClaw platform. The platform is intended to provide businesses with a structured method to build and deploy AI agents—software capable of autonomously performing multistep tasks.
DLSS 5
Nvidia introduced a new AI graphics rendering technology, DLSS 5. According to CEO Jensen Huang, DLSS 5 enhances image realism while reducing computational workload. The system integrates traditional 3D graphics data with generative AI models, which predict and fill in missing visual details.
Partnerships and Demonstrations
The conference features various partnership announcements and demonstrations showcasing Nvidia's AI capabilities across different industries, including healthcare, robotics, and autonomous vehicles.
Kevin Cook, a senior equity strategist at Zacks Investment Research, indicated that attendees expect to receive updates on Nvidia's relationship with Groq. Nvidia reportedly paid $20 billion late last year to license technology from Groq, an inference company.
"Interest surrounds this collaboration, particularly after Groq's founder Jonathan Ross, president Sunny Madra, and other team members joined Nvidia to advance the licensed technology."