Last week, Conapto’s Stefan Nilsson, Jesper Kvist, and Jillian Ochs attended Nvidia GTC 2025, gaining valuable insights into the latest AI advancements. One thing was clear – AI is growing fast, pushing performance and power demand to new levels. The event featured Nvidia’s CEO Jensen Huang’s well-visited keynote, showcasing the next generation of AI, including Blackwell Ultra, new Vera Rubin platform as well as Rubin Ultra, new DGX systems, SpectrumX Ethernet, and AI-driven robotics.

No Nvidia GTC would be complete without groundbreaking GPU announcements, and GTC 2025 was no exception. Nvidia introduced the next generation of Blackwell Ultra GPUs, setting new records in computing performance and efficiency. Designed for AI-heavy workloads, deep learning models, and high-performance computing, these GPUs promise faster, more efficient processing than ever before.

As Blackwell GB200 is now shipping, one of the key highlights was the launch of the Blackwell Ultra GB300, designed to deliver a significant performance boost while lowering operational costs. This powerhouse will be pivotal for AI research, gaming, scientific computing, and autonomous vehicle development. As AI applications grow in complexity, the GB300 will be a crucial enabler of next-generation innovation.

Beyond hardware, GTC 2025 painted a clear picture of how AI, GPU acceleration, and cloud computing will drive digital transformation across industries. Nvidia’s continued investment in AI infrastructure and energy-efficient computing signals a future where real-time AI applications become the norm, from autonomous driving to medical research.

For Conapto, the key takeaway is that sustainability must evolve alongside performance. As AI models scale, the infrastructure supporting them needs to keep pace – not just in power and cooling, but in energy efficiency and responsible resource use.

Scaling AI Infrastructure Sustainably

Yesterday: 10kW per rack
Today: 120kW per rack
Tomorrow: 600kW per rack

With such rapid growth, the big question is: How do we scale AI infrastructure while staying sustainable? At Conapto, we believe the answer lies in smart infrastructure choices:

• Direct-to-chip liquid cooling to manage extreme heat loads efficiently.
• Heat recovery for district heating, turning excess energy into a valuable resource.
• Fossil-free and renewable power sources to minimize environmental impact.
• Modular design for rapid, cost-effective expansion.

As AI infrastructure evolves, so does its real-world impact. One exciting development at GTC 2025 was the exhibition floor, where attendees had the chance to see a variety of robots and self-driving cars. During his keynote, Nvidia’s CEO Jensen Huang emphasized the growing role of robotics:

Physical AI and robotics are moving so fast, everybody, pay attention to this space. This could very well likely be the largest industry of all.

While robots are still in their early stages, AI-driven automation is advancing quickly. Nvidia envisions a future where AI-powered robots revolutionize industries by performing labor-intensive tasks, powering autonomous vehicles, and interacting with the physical world in ways never seen before.

Nvidia GTC 2025 confirmed what we already knew: AI is the driving force of the future, and infrastructure must keep up. For Conapto, this means continuing to innovate in high density cooling, sustainable power solutions, and scalable data center design, ensuring that our customers can harness the power of AI without compromising on efficiency or sustainability.

We’re excited for what’s ahead. The AI revolution is here, and we’re ready to build the foundation for its future.