Beirit Harvey

On a recent episode of AI and the Future of Work, host Dan Turchin, PeopleReign CEO, sat down with Darrick Horton, CEO and co-founder of TensorWave, to unpack what it really takes to build next-generation AI infrastructure on AMD GPUs.

AI at Scale: A Conversation with TensorWave’s CEO

On a recent episode of AI and the Future of Work, host Dan Turchin, PeopleReign CEO, sat down with Darrick Horton, CEO and co-founder of TensorWave, to unpack what it really takes to build next-generation AI infrastructure on AMD GPUs.

The conversation ranges from why TensorWave bet on AMD over NVIDIA, to why power, not GPUs, is fast becoming the biggest constraint on AI, to how AI architectures will evolve beyond today’s transformer-dominated landscape.


Why TensorWave Chose AMD and Open Ecos

On a recent episode of AI and the Future of Work, host Dan Turchin, PeopleReign CEO, sat down with Darrick Horton, CEO and co-founder of TensorWave, to unpack what it really takes to build next-generation AI infrastructure on AMD GPUs.

The conversation ranges from why TensorWave bet on AMD over NVIDIA, to why power, not GPUs, is fast becoming the biggest constraint on AI, to how AI architectures will evolve beyond today’s transformer-dominated landscape.


Why TensorWave Chose AMD and Open Ecosystems

Early in the episode, Horton explains why TensorWave chose to build its cloud on AMD Instinct™ GPUs instead of following the NVIDIA-heavy status quo. That choice wasn’t just about hardware, it was about avoiding vendor lock-in and open ecosystems.

That philosophy shows up throughout the discussion: Horton talks about giving customers real alternatives in a capacity-constrained market and why open, standards-based infrastructure matters for the long term, not just the next benchmark.


Training Clusters Vs. Inference Clusters

Another key theme is why TensorWave focused first on training clusters rather than just building out inference capacity. Horton walks through how large, flexible training clusters can be repurposed for different workloads over time, from pre-training and fine-tuning to high-throughput inference.

Horton firmly believes that by choosing an AMD-only GPU strategy gives TensorWave “a clear focus” and is “...the way to win and be successful in this industry is to be the world’s best at the thing that we’re choosing to do, the thing that we're choosing to do and focus on is being the best provider of AMD infrastructure.” 

For teams building complex AI products, that versatility matters more than a single-purpose inference stack. It’s about designing infrastructure that can evolve as models, frameworks, and business needs change.

Power is the New Bottleneck

One of the most important points in the episode: GPUs aren’t the only choke point anymore, power is. As models grow and clusters scale, the limiting factor increasingly becomes megawatts, cooling, and data center design, not just chip supply.

Horton and Turchin dig into:

 * Why data center architecture has to evolve in parallel with GPUs
 * How liquid cooling, higher rack densities, and smarter scheduling help
 * Why energy efficiency and “tokens per watt” will matter as much as raw FLOPS

Horton notes, “The rate of growth that we’re on and the rate of growth of power infrastructure, including these newer and innovative solutions, I think come 2027/2028 you're [going to] run out of power. But a lot of these newer solutions like… small modular nuclear reactors, other forms of green energy, I think we’ll start to see a boom of that more around 2030.” 

That mindset extends to how TensorWave thinks about power, efficiency, and sustainability across the stack.


Beyond Transformers and Toward Greener AI

Looking ahead, Horton predicts that AI architectures will continue to evolve beyond transformers, driving constant shifts in how compute is used and where infrastructure needs to live.

He and Turchin touch on:

 * How new architectures will reshape cluster design and scheduling.
 * Why green energy and better feedback loops between workloads and infra are now strategic, not “nice to have”.
 * How tighter integration between AI workflows and infrastructure will separate the next generation of NeoCloud providers from legacy hyperscalers.


Listen to the Full Conversation

If you care about AI infrastructure, AMD vs. NVIDIA, power constraints, or the future of NeoClouds, this episode is a must-listen.

Listen to the full episode here: https://www.buzzsprout.com/520474/episodes/18248720


About TensorWave

TensorWave is the AI neocloud purpose-built for performance. Powered exclusively by AMD Instinct™ Series GPUs, we deliver high-bandwidth, memory optimized infrastructure that scales with your most demanding models–training or inference. 

Ready to get started? Connect with a sales engineer. 

AMD Instinct™ MI355X GPUs Now Available

AI at Scale: A Conversation with TensorWave’s CEO

Subscribe to our Blog

Stay ahead of the curve with the latest in AI, AMD accelerators, and all things TensorWave.

Product

Resources

Company

© 2025 TensorWave Inc. - All rights reserved.