關於這一集
What does it actually mean to own the GPU layer of AI? In this episode, John Kostros sits down with Sunny Smith, Founder and CTO of Massed Compute, one of the few operators building at the true infrastructure layer: owning hardware, deploying GPUs at scale, and powering what he calls the "underlying carrier layer" for next-gen cloud.
Sunny has lived through every major tech wave, the PC revolution, the internet era, big data, and now AI. He brings a rare perspective: a full-stack executive who still codes, who went from selling an ISP in the 90s to buying 20 servers and 160 GPUs and building a multi-tiered GPU cloud from the ground up.
This episode is a no-fluff, deep dive into the real economics and operational realities of AI infrastructure in 2025.
In this episode, we cover:
Why most "Neo Clouds" don't actually own their GPUs and why that matters for you as a buyer
The "dropshipping infrastructure" problem and how it breaks down on the support side
Why storage is the true Achilles heel of AI (not compute)
The data locality problem: why you're landlocked by where your data lives
How MassCompute's "Local Metal" offering solves the enterprise on-prem AI dilemma
Why agentic AI is now running real-time cluster optimization and replacing what used to take days of human effort
The GPU scarcity market dynamics: preemptible contracts, 40% down payments, and a seller's market unlike anything Sunny has seen in 30 years
What "post-Kubernetes" infrastructure might look like and why optimization is the real moat
Why enterprise GPU demand could 10x once Fortune 1000 companies finally lean in
The capital question: AI has now consumed more inflation-adjusted dollars than building the entire US interstate highway system
Who Should Listen:
CTOs, Engineering Leaders & Cloud Architects. If you're evaluating GPU infrastructure providers, this episode will teach you the right due diligence questions to ask: starting with "Do you actually own the GPU?"
FinOps & Procurement Teams: Understand why NVMe prices have tripled, why storage is no longer a rounding error on your cloud bill, and how to think about OpEx vs CapEx for GPU infrastructure.
AI Builders & Developers: Whether you're fine-tuning models, building RAG pipelines, or deploying inference workloads, Sunny breaks down the infrastructure decisions that are silently costing you performance.
Enterprise Decision-Makers in FinTech, Healthcare & Regulated Industries
Learn why the "move everything to the cloud" era is over for IP-sensitive data, and what the next-gen on-prem + cloud hybrid model looks like.
Founders & Operators in the Neo Cloud Space — Get a rare insider view of how a bootstrapped GPU infrastructure company competed, survived, and evolved alongside the hyperscalers.
About Massed Compute:
Massed Compute is a GPU infrastructure company that owns and operates its hardware, offering bare metal, virtual machines, and Local Metal (on-prem GPU leasing). They serve developers, AI companies, and enterprises needing high-performance compute with white-glove, layer-7 support.
About DataStorage.com:
DataStorage.com is the go-to resource for builders navigating cloud infrastructure, GPU economics, and AI data strategy. We go deep with the operators, founders, and engineers actually shaping the future — no hype, just signal.
🔔 Subscribe for new episodes every month.
