What are the key takeaways from this Gradient Dissent episode?

Key insights include: **Purpose-built storage architecture:** CoreWeave's LOTA cache and object storage system optimizes GPU utilization by maximizing data throughput directly to GPUs, making different design assumptions than public clouds that must serve diverse workloads like ecommerce sites with different read-write patterns and consistency requirements.; **Liquid cooling infrastructure advantage:** Building data centers exclusively for AI workloads enables CoreWeave to deploy liquid cooling at scale across all facilities, while public clouds struggle with fungibility requirements. Some latest-generation GPUs physically require liquid cooling and cannot run without it, creating supply constraints elsewhere.; **Network latency becomes less critical:** AI inference workloads spend most processing time inside the GPU rather than on network calls, enabling flexible multi-region deployment strategies. This allows dramatic improvements in availability and burst capacity management compared to traditional applications where network positioning matters significantly.

What did Corey Sanders and Lucas Bewald discuss on Gradient Dissent?

CoreWeave SVP Corey Sanders explains how the $41B AI cloud differentiates from AWS, Azure, and GCP through specialized infrastructure like liquid cooling, custom object storage, and laser focus on AI workloads rather than general-purpose computing. Key topics include: **Purpose-built storage architecture:** CoreWeave's LOTA cache and object storage system optimizes GPU utilization by maximizing data throughput directly to GPUs, making different design assumptions than public clouds that must serve diverse workloads like ecommerce sites with different read-write patterns and consistency requirements.; **Liquid cooling infrastructure advantage:** Building data centers exclusively for AI workloads enables CoreWeave to deploy liquid cooling at scale across all facilities, while public clouds struggle with fungibility requirements. Some latest-generation GPUs physically require liquid cooling and cannot run without it, creating supply constraints elsewhere..

How long is this episode of Gradient Dissent?

This episode is 53 minutes long. SignalCast provides an AI-generated summary so you can get the key insights in about 3 minutes.

Gradient Dissent

Inside the $41B AI Cloud Challenging Big Tech | CoreWeave SVP

January 6, 2026

53 min episode · 2 min read

Corey Sanders,Lucas Bewald

Episode

53 min

Read time

2 min

Topics

Productivity, Remote Work, Leadership

AI-Generated Summary

Published Jan 6, 2026

Key Takeaways

✓Purpose-built storage architecture: CoreWeave's LOTA cache and object storage system optimizes GPU utilization by maximizing data throughput directly to GPUs, making different design assumptions than public clouds that must serve diverse workloads like ecommerce sites with different read-write patterns and consistency requirements.
✓Liquid cooling infrastructure advantage: Building data centers exclusively for AI workloads enables CoreWeave to deploy liquid cooling at scale across all facilities, while public clouds struggle with fungibility requirements. Some latest-generation GPUs physically require liquid cooling and cannot run without it, creating supply constraints elsewhere.
✓Network latency becomes less critical: AI inference workloads spend most processing time inside the GPU rather than on network calls, enabling flexible multi-region deployment strategies. This allows dramatic improvements in availability and burst capacity management compared to traditional applications where network positioning matters significantly.
✓Customer engagement at scale: CoreWeave's CTO actively participates in customer Slack channels with double the message volume of other employees, providing hands-on technical support to a much larger percentage of the customer base than hyperscale clouds can offer their non-top-tier accounts.

What It Covers

CoreWeave SVP Corey Sanders explains how the $41B AI cloud differentiates from AWS, Azure, and GCP through specialized infrastructure like liquid cooling, custom object storage, and laser focus on AI workloads rather than general-purpose computing.

Key Questions Answered

•Purpose-built storage architecture: CoreWeave's LOTA cache and object storage system optimizes GPU utilization by maximizing data throughput directly to GPUs, making different design assumptions than public clouds that must serve diverse workloads like ecommerce sites with different read-write patterns and consistency requirements.
•Liquid cooling infrastructure advantage: Building data centers exclusively for AI workloads enables CoreWeave to deploy liquid cooling at scale across all facilities, while public clouds struggle with fungibility requirements. Some latest-generation GPUs physically require liquid cooling and cannot run without it, creating supply constraints elsewhere.
•Network latency becomes less critical: AI inference workloads spend most processing time inside the GPU rather than on network calls, enabling flexible multi-region deployment strategies. This allows dramatic improvements in availability and burst capacity management compared to traditional applications where network positioning matters significantly.
•Customer engagement at scale: CoreWeave's CTO actively participates in customer Slack channels with double the message volume of other employees, providing hands-on technical support to a much larger percentage of the customer base than hyperscale clouds can offer their non-top-tier accounts.

Notable Moment

Sanders reveals that Microsoft and Google are both CoreWeave customers, using the specialized AI infrastructure for specific workloads because the purpose-built architecture delivers capabilities that general-purpose clouds cannot easily replicate without abandoning their fungibility requirements across diverse use cases.

Know someone who'd find this useful?

Episode Transcript

I don't care if the APIs are consistent and commoditized. The level of quality, performance, and capability and experience that we deliver today will not win workloads in two years. And for anyone who's deployed on a public cloud, especially with GPUs, and you suddenly have a personal capacity, like, you're always sweating. And so what I like to think about for us is how do we go make all of that hudsty then go away? K. I think CoreWeave's already got that. Coreweave is the best place to run training AI workloads, and it's why people use us. It's focused on basically bringing as much possible data and throughput into the DPU as you can. And that's a very unique requirement that AI workloads shouldn't have because GPU is the most expensive asset across all of those symphonetry. And so that allows us to make these assumptions that are simplifying for us and and, daunting for the public cloud such that even the public clouds will come in and talk to us about enabling us for some of their workloads or passengers. You're listening to Gradient Dissent, a show about making machine learning work in the real world, and I'm your host, Lucas Bewald. Alright. Today, I'm talking with Corey Sanders, who is currently the SVP of product at CoreWeave, and prior to that, a long time an executive at Azure where he worked on compute projects and a whole bunch of other things. It might seem like maybe this is CoreWeave sponsored somehow or something, but, this is actually a conversation that I would be excited to have regardless of where I was working. But you should be aware that, Corey is one of my colleagues now that Weights and Biases is bought by by CoreWeave, but we try to stay as objective as possible and keep it interesting. I hope you enjoy it. Alright. Corey, do you wanna sing a sim? Should we, should we do this? The the, the can I be on your podcast? This is a little, a little beat that I that I created, begging Lucas to allow me to join this session. So Yeah. And I guess we you know, we'll we'll put a lot of disclaimers in here, but we're now coworkers. Alright. So this certainly is not an unbiased, interview. And I guess I haven't really all about biases, Lucas. We're all about biases here. And I haven't mentioned this in the past, but, you know, since, Weights and Biases about by CoreWeave, we actually run all of these podcasts through CoreWeave compliance. And I'm a little disappointed that we've never triggered anything, including, you know, the episode with Martin Shkreli, who's, like, a little controversial. And, you know, we try I try to make this interesting, but, we've not ever had compliance complain about anything. So maybe this could this could be our first time. Let's go for it. Let's go for the win here, Lucas. There's gonna be, …

Get the full transcript (10,769 words) + summary by email — free

One-time email with the complete transcript and AI summary of this episode. No account needed.

One email, no spam. We’ll also show you what SignalCast does.

Browse all Gradient Dissent transcripts →