10+ hours of podcasts, in 5 minutes.
20 quotes from 2 episodes on Latent Space and TBPN, each with a timestamped link to the source.
Evan Conrad argues that trading AI compute requires a neutral exchange operator that runs physical data centers without owning any hardware. Hardware providers cannot run this market because market makers avoid consuming their own supply to attract independent financing.
Nvidia benefits from a compute exchange existing but cannot build one because the company lacks the neutrality required for traders to trust the platform.
“So our impression is that Nvidia can't make it, but benefits tremendously from it existing. The reason why they can't make it is because they would then not be independent enough for people to trade on it.”
Operators running inference workloads on a single B300 chip can generate $40 an hour serving models.
“Right now if you took a B300 and you tried to run an inference on it, um you can make $40 an hour or so um selling that Any model, basically…”
Venture capitalists currently hold the financial risk of the artificial compute bubble.
“So if you're looking for a bubble, it's held currently by the venture capitalists…”
From Evan Conrad on Why Compute Futures Need Physical Settlement
Evan Conrad explains that hardware providers and cloud hosts cannot operate a compute exchange because their internal demand biases the market. A functional forward market requires an independent operator that gives up lucrative cloud hosting deals.
“We don't own the hardware. The reason why we don't own the hardware is because if we did, we would be able to manipulate the forwards market. So, we have to be Switzerland, we have to be independent.”
“This is the thing that benefits the entire industry because it makes the financing costs go down. But nobody can make it unless they give up the ability to do all the big inference money and all the big cloud deals. You have to kill your business in order to make this.”
“If you don't trade on it, then you don't have enough liquidity for lenders to finance against the forwards market. This is also why Crusoe and Coreweave and Lambda why the hyperscalers can't make it, why Base10 and Fireworks and Modal can't make it cuz they have too much demand on that side.”
Trading compute relies on physical delivery and direct infrastructure control. Evan Conrad notes that market operators end up running the entire facility, building programmable data centers, and testing hardware to eliminate faulty chips before customers receive them.
“That market operator also needs to do what's called physical settlement…”
From Evan Conrad on Why Compute Futures Need Physical Settlement
“And at a certain point, you're just running the whole hotel…”
“When you get a cluster, a good cloud, what it does is it tries to kill all the GPUs before they ever get in the hands of the customer…”
Standardizing compute requires matching multiple variables at once. Evan Conrad points out that while this approach widens the spread between buyers and sellers, buyers who write flexible workloads can make back $25 for every $100 they spend.
“What we did instead was we created an order book that had multiple different constraints…”
“The trade-off of this is that you have less concentrated liquidity, and so the prices are going to have more spreads between the buy and the sell.”
“…a customer who buys capacity from us, and they spend $100 to buy that capacity, will make back $25 on average from selling it back to somebody else.”
Compute clusters carry fixed debt payments over 3 to 5 years. Because unsold capacity expires instantly, Evan Conrad states that operators cut off flexible batch inference jobs with an hour of notice to maximize utilization.
“Clusters are financed 3 to 5 years. Um so the the payments are over that period of time.”
“Basically, like, if I have a workload that is, um, uh, that's batch inference, basically, where I can be interrupted at any point, I don't really care. Um, then, my ability to be cut off, um, is probably like, I need an hour of time to know, um, that I'm going to be cut off and then I'm good.”
“To get closer to something that can be standardized, you first need to make all the operations of the thing the same…”
“If you have a product that can use lots of different types of capacity, then you might really like SF Compute because you will be able to use lots of different types of capacity and you'll make more money…”
“To make that work you also need like a fast transit path between one cluster and another…”
“If you have a live job, you're you are probably paying an SLA when you break for your customer…”
“We're virtualizing data center and making it programmable…”
From Evan Conrad on Why SF Compute Virtualizes InfiniBand Fabrics
“Right now, compute kind of looks like the early days before there was a power grid…”
From Evan Conrad on Why Compute Futures Need Physical Settlement
How we attribute quotes. Every quote was matched against the episode transcript, so the words and the timestamp are real (we trim filler words like "um", nothing else). The name comes from our written summary of the episode, and we use it only when a separate check of the captions finds that person on the episode. YouTube gives us no voice-by-voice transcript, so open the timestamp to hear who is talking. See a wrong name? Tell us and we fix or remove it.
10+ hours of podcasts, distilled into one 5-minute read. Free, every Sunday.
For now, every subscriber gets both newsletters. No spam. Unsubscribe with one click.