Jordan (@gpufather) header
Jordan (@gpufather) profile picture

co-founder @RacksHQ

Joined October 2026
Photo by Jordan (@gpufather)Photo by Jordan (@gpufather)Photo by Jordan (@gpufather)Photo by Jordan (@gpufather)Photo by Jordan (@gpufather)Photo by Jordan (@gpufather)Photo by Jordan (@gpufather)Photo by Jordan (@gpufather)Photo by Jordan (@gpufather)Photo by Jordan (@gpufather)
Jordan retweeted
Huge congrats to the @Architect_Fi team. We’re glad to witness this massive milestone for compute markets and our industry. Only the tip of the iceberg!
Announcement: Architect, @Luxor, and @hyperbolic_labs have executed the first OTC option on compute. Through our CFTC Introducing Broker we structured a bespoke $5M put option on B200 GPUs settling against @silicondata’s index. Luxor priced and was counterparty to Hyperbolic.
2
5
142
Jordan retweeted
In the past few weeks I’ve learned the best part about building a useful product isn’t even the product itself It’s the workflows you cultivate along the way in order to develop and maintain the product The machine that builds the machine
1
1
5
128
Jordan retweeted
Tired of waiting behind red tape? Racks fixes this.
CME's compute futures will not launch Monday. NYMEX consented on November 4, at the CFTC's request, to another 60 days of review — the new deadline is January 8, 2027. The contracts, cash-settled futures on Silicon Data's H100 and B200 rental indexes at 730 GPU-hours each, were first set for October 5. The first regulated market for hedging GPU rent now slips into 2027 before trading a single contract. $CME
3
6
90
Jordan retweeted
The Great Rack Spread
The "Rack Spread" is unfolding, token prices are falling while GPU prices climb Change since mid-July: - B300 GPU rentals up 106% - B200 up 66% - H100 up 17% - Average token prices down 58% Charts of the Week: a16z.news/p/charts-of-the-we…
1
3
74
Jordan retweeted
robinhood chain
42
32
2
281
8,786
Jordan retweeted
Everyone wants one thing
1
4
3
9
2,994
Delaying compute futures has a real cost for US innovation. Some of the largest data center lenders have hit their risk limits. Give them a liquid market to hedge compute price risk and they can free up lending capacity. It's the lack of a hedging instrument that has caused credit spreads to blow out. We support the @CFTC’s commitment to getting this right. The good news is we already have standards for building these markets properly: independent benchmarks, transparent methodologies and regulatory access to the underlying data. Apply them to compute and let well-designed futures move forward. Our comment letter: regulations.gov/comment/CFTC…
16
40
6
267
42,847
Jordan retweeted
Given the recent volatility, the need for price discovery and risk-management tools for compute has never been clearer. Thank you @citicsintl for the podcast opportunity—and for the chance to connect with clients across Asia around @CMEGroup compute futures. The compute market is growing up. 📈youtube.com/watch?v=39inZajI…
3
16
1,691
Jordan retweeted
The need for a product (like @RacksHQ) is becoming more obvious with the nascent state of compute markets. @BrettHarrison and @andrawesbahou giving free alfa here in their recent @Blockworks discussion
Who cares what a GPU rents for? the renter, when their cloud bill goes up. the provider, when their revenue goes down. and traders with skin in the game.
2
11
524
Jordan retweeted
Who cares what a GPU rents for? the renter, when their cloud bill goes up. the provider, when their revenue goes down. and traders with skin in the game.
1
3
1
11
850
Massive 40% repricing of NVIDIA B300 hardware. When @Kimi_Moonshot realeased KimiK3 in mid-july, they *single handedly* caused the most violent repricing of NVIDIA AI Infrastructure, (40% in 3 months) with no stop in sight. They caused a major depletion of Supermicro's short lead-time hardware in end September, which exacerbated the pricing. Lead times exploded from 4 weeks to 32 weeks. Kimik3 is uniquely suited to run on B300 hardware, which is why our Bloomberg: $DESKB300 index exploded over the summer.
Nvidia hardware is an appreciating asset, with B300 value up 40% in 3 months. "The B300 Summer". In July, @Kimi_Moonshot released KimiK3, expanding B300 demand and value. By late September @Supermicro "Gold SKU" supply on 4-week lead time was depleted, triggering a continued value increase, with no sign of slowing down. Residual value for Nvidia hardware is increasing. @NVIDIAAIInfra @JensenHuang
4
6
1
71
158,987
Compute wars in a difficult spot. A lot of irrational behavior that amplifies the worst SV tendencies “If you overbuy, you could bankrupt the company by having too many lease commitments, and if you underbuy, you could knock down your growth trajectory and have a bunch of unhappy users,”
5
16
4
117
30,002
Jordan retweeted
“you burned how many tokens today?”
115
441
51
11,418
353,518
Jordan retweeted
Every commodity you can't store ends up with a futures market. That's why we built Racks. Racks Season Zero is now live ↓
I just published an essay at The Sequence about the case for AI compute as a commodity: thesequence.substack.com/p/t… Start with a strange property of GPUs: an unused hour of capacity disappears forever. You can warehouse wheat, store oil, and resell a chip. You cannot recover yesterday’s idle GPU-hour. That makes compute surprisingly similar to electricity and freight: expensive infrastructure selling a service whose value depends on timing, location, and availability. Commodity markets developed ways to manage precisely these problems. Standardized grades made grain tradable. Futures helped producers and buyers manage prices. Credit financed the infrastructure behind supply. AI is beginning to build its own version of that machinery. But the details matter. Eight GPUs scattered across unrelated machines cannot necessarily replace one tightly connected cluster. A price hedge can protect your budget without securing a single GPU. The essay explores how compute benchmarks, capacity contracts, derivatives, and financing could fit together, and why hardware turnover and fragmented markets make this difficult. We spend enormous effort improving the machines that produce AI. The markets that allocate their capacity deserve just as much attention. How we price and finance compute will ultimately help determine who can build with it and at what scale.
8
5
1
23
1,018