Liquid Compute

[Liquid Compute @liquidcompute]

RT @ronitrjain: Our LCI-H200 Index is up 3.7% this week, following the election.

The huge takeaway is if a dem majority ends up slowing da…

[Nawroz Minsaria @navrose_nm7]

@stanleylee21 🎯good to finally see a nuanced sensible take on all of this

[Liquid Compute @liquidcompute]

$CRWV is up over 20% today. CoreWeave is leaning harder on customer prepayments and on loans backed by investment-grade customers, and says it won’t issue more unsecured debt through 2027.

Most mid-sized neoclouds don’t have investment-grade customers to borrow against. With lenders pulling back, they need customers to prepay more of each cluster, and those prepayments are paid for out of venture rounds, the most expensive money in the deal.

That is the gap we finance. Last month we arranged a bridge facility that let a neocloud pay for its GPUs before delivery, months before its term lenders would fund.

[Stanley Lee @stanleylee21]

CoreWeave $CRWV is leaning harder on customer prepayments and returning to IG offtake for its next DDTL because public credit got flooded with supply and drew the line. That's as much a sign of compute's first credit cycle as it is a "broken" market.

In these cycles, the financing burden shifts to whoever can carry it. At the top of the mkt that's IG balance sheets, who have access to some of the world's cheapest capital. In the middle market it's actually venture rounds, the most expensive, carrying this through prepayments as lenders take a step back/have re-racked levels wider.

This actually makes inference company raises, not neocloud spreads, the signal to watch in the medium term. As long as those keep getting done, customers keep paying and lenders get comfortable funding more of each cluster again. Over time, prepays come back down and cheaper capital finds its way back to the good operators. Same path every financed asset class takes.

[Stanley Lee @stanleylee21]

CoreWeave $CRWV is leaning harder on customer prepayments and returning to IG offtake for its next DDTL because public credit got flooded with supply and drew the line. That's as much a sign of compute's first credit cycle as it is a "broken" market.

In these cycles, the financing burden shifts to whoever can carry it. At the top of the mkt that's IG balance sheets, who have access to some of the world's cheapest capital. In the middle market it's actually venture rounds, the most expensive, carrying this through prepayments as lenders take a step back/have re-racked levels wider.

This actually makes inference company raises, not neocloud spreads, the signal to watch in the medium term. As long as those keep getting done, customers keep paying and lenders get comfortable funding more of each cluster again. Over time, prepays come back down and cheaper capital finds its way back to the good operators. Same path every financed asset class takes.

[Ronit Jain @ronitrjain]

The biggest unlock for the compute ecosystem will be the ability for liquidity providers to price puts.

> Lenders will be able to use it for backstops on their loans

> Neo-clouds will be able to lease more capacity on-demand and on shorter terms by protecting their downside through hedging

> AI companies reap the benefits of shorter-term compute access

We are executing these right now @liquidcompute

[Stanley Lee @stanleylee21]

Yes and this is why the neocloud credit crunch and the prepayment shift are the exact same story. The prepayment didn't kill the credit question for lenders…it changed who's answering it. The customers paying higher upfronts are now financing the buildout, and most of those customers are venture-funded inference companies paying deposits out of their last round.

That's why a 3-year B300 is cheaper than monthly right now, the discount is from the price of the customer's capital, not just the hardware alone.

Seeing a correction in the neo cloud market too (which is healthy)! Operators who can't fund a deposit lose the allocation but operators with real dock-to-live numbers get cheaper capital. The spread between these two is where we are at in the cycle and quite frankly this is normal for any commodity market growing (particularly one financed by credit).

The next step is lenders financing the end compute buyers against these contracts so they don't have to sell equity every three months to lock in cheap compute, and that's finally starting to happen now.

[Stanley Lee @stanleylee21]

Thank you for the shoutout on some of our compute credit research! This series from @alexsongis is the best breakdown I've read of why aircraft finance is the right analogy for GPUs and where it breaks. The engine point in particular is really sharp. There are no regular "shop visits" resetting GPU collateral, so the structure has to do the work that maintenance reserves do in an engine deal. These innovative structures are starting to work though. We're building structures where puts on GPU-hour rental rates get pledged to the lender, which is honestly the closest thing this market has to a maintenance reserve. And the backup servicing layer Alex is right to flag is starting to show up, folks like us @liquidcompute are being named remarketing agent in deals that are closing right now. For anyone originating in the space, this is a must right now. Great series for folks to keep following from Alex.

[Stanley Lee @stanleylee21]

For the first time, options on compute are trading against a live index. If you're a neocloud whose lender is nervous about residual value, an AI company worried about locking in pricing ahead of more supply crunches, or a market maker that wants the other side of a structural hedging flow, reach out to us @liquidcompute

[Liquid Compute @liquidcompute]

When GPU rates fall, your revenue falls. Your power bill and debt payments don’t.

A put on the Liquid Compute Index pays you the difference below your strike, in cash. No GPUs change hands.

Neoclouds: protect your floor. AI companies: cap your costs.

Open a hedge now to protect yourself against future uncertainty https://t.co/5jxvqjqvCg

[Stanley Lee @stanleylee21]

Thank you to @CosmeLaw for citing our research in this. As the ecosystem evolves, the credit market has zeroed in on one of your key takeaways: a foreclosed cluster is worth less pulled from the rack than running in place, so recovery runs through remarketing rather than liquidation/sale of hardware. This is exactly why we think the remarketing agent and a reference price belong in loans from day one and are offering this service. At @liquidcompute we are building the market infrastructure for compute, a rental price index, a venue for buying and selling capacity, cash settled hedges on the index, and the financing structures that sit on top of all of it. If any of that is of interest, would encourage folks to reach out!

[Stanley Lee @stanleylee21]

This is a really thoughtful and well written article. The one thing I would comment on/add is that hedge demand in compute doesn’t just have to come from buyers or operators choosing to trade, it can come from the capital behind them doing the financing. There is an enormous amt of risk sitting in the financing stack right now that isn’t yet moveable but would be with hedges. This is a key difference from bandwidth imo. Fiber in the ground cost nothing to use and no one had a loan against an asset that was shrinking in value whereas GPUs are bought with debt and worth less every month (leaving a financier exposed to residual value). This is somewhat relevant to the consolidation point too, operators can consolidate into the hyperscalers (and I genuinely agree they likely will) but the 100+ credit funds and banks that financed them don’t.

The “everyone on the same side in an unwind” argument has a lot of merit too but is also why this can work. CDS markets grew over time after early blow ups because the possibility of defaults and need to hedge created a structural flow of people who had to buy protection. This same structural flow/transfer of risk premium is exactly what it takes for market makers and liquidity providers to care enough to trade the product in the first place. I think so much of what survives from this compute financial buildout will look like aircraft finance and credit markets…not as sexy as oil futures but still a real market with real risk transfer and a real use case.

This piece was outstanding though, thank you for sharing!

[Liquid Compute @liquidcompute]

GPU prices flat, neocloud stocks down.

Markets have swung hard since the midterms, with data center build-outs now a central issue after the results.

What does that mean for compute prices?

[Liquid Compute @liquidcompute]

Our CPO @stanleylee21 writes about this in his latest piece: https://t.co/w91Z6YcZ4q

[Liquid Compute @liquidcompute]

GPU prices flat, neocloud stocks down.

Markets have swung hard since the midterms, with data center build-outs now a central issue after the results.

What does that mean for compute prices?

[Stanley Lee @stanleylee21]

Some thoughts on what midterm election results mean for compute markets and compute financing... https://t.co/FK7j3k3E5b

[Liquid Compute @liquidcompute]

Lending on Compute

Liquid Compute and K8 Capital are pleased to host Lending on Compute, an evening of conversation and connection in New York City on November 11, focused on GPU financing, compute markets and the credit behind them.

AI companies and infrastructure providers are committing unprecedented amounts of capital to compute, and how that capital gets financed is becoming one of the defining questions in credit. GPUs are emerging as a new asset class, with their own questions around valuation, depreciation, re-marketability and counterparty risk.

The market is also shifting quickly. Rental rates on older GPU generations have risen this year rather than falling as new chips arrive, challenging common assumptions about depreciation. At the same time, tighter capital markets are making long-term financing slower and harder to secure, leaving many neoclouds to fund large hardware deposits and bridge the gap to permanent financing. Lenders are rethinking how they underwrite the asset, and new pools of capital are looking for a way in.

Join @stanleylee21 , Chief Product Officer of Liquid Compute, and Chris Frissora of K8 Capital for a broad discussion of where compute financing stands today and where it is headed. Topics will include what recent pricing trends mean for GPU collateral, how lenders are underwriting GPU-backed loans, how tenor and pricing in the compute market shape credit risk, and the role of offtake contracts and prepayments. Stanley and Chris will also share some of the new structures Liquid Compute and K8 Capital have been developing together, from bridge financing to compute prepayments.

Join us to exchange ideas, meet potential lending partners, and connect with the people building the financial foundations of the compute economy.

Sign up: https://t.co/37MWKJrFIS

[Liquid Compute @liquidcompute]

Proud to be a compute partner to @pika_labs 📷

Huge congrats on launching Pika H3 Flex, debuting today as the #1 open weight model on DesignArena's t2v board.

Big launch from a great team!

[Pika @pika_labs]

Meet Pika H3 Flex, debuting today as the #1 open weight model on DesignArena’s t2v board.

It’s built for professional iteration, with exportable audio stems, resolutions up to 4k, and faster-than-realtime generation. It can produce a 15-second clip with soundtrack in as little as 7 seconds.

So, yeah. It’s kind of a flex 💅

[Liquid Compute @liquidcompute]

Last week we submitted our comments to the @CFTC on compute derivatives. This week our co-founder and CEO @ronitrjain spoke at the CFTC’s first Frontier Forum on how super intelligence will change US and global financial markets.

Ronit’s case: the benchmarks for grain, oil and interest rates trade on regulated US exchanges, and that has given America real influence over global markets. Compute should be next. Proper regulation of compute derivatives is how America keeps its lead in both SI and finance.

Thank you to Chairman Selig and the Innovation Task Force for including us.