Today we're launching Live Inventory: a second way to buy GPU capacity. Providers post their clusters directly — model, count, location, term, price per GPU/hour, and go-live date. 31 clusters listed today: 32x H200, US Northwest, $2.81/GPU/hr, Sep 30 48x H100, Asia Pacific, $2.70, Nov 24 256x B200, Europe West, $4.68, Nov 30 72x B300, US, $3.95, Jan '27 1,152x GB300 NVL, US West, $6.50, Feb '27 Every listing comes from a verified provider, with transparent pricing and specs. Live now on the Compute Exchange Marketplace. Try it out: compute.exchange

The institutional marketplace for compute procurement— independent, neutral, and purpose-built for professional market participants.

the more I reread this post, the more I become convinced this is NVIDIA's manifesto of becoming the "central bank of compute." some thoughts on this econ model + nvidia's role: > econ101 tells us a central bank has several goals - (1) anchor inflation, (2) full employment/utilization, (3) lender of last resort. Commercial banks on the other hand "create money" in the economy by lending and thus increasing GDP > the loan in GPU financing world is datacenter buildout (of NVIDIA factories), the "GDP growth" is essentially token/inference spend. The primary "commercial banks" are neoclouds (Coreweave, Nebius et al.) that contract with the labs/inference platforms, which manufacture tokens and allow for "GDP growth". This is the first layer of compute. Lets call it C1. > in addition to directly "lending" to the giants as C1, wholesale compute can also get unbundled into PAYG/short-term duration markets. this is the whole "GPU marketplace layer" - the world of SF compute/Vast AI/compute exchange that sell bare metal. We can call this C2 > all the orderbooks on C1 and C2 can then get bundled into a bunch of synthetic financialization tools on the outside layer, which we can label as C3. Indexes, cash-settled exchanges, residual puts, insurance stuff all belong here. so why is nvidia a central bank? going back to econ101 - (1), the "inflation" in compute land is actually depreciation curves. NVIDIA can control the depreciation curve by deciding when to release a new series of chips (eg. Vera Rubin). (2) NVIDIA's goal is to enable full utilization in the economy of their GPUs (which spurs more demand + higher prices etc.) and so will let the financing stack go faster. (3) NVIDIA has the balance sheet and resources to "bail" failing neoclouds/marketplaces out that don't see the inference demand growth as the lender of last resort. this is where the 25% underwriting comes from.

Diffusion models are outrunning the infrastructure that serves them. So we're partnering with @tensorwave and @DecartAI to dig into the gap. AI Research Paper Club #002 brings together a technical crowd around what's next for diffusion and multimodal models — training costs, inference latency, and the architecture decisions that are still genuinely open.

Token Forwards are live on Compute Exchange — covered today by @dakincampbell in @theinformation . Privately negotiated contracts that lock token prices for up to six months, across open-weight models from six inference providers. Buyers spec the model, price, and throughput they need — providers compete for the bid. Why now: token spot prices that held for weeks can move hourly, and companies like Uber and ServiceNow burned through annual AI budgets in months. Inference is heading toward two-thirds of AI workloads by year end. Costs that volatile deserve a market. Full story: theinformation.com/newslette…

H100 forward contracts are in contango on @Silicon_Data says CEO @carmenli This means the market is pricing in a compute *crunch* and pays more for a GPU in the future than today. Bullish compute 🐂

Used H100s and A100s are among the most requested GPUs in AI. Today we launched the Hardware Market: used and refurbished GPUs, independently verified by SiliconMark. Requests already run from hundreds to tens of thousands of units. Read more: siliconangle.com/2026/07/17/…

Last week in Paris, our CEO @carmenli Li joined Sam V. Tabar (@WhiteFiber_ ), Piotr Tomasik (@tensorwave ), Val Bercovici (@weka ), and Parimal Pandya (@Akamai ) on stage at the Carrousel du Louvre, moderated by Robin Wauters. The question on the table: what happens when compute stops being an IT line item and starts behaving like a financial asset? GPU capacity already trades on term length, price, and counterparty. Procurement processes just haven't caught up. 9,000 people at RAISE Summit this year, and compute economics ran through nearly every conversation. Merci, Paris. 🇫🇷

Eight months ago, when we started building the LLM Token Expenditure Index, the goal was simple: bring transparency to how much, as a society, we’re actually spending on AI per million token. I do not think it was that useful for me to tell you what you’re paying for one model—you already know your own bill. What I found much more interesting was the market as a whole: a volume-weighted view across models that shows our collective willingness to pay for AI. Eight months later, it’s exciting to see the index now sitting at the center of this conversation. This is exactly why we built it. Happy to chat.

Markets were jittery this week on news reports that Meta may be selling compute, raising concerns about excess supply. We thought we’d share some perspective using our rental term curves on why this news, if true, doesn’t have to be bearish for GPU rental prices. Last year, ahead of the agentic AI boom, Meta aggressively locked in a large amount of compute capacity. It quite possibly secured more than it ultimately needed. It was a smart strategic bet to secure scarce supply early, with the flexibility to either deploy it internally or sell excess into a tight market. They acquired valuable real options at the time. As recently as last November, the compute market looked very different. Spot and forward GPU rental rates were much lower, and the term curve was sharply backwardated. This is classic commodity behavior when the market anticipates new supply coming online and pressuring prices lower. Since then, the picture has changed dramatically. As shown in our H100 term rate curves below, the entire curve has both risen sharply in level and flattened significantly, moving out of its steep backwardation. In fact, rental rates have firmed further around the 1-year term over just the past week (Jun 25 – Jul 2), with multiple providers raising prices. For all the concern about a glut, the rental market is doing the opposite of pricing one in: rates are firming, not softening. It now makes perfect financial sense for Meta to shed some of its older secured capacity while continuing to invest in newer, more powerful clusters. The real option they purchased has appreciated meaningfully. At the same time, demand for their specific models and use cases may not have materialized as strongly or as quickly as anticipated. This looks like firm-level rebalancing rather than a signal about the broader market. Reallocating from legacy commitments toward frontier hardware is what a maturing market looks like: optimization, not weakness. Little in our data suggests the demand tailwinds from agentic AI and inference are softening. If anything, the term structure of GPU rental rates points to a market that’s tightening, not loosening. Our forward and term curves are updated daily at silicondata.com. Happy Fourth! 🇺🇸🎆

AI Infra Night is tomorrow 🌆 Chips, power, capital, and the marketplace behind the GPU economy — live in NYC.

AI Infra Night — June 4, NYC 🌆 Chips, power, capital, and the marketplace behind the GPU economy. An evening with cloud operators, investors, and infra leaders. Co-hosted by TowersTeam, @evergrid + us, at Wix NY HQ. Part of #NYTechWeek. partiful.com/e/9Vb2HFPhUit5R…

Attending #NVIDIAGTC in DC? Meet @computeexchange & @Silicon_Data #CEO @carmenli in person at the conference (Oct 28-29). Let’s connect, explore #compute markets, #AIInfrastructure, & #GPU sourcing. Time is limited — Use the QR code to lock in a slot. #AI #CloudComputing

🚀 LAUNCHED TODAY: RFQ Hub 🚀for #compute 1 RFQ → 75+ providers → quotes in minutes. Save 30%–80%. 👉 Submit specs: compute.exchange/

@computeexchange is at #pytorchcon today Thurs. Oct. 23. Schedule your time to meet! Learn how to save 30% to 80% on compute & about our new "Referral" program & "RFQ Hub." #compute #CloudComputing #ai #Infrastructure

@computeexchange is at #PyTorchCon this week! If your team’s growing and you need quality #GPUs (for Less) — let’s talk 💬 📲 Scan the QR in the post to book time with me during the event. #AI #DeepLearning #GPUs #PyTorchCommunity

Live now: our latest GPU auction on @computeexchange. A new benchmark for transparency & access in the compute economy. Proud of the team making it happen! #AI #ComputeMarkets #GPU

New leadership at @computeexchange : @carmenli steps in as CEO, bringing together @Silicon_Data intelligence layer with our compute marketplace. A unified vision for the AI economy. Read more: einpresswire.com/article/855…

Every era has its defining resource. 🌾 1800s: Agriculture 🛢️ 1900s: Oil 💻 Today: Compute Yet compute markets are still broken. We’re changing that—transparent spot trading now, futures next. Resource in 1st comment ⬇️