# GPUQuant > GPUQuant brings GPU rental prices, compute markets and AI infrastructure into one place. GPUQuant is an independent market-intelligence and trading interface at https://gpuquant.com. It publishes cloud rental list-price benchmarks for H100, B200, H200, A100, a separate specialist GPU cloud index, published Silicon Data and Ornn indices, and a sourced map and directory of AI infrastructure. Tradable Compute connects to the named venues; orders are signed in the user's browser and executed by the named venue. GPUQuant does not hold customer funds. Every documentation page is also served as markdown at the same address with `.md` appended. The full documentation in one file is at /llms-full.txt. GPUQuant provides a public read-only MCP endpoint at https://gpuquant.com/api/mcp. Its tools are search_infrastructure, get_facility and get_company. Use them for infrastructure records, canonical URLs, source links and record verification context. Unknown values and older source dates remain explicit; a new request is not a fresh verification. The MCP excludes GPU price feeds, licensed market series, raw source excerpts, imagery, geometry, trading accounts and order execution. ## Start here - [GPUQuant Documentation](https://gpuquant.com/docs.md): What GPUQuant publishes, where each number comes from, and how the benchmarks, the settlement indices and the tradable instruments relate to each other. ## Benchmark - [Reference Price Methodology](https://gpuquant.com/docs/reference-price.md): How the GPUQuant Reference Price for each published GPU is calculated from the list prices of AWS, Azure, Google Cloud and Oracle: what is measured, the five steps, and why each provider counts once. - [Neocloud GPU Price Index](https://gpuquant.com/docs/neocloud-index.md): How the neocloud index is calculated from the published H100 and B200 rates of the specialist GPU clouds: which rate counts as a quote, how each is normalized, and why it is never averaged with the hyperscaler figure. - [Per-GPU Normalization and Canonical Variants](https://gpuquant.com/docs/normalization.md): How a published machine price becomes USD per GPU-hour: the eight-GPU node, the arithmetic for each provider, the Google Cloud rebuild, the canonical variant of each GPU, and the 730-hour GPU-month. - [Time, Holds and Coverage Changes](https://gpuquant.com/docs/history.md): Why the benchmarks are monthly, how a held price is shown, what happens when a provider is missing, and when a period change is quoted and when it is refused. ## Data - [Data Sources](https://gpuquant.com/docs/data-sources.md): The first-party price feeds GPUQuant reads: the AWS EC2 Price List archive, the Azure Retail Prices API, the Google Cloud Billing Catalog and the Oracle price list, with what each one can and cannot tell you. - [Coverage and Exclusions](https://gpuquant.com/docs/coverage.md): What is stored right now: the coverage matrix by GPU and provider, every unresolved mapping and excluded SKU with its reason, the ingestion schedule and the last successful run. ## Markets - [NVIDIA GPU Compute Futures Explained](https://gpuquant.com/docs/compute-futures.md): What the CME Group and Silicon Data H100 and B200 rental-index futures are, what one contract represents, what they settle against, and how a cloud list-price benchmark differs from a futures price. - [Silicon Data Indices and the Forward Curve](https://gpuquant.com/docs/silicon-data.md): The Silicon Data GPU rental indices the compute futures settle against, the forward curve, the RAM and LLM token indices, how each relates to the GPUQuant benchmarks, and what the site is licensed to show. - [Three Prices for One GPU-Hour](https://gpuquant.com/docs/price-indices.md): How the GPUQuant reference price, Silicon Data’s indices and Ornn’s Compute Price Index each arrive at a dollar figure for an hour of the same GPU, why they differ, and what the site shows of each. - [Tradable Compute](https://gpuquant.com/docs/tradable-compute.md): The instruments on the Tradable Compute page, what each is a claim on, how an order is signed and executed without GPUQuant holding funds, what it costs, and the risks of a leveraged perpetual. - [USD.AI Earn](https://gpuquant.com/docs/usdai-earn.md): Add wallet USDC to sUSDai, follow your variable yield, withdraw to USDC, and move it back to trading from Portfolio. ## Site - [What the Map Draws](https://gpuquant.com/docs/map.md): How the map draws a cloud region, what the colour means, and what the facility and relationship layers show and do not show. - [Embed the Map](https://gpuquant.com/docs/embed-map.md): Add the maintained GPUQuant map to your site with JavaScript or an iframe, including exact colors, fixed or editable filters, independent popup controls and all configuration defaults. - [The Figures on the Landing Page](https://gpuquant.com/docs/landing-figures.md): Where each figure on the landing page comes from, and how the capital expenditure clock is computed from reported totals. - [Frequently Asked Questions](https://gpuquant.com/docs/faq.md): Short answers to the questions readers ask most about the prices, the charts, the contracts and the site. ## Reference - [Instrument Reference](https://gpuquant.com/docs/instruments.md): Every instrument on Tradable Compute, one page each: what it is, its place in the compute economy, the terms of the contract, what to read beside it, how it trades from this site, and its risks. - [Silicon Data H100 Rental Index Futures Explained](https://gpuquant.com/docs/instruments/h100-futures.md): The Silicon Data H100 Rental Index futures CME Group plans to list on NYMEX: what one contract represents, what it settles against, when it lists, and how it will trade through a connected broker. - [Silicon Data B200 Rental Index Futures Explained](https://gpuquant.com/docs/instruments/b200-futures.md): The Silicon Data B200 Rental Index futures CME Group plans to list on NYMEX: what one contract represents, what it settles against, when it lists, and how it will trade through a connected broker. - [CHIP Perpetual: USD.AI on Tradable Compute](https://gpuquant.com/docs/instruments/chip.md): The CHIP perpetual on Hyperliquid: the governance token of USD.AI, a protocol that lends dollars against GPUs, and what a leveraged view on GPU-backed credit is and is not. - [RENDER Perpetual: Render Network on Tradable Compute](https://gpuquant.com/docs/instruments/render.md): The RENDER perpetual on Hyperliquid: the token of the Render Network, a marketplace for idle GPU capacity, and what it is exposed to in the compute economy. - [IO Perpetual: io.net on Tradable Compute](https://gpuquant.com/docs/instruments/io.md): The IO perpetual on Hyperliquid: the token of io.net, which pools GPUs from independent operators into clusters for machine learning, and what it is exposed to. - [TAO Perpetual: Bittensor on Tradable Compute](https://gpuquant.com/docs/instruments/tao.md): The TAO perpetual on Hyperliquid: the token of Bittensor, a network that pays for machine-learning work in its own token, and what it is exposed to in the compute economy. - [CRWV Perpetual: CoreWeave on Tradable Compute](https://gpuquant.com/docs/instruments/crwv.md): The CRWV perpetual on Hyperliquid: one share of CoreWeave, the largest independent GPU cloud, and how its price relates to the neocloud index GPUQuant computes. - [NBIS Perpetual: Nebius on Tradable Compute](https://gpuquant.com/docs/instruments/nbis.md): The NBIS perpetual on Hyperliquid: one share of Nebius Group, a neocloud building GPU clusters in Europe and the United States, and how it relates to the neocloud index. - [IREN Perpetual: IREN on Tradable Compute](https://gpuquant.com/docs/instruments/iren.md): The IREN perpetual on Hyperliquid: one share of IREN, a bitcoin miner turning its power into GPU clusters, and what it is exposed to in the compute economy. - [NCLD Perpetual: Roundhill Investments on Tradable Compute](https://gpuquant.com/docs/instruments/ncld.md): The NCLD perpetual on Hyperliquid: one share of the Roundhill Neocloud ETF, a basket of GPU-as-a-service and AI datacenter companies, and what the basket carries. - [ORCL Perpetual: Oracle on Tradable Compute](https://gpuquant.com/docs/instruments/orcl.md): The ORCL perpetual on Hyperliquid: one share of Oracle, the fourth hyperscaler in the GPUQuant reference price, and how its cloud shows up in the data on this site. - [MSFT Perpetual: Microsoft on Tradable Compute](https://gpuquant.com/docs/instruments/msft.md): The MSFT perpetual on Hyperliquid: one share of Microsoft, whose Azure GPU prices are one leg of the reference price, and how the cloud shows up in the data. - [GOOGL Perpetual: Alphabet on Tradable Compute](https://gpuquant.com/docs/instruments/googl.md): The GOOGL perpetual on Hyperliquid: one Class A share of Alphabet, whose Google Cloud GPU prices are one leg of the reference price, and how the cloud shows up in the data. - [AMZN Perpetual: Amazon on Tradable Compute](https://gpuquant.com/docs/instruments/amzn.md): The AMZN perpetual on Hyperliquid: one share of Amazon, whose AWS GPU prices are one leg of the reference price, and how the cloud shows up in the data. - [DRAM Perpetual: Roundhill Investments on Tradable Compute](https://gpuquant.com/docs/instruments/dram.md): The DRAM perpetual on Hyperliquid: one share of the Roundhill Memory ETF, a basket of memory makers, and why the memory cycle sits under every accelerator price. - [MU Perpetual: Micron on Tradable Compute](https://gpuquant.com/docs/instruments/mu.md): The MU perpetual on Hyperliquid: one share of Micron, one of the three makers of the high-bandwidth memory stacked on every accelerator, and what it is exposed to. - [SKHY Perpetual: SK hynix on Tradable Compute](https://gpuquant.com/docs/instruments/skhy.md): The SKHY perpetual on Hyperliquid: one SK hynix ADS, a tenth of a Korean share of the largest HBM supplier to NVIDIA, and what it is exposed to. - [SNDK Perpetual: Sandisk on Tradable Compute](https://gpuquant.com/docs/instruments/sndk.md): The SNDK perpetual on Hyperliquid: one share of Sandisk, the NAND flash maker spun out of Western Digital, and where storage sits in the compute economy. - [NVDA Perpetual: NVIDIA on Tradable Compute](https://gpuquant.com/docs/instruments/nvda.md): The NVDA perpetual on Hyperliquid: one share of NVIDIA, which designs the H100, H200 and B200 this site prices, and how its share relates to the rental price of its GPUs. - [AMD Perpetual: AMD on Tradable Compute](https://gpuquant.com/docs/instruments/amd.md): The AMD perpetual on Hyperliquid: one share of AMD, whose Instinct accelerators are the main alternative to NVIDIA’s, and what it is exposed to in the compute economy. - [AVGO Perpetual: Broadcom on Tradable Compute](https://gpuquant.com/docs/instruments/avgo.md): The AVGO perpetual on Hyperliquid: one share of Broadcom, which designs the hyperscalers’ custom accelerators and the networking between them, and what it is exposed to. - [TSM Perpetual: TSMC on Tradable Compute](https://gpuquant.com/docs/instruments/tsm.md): The TSM perpetual on Hyperliquid: one TSMC ADS, five Taiwan shares of the foundry that fabricates and packages every NVIDIA and AMD accelerator, and what it is exposed to. - [ASML Perpetual: ASML on Tradable Compute](https://gpuquant.com/docs/instruments/asml.md): The ASML perpetual on Hyperliquid: one share of ASML, the only maker of the EUV lithography machines the leading-edge fabs run, and what it is exposed to. - [Market Data Series Reference](https://gpuquant.com/docs/market-data.md): Every series on Market Data, one page each: what it measures, how it is built, how it is published, what the site shows of it and under what terms, what to read beside it, and its sources. - [H100 Reference Price: Hyperscaler List Price per GPU-Hour](https://gpuquant.com/docs/market-data/h100-reference-price.md): The GPUQuant H100 reference price: the on-demand list price of one H100 for one hour on the eight-GPU node, averaged across AWS, Azure, Google Cloud and Oracle, monthly since 2023. - [B200 Reference Price: Hyperscaler List Price per GPU-Hour](https://gpuquant.com/docs/market-data/b200-reference-price.md): The GPUQuant B200 reference price: the on-demand list price of one B200 for one hour on the eight-GPU node, averaged across the hyperscalers that publish one, monthly since the first listing. - [A100 Reference Price: Hyperscaler List Price per GPU-Hour](https://gpuquant.com/docs/market-data/a100-reference-price.md): The GPUQuant A100 80GB reference price: the on-demand list price of one A100 for one hour on the eight-GPU node, averaged across AWS, Azure, Google Cloud and Oracle, monthly since December 2021. - [H200 Reference Price: Hyperscaler List Price per GPU-Hour](https://gpuquant.com/docs/market-data/h200-reference-price.md): The GPUQuant H200 reference price: the on-demand list price of one H200 for one hour on the eight-GPU node, averaged across AWS, Azure, Google Cloud and Oracle, monthly since December 2024. - [H100 Neocloud Index: Specialist GPU Cloud List Price per GPU-Hour](https://gpuquant.com/docs/market-data/h100-neocloud-index.md): The GPUQuant H100 neocloud index: the on-demand list price of one H100 for one hour on the specialist GPU clouds, read from each provider’s own price list and averaged, monthly since April 2023. - [B200 Neocloud Index: Specialist GPU Cloud List Price per GPU-Hour](https://gpuquant.com/docs/market-data/b200-neocloud-index.md): The GPUQuant B200 neocloud index: the on-demand list price of one B200 for one hour on the specialist GPU clouds that publish one, read from each provider’s own price list and averaged. - [Silicon Data H100 Rental Price Index (SDH100RT)](https://gpuquant.com/docs/market-data/silicon-data-h100-index.md): SDH100RT, the Silicon Data H100 Rental Price Index: the neo-cloud reading of a modelled daily rental price for H100 capacity, and the index the CME H100 futures settle against. - [Silicon Data H100 Hyperscaler Rental Index](https://gpuquant.com/docs/market-data/silicon-data-h100-hyperscaler-index.md): The hyperscaler reading of the Silicon Data H100 Rental Price Index: the same modelled daily figure for capacity rented from the large clouds, which no contract settles against. - [Silicon Data B200 Rental Price Index (SDB200RT)](https://gpuquant.com/docs/market-data/silicon-data-b200-index.md): SDB200RT, the Silicon Data B200 Rental Price Index: one modelled daily rental price blended across every segment of the B200 market, and the index the CME B200 futures settle against. - [Silicon Data A100 Rental Price Index (SDA100RT)](https://gpuquant.com/docs/market-data/silicon-data-a100-index.md): SDA100RT, the Silicon Data A100 Rental Price Index: the neo-cloud reading of a modelled daily rental price for A100 capacity, shown for context beside the H100 and B200 indices. - [Silicon Data A100 Hyperscaler Rental Index](https://gpuquant.com/docs/market-data/silicon-data-a100-hyperscaler-index.md): The hyperscaler reading of the Silicon Data A100 Rental Price Index: the same modelled daily figure for A100 capacity rented from the large clouds, which no contract settles against. - [Ornn Compute Price Index for the H100 (OCPI)](https://gpuquant.com/docs/market-data/ornn-h100-index.md): The Ornn Compute Price Index for the H100 SXM: a volume-weighted average of the prices H100 rentals actually cleared at, settled daily, shown for the trailing three months with Ornn’s permission. - [Ornn Compute Price Index for the B200 (OCPI)](https://gpuquant.com/docs/market-data/ornn-b200-index.md): The Ornn Compute Price Index for the B200: a volume-weighted average of the prices B200 rentals actually cleared at, settled daily, shown for the trailing three months with Ornn’s permission. - [Ornn Compute Price Index for the A100 (OCPI)](https://gpuquant.com/docs/market-data/ornn-a100-index.md): The Ornn Compute Price Index for the A100 SXM4: a volume-weighted average of the prices A100 rentals actually cleared at, settled daily, shown for the trailing three months with Ornn’s permission. - [Ornn Compute Price Index for the H200 (OCPI)](https://gpuquant.com/docs/market-data/ornn-h200-index.md): The Ornn Compute Price Index for the H200: a volume-weighted average of the prices H200 rentals actually cleared at, settled daily, shown for the trailing three months with Ornn’s permission. - [H100 Forward Curve: What the Market Expects the Rental Index to Be](https://gpuquant.com/docs/market-data/h100-forward-curve.md): The Silicon Data H100 forward curve: the market’s expected path for the H100 rental index from spot out to three years, anchored on the settlement index, shown for the latest seven days. - [B200 Forward Curve: What the Market Expects the Rental Index to Be](https://gpuquant.com/docs/market-data/b200-forward-curve.md): The Silicon Data B200 forward curve: the market’s expected path for the B200 rental index from spot out to three years, anchored on the settlement index, shown for the latest seven days. - [A100 Forward Curve: What the Market Expects the Rental Index to Be](https://gpuquant.com/docs/market-data/a100-forward-curve.md): The Silicon Data A100 forward curve: the market’s expected path for the A100 rental index from spot out to three years, anchored on the neo-cloud index, shown for the latest seven days. - [H100 Residual Value: What One GPU Is Still Worth](https://gpuquant.com/docs/market-data/h100-residual-value.md): GPUQuant’s residual value of one H100: the discounted value of what the GPU can still earn over its remaining life, computed daily from the forward curve, and how closely it tracks Silicon Data’s figure. - [B200 Residual Value: What One GPU Is Still Worth](https://gpuquant.com/docs/market-data/b200-residual-value.md): GPUQuant’s residual value of one B200: the discounted value of what the GPU can still earn over its remaining life, computed daily from the forward curve, and how closely it tracks Silicon Data’s figure. - [A100 Residual Value: What One GPU Is Still Worth](https://gpuquant.com/docs/market-data/a100-residual-value.md): GPUQuant’s residual value of one A100: the discounted value of what the GPU can still earn over its remaining life, computed daily from the forward curve, and how closely it tracks Silicon Data’s figure. - [Silicon Data LLM Token Expenditure Index (SDLLMTK)](https://gpuquant.com/docs/market-data/silicon-data-token-index.md): SDLLMTK, the Silicon Data LLM Token Expenditure Index: what a million tokens of language-model inference costs across labs, weighted by spend, and why its headline can fall while every price rises. - [Ornn Token Price Index for Anthropic (OTPI)](https://gpuquant.com/docs/market-data/ornn-token-index-anthropic.md): The Ornn Token Price Index for Anthropic: the volume-weighted price of a million tokens on Anthropic’s models, settled daily, shown for the trailing month with Ornn’s permission. - [Ornn Token Price Index for OpenAI (OTPI)](https://gpuquant.com/docs/market-data/ornn-token-index-openai.md): The Ornn Token Price Index for OpenAI: the volume-weighted price of a million tokens on OpenAI’s models, settled daily, shown for the trailing month with Ornn’s permission. - [Ornn Token Price Index for Google (OTPI)](https://gpuquant.com/docs/market-data/ornn-token-index-google.md): The Ornn Token Price Index for Google: the volume-weighted price of a million tokens on Google’s models, settled daily, shown for the trailing month with Ornn’s permission. - [Ornn Token Price Index for DeepSeek (OTPI)](https://gpuquant.com/docs/market-data/ornn-token-index-deepseek.md): The Ornn Token Price Index for DeepSeek: the volume-weighted price of a million tokens on DeepSeek’s models, settled daily, shown for the trailing month with Ornn’s permission. - [Silicon Data RAM Index (GDDR6)](https://gpuquant.com/docs/market-data/silicon-data-ram-index.md): The Silicon Data RAM Index: a weighted basket of GDDR6 memory prices published on business days, an input cost under every accelerator, and why it is an index and not a quotable price. - [RamRadar RAM Price Index: DDR5 and DDR4 at Retail, per GB](https://gpuquant.com/docs/market-data/ramradar-ram-price-index.md): The average retail price of a gigabyte of DDR5 and DDR4 memory, as desktop DIMMs and laptop SODIMMs, computed daily by RamRadar from the kits in stock at eBay, Newegg and B&H Photo, shown whole since October 2025 under RamRadar’s citation terms. - [DRAM Chip Spot Prices: DDR5 and DDR4, from DRAMeXchange](https://gpuquant.com/docs/market-data/dramexchange-dram-spot.md): The spot price of a DDR5 16Gb, DDR4 16Gb and DDR4 8Gb chip, as DRAMeXchange publishes it daily, with the history the Internet Archive holds since 2019, and what the site is waiting for before it shows them. - [Memory Module Spot Prices: DDR5 RDIMM and UDIMM, from DRAMeXchange](https://gpuquant.com/docs/market-data/dramexchange-module-spot.md): The spot price of a DDR5 RDIMM 32GB server module and the DDR5 and DDR4 UDIMMs beside it, as DRAMeXchange publishes them, and why no maker’s list price exists for a server module. - [GDDR Chip Spot Prices: GDDR6 and GDDR5, from DRAMeXchange](https://gpuquant.com/docs/market-data/dramexchange-gddr-spot.md): The spot price of a GDDR6 and a GDDR5 8Gb chip, updated roughly monthly by DRAMeXchange, and why GDDR6X and GDDR7 have no row. - [HBM3E Price Estimates: Contract and Spot, per GB](https://gpuquant.com/docs/market-data/hbm3e-price-estimates.md): What HBM3E is said to cost per gigabyte, on long-term contract and on the spot market, as dated estimates quoted from the trade press with the sentence each came from. - [HBM4 Price Estimates: Contract, per GB](https://gpuquant.com/docs/market-data/hbm4-price-estimates.md): What HBM4 is said to cost per gigabyte on contract, and what it is forecast to cost in 2027, as dated estimates quoted from the trade press and analyst notes. ## Pages with the numbers - [Market data](https://gpuquant.com/market-data): the reference price, the neocloud index, the Silicon Data indices and the forward curve, with every provider behind each figure. - [NVIDIA H100 80GB SXM](https://gpuquant.com/gpus/h100): the H100 price history, the providers and the regional offerings behind it. - [NVIDIA B200 180GB](https://gpuquant.com/gpus/b200): the B200 price history, the providers and the regional offerings behind it. - [NVIDIA H200 141GB SXM](https://gpuquant.com/gpus/h200): the H200 price history, the providers and the regional offerings behind it. - [NVIDIA A100 80GB SXM](https://gpuquant.com/gpus/a100-80gb): the A100 price history, the providers and the regional offerings behind it. - [Tradable Compute](https://gpuquant.com/tradable-compute): the instruments, with live prices from their venues. - [Map](https://gpuquant.com/map): every cloud region with a published price for the tracked GPUs, and the facility layer. - [Atlas Search](https://gpuquant.com/atlas): browse and search the sourced facility, company, place and cloud-region records. Each record carries its evidence and verification dates. - [Documentation](https://gpuquant.com/docs): the human-readable entry point. - [MCP server](https://gpuquant.com/mcp): remote HTTP connection instructions, tool inputs, examples and coverage limits for the public infrastructure MCP. - [Community](https://gpuquant.com/community): report a correction or missing infrastructure through the contact form with supporting sources. Reports require maintainer review. - [Contact](https://gpuquant.com/contact): data corrections, questions, press and partnerships. - [Trading terms and risk](https://gpuquant.com/terms): how the trading interface works, fees and risks.