--- title: "Megawatt: what it means · The AI Compute Race · DRYBURGH" source: https://dryburgh.com/vol1/glossary/megawatt description: "Megawatt: A thousandth of a gigawatt, written MW. A kilowatt, written kW, is a thousandth of a megawatt. Chapter 6." --- # Megawatt What does “Megawatt” mean? MegawattA thousandth of a gigawatt, written MW. A kilowatt, written kW, is a thousandth of a megawatt. *Chapter 6.* From *The AI Compute Race*. [Read the book](https://dryburgh.com/vol1#buy) ## Where the book uses it - [Chapter 6 — Firm, Around-the-Clock](https://dryburgh.com/vol1/notes/chapter-6)Where the electricity comes from: grid queues, transformers, turbines and the firm power a datacenter needs every hour of the year, on calendars the buildout does not control. ## Where the record uses it - Campus [ByteDance (TikTok) Pecém](https://dryburgh.com/vol1/ledger/bytedance-tiktok-pecem) - Campus [Z.ai domestic-chip campus](https://dryburgh.com/vol1/ledger/z-ai-domestic-chip-campus) - Dated test [Chapter 6, firm service is the constraint, due 31 December 2028](https://dryburgh.com/vol1/tests#2028-12) - Dated test [Chapter 12, claim 3, a bargain kept, due 31 December 2028](https://dryburgh.com/vol1/tests#2028-16) - Dated test [Chapter 12, claim 3, prevention, due 31 December 2028](https://dryburgh.com/vol1/tests#2028-17) ## Sources that use it - Harsh Singh Banwait, “NVIDIA Vera Rubin NVL72 on CoreWeave: 10x More Tokens Per Megawatt Than Blackwell,” CoreWeave blog, 21 July 2026 (CoreWeave’s own measurement, on DeepSeek R1: “At a matched interactivity target, Vera Rubin NVL72 generated 10x tokens-per-second per megawatt compared to NVIDIA GB200 NVL72 on the same DeepSeek R1 workload”; the conditions, “with all major, state-of-the-art inference optimizations turned on,” among them “large-scale expert parallelism, NVFP4 precision, multi-token prediction, and disaggregated prefill and decode”; the page’s other wording of the result, “up to 10x higher tokens per megawatt”; what it means for customers, in the company’s words: “run far more reasoning-model traffic within the same power budget or run the same traffic on far less power”; CoreWeave sells the capacity it measured, and NVIDIA’s blog of the same day relays the figure as its own platform’s; the comparison on both systems, “This evaluation compares the performance on each platform with all major, state-of-the-art inference optimizations turned on”; read in the page as served in October 2026, which carries its publication date). [In the Sources](https://dryburgh.com/vol1/sources/company#coreweave-vera-rubin-jul26) - Dylan Patel, “Meta Discusses AI Hardware and Co-packaged Optics,” *SemiAnalysis*, 15 September 2022 (a report of a talk at the AI Hardware Summit in Santa Clara by Alexis Black Bjorlin, Meta’s vice president of infrastructure hardware: “They commented that a large training cluster could be as much as 6 Megawatts today. They said that these training clusters would be 64 Megawatts in the future”; the figures are the post’s account of the talk, and neither is on a slide the post reproduces nor in any recording or program found on 5 October 2026) (SemiAnalysis sells subscriptions, consulting and its own models of chip, memory and datacenter supply into the markets it measures; held whole in the Library since Sourcing’s capture of 5 October 2026). [In the Sources](https://dryburgh.com/vol1/sources/research#semianalysis-meta-sep22) ## Related terms - [Gigawatt](https://dryburgh.com/vol1/glossary/gigawatt) - [Kilowatt](https://dryburgh.com/vol1/glossary/kilowatt) - [Evidenced](https://dryburgh.com/vol1/glossary/evidenced) - [kW](https://dryburgh.com/vol1/glossary/kw) - [Megawatt-hour](https://dryburgh.com/vol1/glossary/megawatt-hour) - [MW](https://dryburgh.com/vol1/glossary/mw) - [Backstop](https://dryburgh.com/vol1/glossary/backstop) - [Batch Zero](https://dryburgh.com/vol1/glossary/batch-zero) - [Behind the meter](https://dryburgh.com/vol1/glossary/behind-the-meter) - [Beignet](https://dryburgh.com/vol1/glossary/beignet) - [Bridge power](https://dryburgh.com/vol1/glossary/bridge-power) - [Capacity auction](https://dryburgh.com/vol1/glossary/capacity-auction) From the glossary of The AI Compute Race. - [Campuses](https://dryburgh.com/vol1/ledger) - [Tests](https://dryburgh.com/vol1/tests) - [Scorecard](https://dryburgh.com/vol1/scorecard) - [Data](https://dryburgh.com/vol1/data/releases) - [Chapter by chapter](https://dryburgh.com/vol1/companion) - [The record at a glance](https://dryburgh.com/vol1/observatory) - [The machine](https://dryburgh.com/vol1/machine) - [One campus, four futures](https://dryburgh.com/vol1/hyperion) - [Glossary](https://dryburgh.com/vol1/glossary) - [People](https://dryburgh.com/vol1/people) - [Sources](https://dryburgh.com/vol1/sources) - [Notes](https://dryburgh.com/vol1/notes) - [Where the book was wrong](https://dryburgh.com/vol1/corrections) - [Where its sources were wrong](https://dryburgh.com/corrections?book=vol1) - [The Paper](https://dryburgh.com/vol1/paper) - [Consent](https://dryburgh.com/vol1/consent) - [Who builds and pays](https://dryburgh.com/vol1/parties) *The AI Compute Race: Five Races, One Machine* is a book by Lee S. Dryburgh, the first volume of THE COMPUTE ECONOMY, published by Dryburgh, an imprint of Contestra OÜ, Tallinn, Estonia; its record closes on 8 November 2026. [About the book](https://dryburgh.com/vol1)[Tell me when it is out](https://dryburgh.com/vol1#notify)