AWS has published official per-accelerator-hour rates for EC2 Capacity Blocks taking effect October 7, 2026 — but without any prior rates to compare them against.
Correction, 1 October 2026. This piece originally reported five rates and a spread of about 2.7 times. The AWS announcement in fact lists seven, including P4de at $2.546 and P4d at $1.696, giving a spread of about 9.5 times. The omission was this publication’s: the quoted passage was read in truncated form and the shorter list was mistaken for the whole. The rates, the spread and the chart above have been corrected, and the central point — that no prior rates are published, so no direction of change is knowable — is unaffected.
AWS publishes these rates without publishing the ones they replace. Nothing below calls this a price rise or a price cut, because the page makes no direction of change knowable.The spread across P4, P5 and P6 is seven different accelerators sold at the same time, not one accelerator over time. The derived ratios are this publication’s arithmetic. Nothing here is investment advice.
AWS has published a pricing page for EC2 Capacity Blocks listing seven accelerator options effective October 7, 2026. The listed rates, per accelerator-hour and across all available AWS Regions, are: P6-B300 at $16.146, P6-B200 at $14.208, P5en at $7.895, P5e at $6.866, and P5 at $5.970, P4de at $2.546 and P4d at $1.696.
In AWS GovCloud Regions, two rates differ: B300 is listed at $16.819 and B200 at $14.801. The page states that all other Capacity Block prices — and prices on fixed models including On-Demand and Savings Plans — remain unchanged.
Capacity Blocks are a distinct reservation mechanism. They allow customers to reserve accelerator capacity for a defined period, priced per accelerator-hour, as an alternative to On-Demand purchasing. The page is a price list, not a specification document.
The key insight: The page publishes no prior rates. No direction of change — up, down, or flat — can be established from what AWS has published. A rate card without a baseline is a menu, not a movement. AWS also says why, in a line added to this piece after publication: the page states that Capacity Blocks reservation prices “are dynamic and are updated periodically based on supply and demand”, and that a block “is charged at the prevailing rate at the time of purchase, even if the Capacity Block is scheduled to start after the price is updated”. So there is no fixed predecessor to publish. The rate is a moving quote, not a list price.
These are seven different accelerators offered at the same time, not one accelerator priced over time. The spread is a menu, not a trend.
What the Numbers Actually Show
The seven rates in the announcement span $1.696 to $16.146. That is a spread of approximately 9.5 times, by this publication’s arithmetic on the listed figures. But that number requires immediate context.
These are seven different accelerator products sold simultaneously, not one product priced at different points in time. The 2.7-times spread is a menu of tiers. It is not inflation, not a pricing trajectory, and not evidence that anything became more expensive.
Within the newest generation, the gap is narrower. B300 at $16.146 sits approximately 13.6 per cent above B200 at $14.208 — again, this publication’s division of two listed rates. The page offers no explanation of what that difference reflects. It is a price list, not a specification.
The Structural Read
The GovCloud premium is the one clean internal comparison the card permits. $16.819 against $16.146 on B300 is approximately 4.17 per cent. $14.801 against $14.208 on B200 is also approximately 4.17 per cent, by this publication’s arithmetic.
The same premium appearing on both products is a consistency worth noting. The page gives no reason for it. This publication is not supplying one.
What the page does not contain matters as much as what it does. There are no previous rates, no percentage change, and no list of which Regions qualify as available. There is no minimum or maximum block duration, no booking window, and no advance-reservation limit.
There is no utilisation figure, no indication of demand, and no competitor comparison. AWS gives no rationale for setting any rate where it has.
What is narrow and real: a current, official, per-unit price for reserved frontier accelerator capacity from one major vendor, on a stated date. That is a datapoint worth having precisely because it is published rather than inferred.
Three Implications
IMPLICATION 1 — THE DATAPOINT HAS VALUE PRECISELY AS STATED A published, official, per-accelerator-hour rate from AWS is a concrete input for anyone building infrastructure budgets or modelling AI workload costs. Its value comes from being sourced and dated, not from implying movement.
IMPLICATION 2 — THE TIERED MENU SIGNALS MARKET SEGMENTATION, NOT TRAJECTORY Seven accelerator products listed simultaneously at different price points reflect a segmented product line. The spread across P5, P5e, P5en, P6-B200 and P6-B300 tells buyers what options exist, not whether the market is moving in any direction. Treating the spread as a trend would be a misread of the data.
IMPLICATION 3 — WHAT IS ABSENT SETS THE FLOOR FOR ANALYSIS Prior rates, percentage changes, Region lists, block durations, booking windows, utilisation data, and competitive pricing are all absent from this page. Any analysis that imports those elements is adding assumptions, not reading the source. The honest baseline is what AWS actually published.
Business Engineer Framework
The FDE Framework: Founders, Distributors, Enablers
AWS sits squarely in the Enabler layer — the infrastructure substrate on which AI builders operate. Understanding where a vendor sits in the FDE stack changes how you read a pricing page. A rate update from an Enabler is a signal about access cost, not market direction. The Map of AI makes this stack legible across 200+ companies.
AWS has put seven official accelerator-hour rates on the record for October 7, 2026. That is the complete, verifiable claim the page supports. No prior rate appears anywhere on it, so no direction of change is knowable, no performance comparison follows from the prices alone, and the 2.7-times spread is a product menu, not a market signal. Read it as what it is: a dated, sourced cost of access — and nothing more than that.
Every rate above comes from the AWS EC2 Capacity Blocks pricing page, read directly on 1 October 2026. The page publishes the rates effective 7 October 2026 but does not publish the rates they replace, and gives no percentage change. It does state that these reservation prices are dynamic and updated periodically based on supply and demand, and that a block is charged at the prevailing rate at the time of purchase — a fact added to this piece shortly after publication, and one that explains the absence of a baseline rather than excusing it.
Nothing above therefore describes any rate as an increase or a reduction, because the page makes no direction of change knowable.Three figures above are this publication’s arithmetic on the listed rates rather than numbers AWS states: the spread of about 9.5 times between P4d and P6-B300, the gap of about 13.6 per cent between P6-B300 and P6-B200, and the GovCloud premium of about 4.17 per cent on both P6 products.
The spread across P4, P5 and P6 compares seven different accelerators offered at the same time. It is not one accelerator priced over time, and it is not evidence of inflation or of any price trajectory.The page is a price list and carries no specifications. Nothing above claims any accelerator is faster, more capable or better value per unit of work than another, and no reason is offered for the GovCloud premium because the page gives none.
Also absent: prior rates, which Regions count as available, minimum or maximum block durations, how far ahead capacity may be reserved, utilisation, competitorpricing, and AWS’s rationale. Amazon is a listed company. Nothing above predicts anything, and nothing here is investment advice.
Gennaro is the creator of FourWeekMBA, which reached about four million business people, comprising C-level executives, investors, analysts, product managers, and aspiring digital entrepreneurs in 2022 alone | He is also Director of Sales for a high-tech scaleup in the AI Industry | In 2012, Gennaro earned an International MBA with emphasis on Corporate Finance and Business Strategy.
Scroll to Top
Discover more from FourWeekMBA
Subscribe now to keep reading and get access to the full archive.
