Nvidia B200 Rent Jumped 30% in Two Weeks to $9.70 an Hour

The price of renting an Nvidia B200 rose from $7.45 to $9.70 per GPU-hour between 11 and 25 October 2026 on the Ornn Compute Price Index, a 30.2% rise by our arithmetic from Ornn’s settled values. All five GPUs in Ornn’s free data rose over the same two weeks.

Analyst Shanu Mathew set out the moves in a post on X on 26 October 2026, drawing on Ornn’s datasets. We checked every GPU rental price in this piece against Ornn’s public daily series and every token price against Ornn’s Token Price Index table, both read on 27 October; the reservation rate is Mathew’s reading.

Video explainer · Nvidia GPU rents

A 16:9 video explainer narrated from the published text of this piece and three related FourWeekMBA pieces.

The key insight: As we read it, the B200 led a move that touched all five GPUs in Ornn’s free data, but the index records prices, not how much trading sits behind them.

What Ornn’s Index Shows

Ornn describes each index value as a “volume-weighted, winsorized average of transacted GPU rental prices” that settles once per trading day. Between 11 and 25 October, by our arithmetic from its settled values, the B200 rose 30.2%, the H100 SXM 14.3% ($2.65 to $3.03), the A100 SXM4 13.5% ($0.96 to $1.09), the H200 12.3% ($5.61 to $6.30) and the RTX PRO 6000 7.7% ($1.42 to $1.53).

The B200 settled at $9.76 on 26 October, and Ornn’s own table shows it up 21.4% over one month. The B200 and H100 indices carry Bloomberg tickers, ORNNB200 and ORNNH100, according to Ornn’s data page.

Ornn Compute Price Index, US$ per GPU-hour, 11 and 25 October 2026: B200 $7.45 to $9.70 (+30.2%), H200 $5.61 t
Ornn Compute Price Index, US$ per GPU-hour, 11 and 25 October 2026: B200 $7.45 to $9.70 (+30.2%), H200 $5.61 to $6.30 (+12.3%), H100 SXM $2.65 to $3.03 (+14.3%), RTX PRO 6000 $1.42 to $1.53 (+7.7%), A100 SXM4 $0.96 to $1.09 (+13.5%). Percent changes are our arithmetic from Ornn’s settled values. Source: Ornn Data public daily series.
Change in the Ornn Compute Price Index between 11 and 25 October 2026: B200 +30.2%, H100 SXM +14.3%, A100 SXM4
Change in the Ornn Compute Price Index between 11 and 25 October 2026: B200 +30.2%, H100 SXM +14.3%, A100 SXM4 +13.5%, H200 +12.3%, RTX PRO 6000 +7.7% (our arithmetic from Ornn’s settled values). Source: Ornn Data public daily series.

What Was Posted on X

Shanu Mathew posted on 26 October that Ornn’s GPU spot price indices rose across all five public series over the two weeks to 25 October, with the B200 up 30.2% from $7.45 to $9.70 per GPU-hour, and noted that without transaction volumes, fleet counts and contributor concentration the post cannot establish how much activity supports those moves.

An X user posted that Deutsche Bank estimates Nvidia GB300 GPUs can rent for up to about $14 an hour versus $3.50 for H100s, with some contracts generating about $60M per megawatt a year; the poster puts 1 GW of AI compute at about $48B of annual revenue based on blended pricing.

Why the Start Date Matters

Nvidia B200, US$ per GPU-hour: spot $9.70 on 25 Oct and $7.45 on 11 Oct (Ornn Compute Price Index), against a
Nvidia B200, US$ per GPU-hour: spot $9.70 on 25 Oct and $7.45 on 11 Oct (Ornn Compute Price Index), against a one-year reservation mark of $5.67 as read by Shanu Mathew from Ornn’s published marks. The 71% premium is Mathew’s arithmetic.

Mathew noted that “The starting date changes the interpretation.” On Ornn’s series the B200 settled at $8.09 on 30 September, so the rise to 25 October is 19.9% by our arithmetic, against 30.2% from 11 October, when the index had dipped.

The gap between the B200 and the H200 moved too. By our arithmetic from Ornn’s values, the B200 cost 68% more per hour than the H200 on 26 September, 22% more on 15 October and 54% more on 25 October.

What Mathew Found Beyond Prices

Two of Mathew’s readings come from Ornn data we could not open without an API key. Mathew wrote that the B200’s tracked fleet saw its rented share fall from approximately 94.3% to 87.8% over the month, and that “A growing fleet could explain falling occupancy even if the number of rented GPUs increased.” Mathew added: “Ornn doesn’t disclose the fleet counts needed to distinguish those possibilities.”

Mathew also wrote that B200 spot “sits 71% above Ornn’s published one-year reservation rate of $5.67/GPU-hour, versus 51% a week earlier”, and that the one-year and three-year reservation rates were identical in captures on 18 and 26 October. In Mathew’s words: “Unchanged published marks don’t establish whether newly negotiated contracts repriced.”

Anthropic and OpenAI Token Prices Fell

Ornn publishes a second index for what AI labs charge, in dollars per million tokens. Its table, settled 25 October, shows Anthropic at $0.85, down 43.5% over one month, OpenAI at $0.27, down 34.3%, and DeepSeek at $0.04, down 35.2%, while Google rose 9.3% to $0.60. Over roughly the same month, the B200 rental index rose 21.4% on Ornn’s own figures.

Connecting the Coverage

Earlier today we reported that traders can now bet 10x on Nvidia GPU rents through Injective’s new perpetual futures. On 24 October we reported that Nvidia H100 rent hit $2.88 an hour, up 44% since December, on Silicon Data’s index, and on 25 October that Nvidia B300 rent was up 52% since May as AI token prices fell.

On 25 October we also reported that Nvidia GPU rents are now priced to rise for 36 months, and on 26 October that neocloud stocks split after Musk’s warning as their bonds hit lows.

The Structural Read

As we read it, the spot and term prices tell different stories: on Mathew’s reading, B200 spot sits 71% above Ornn’s published one-year reservation rate, and the one-year and three-year marks were unchanged between 18 and 26 October.

As we read it, the start date carries much of the headline: on Ornn’s values the B200 is up 30.2% from its 11 October dip and 19.9% from 30 September.

As we read it, the B200 rental index and the Anthropic and OpenAI token price indices moved in opposite directions over the month in Ornn’s own tables: the B200 up 21.4%, Anthropic down 43.5% and OpenAI down 34.3%, while Google’s token index rose 9.3%.

Shanu Mathew, on X, 26 October 2026

“Unchanged published marks don’t establish whether newly negotiated contracts repriced”

Three Implications

FOR RENTERS On Ornn’s index the B200 settled at $9.76 on 26 October, up 21.4% over one month according to Ornn’s table.

FOR NEOCLOUDS Mathew wrote that the B200’s tracked fleet saw its rented share fall from approximately 94.3% to 87.8% over the month, and that a growing fleet could explain it; Ornn does not publish the fleet counts.

FOR TRADERS Ornn says its B200 and H100 indices are listed on Bloomberg as ORNNB200 and ORNNH100, and that each value settles once per trading day.

The Business Engineer Lens

This story maps onto the Business Engineer framework The AI Supercycle.

The framework describes the AI economy as a layered industrial stack, in which “the binding constraint sits at a different layer than it did six months ago”.

As we read it, the rental indices are one place to watch that constraint at the compute-capacity layer: a two-week jump in what an hour of B200 time costs. The framework was published before this data and does not discuss Ornn’s indices.

What Is Not Established

In Mathew’s words: “Without transaction volumes, fleet counts and contributor concentration, I can’t establish how much activity supports those moves.” Ornn’s free data gives prices only. Its utilization readings and forward marks need an API key, so the rented-share and reservation figures above are Mathew’s readings, not ours. A two-week move on one index does not show where rents go next.

This Week In The AI Supercycle

Execution Fault Lines, Concentrated Demand, and Geopolitical Firewalls

This week’s issue: the AI build is not short of demand for its parts. It is short of cheap money, and its end revenue rests on a narrow base of large, often venture-funded buyers.

Read the full issue →

Business Engineer Framework

The AI Supercycle

A Business Engineer framework for the AI build, from energy at the bedrock to governance at the top.

Read the Map of AI →

The Bottom Line

On Ornn’s index the B200 went from $7.45 to $9.70 an hour in two weeks, and all five tracked Nvidia GPUs rose; how much trading sits behind those prices is not public.

95,000+ executives read Business Engineer for the AI strategy frameworks cited by ChatGPT, Claude, and Perplexity.

A note on sourcing. GPU rental prices are from Ornn Data’s public daily series of the Ornn Compute Price Index, and token prices from Ornn’s Token Price Index table, both read on data.ornn.com and Ornn’s free API on 27 October 2026. Percent changes and the B200-to-H200 premiums are our arithmetic from Ornn’s settled values. The rented-share figures, the one-year reservation rate and the 71% and 51% premiums are Shanu Mathew’s readings of Ornn data posted on X on 26 October 2026; that data needs an Ornn API key, which we did not use. This is not investment advice. We did not contact Ornn.

Sources: Ornn Data, Compute Price Index and Token Price Index tables (settled 25-26 Oct 2026) · Ornn Data, B200 market page and public daily series (api.ornnai.com, read 27 Oct 2026) · Shanu Mathew on X, 26 Oct 2026: analysis of Ornn datasets

Scroll to Top
FourWeekMBA