NVIDIA’s Blackwell family – the B200, the Blackwell Ultra B300, the rack-scale DGX B300 system, and the 72-GPU NVL72 racks (GB200 NVL72 and the newer GB300 NVL72) – sits at the center of AI infrastructure spending in 2026. Buyers are juggling on-prem capex, cloud GPU-hour rates, and rapidly shifting availability. This breakdown collects the latest verified pricing signals, updated through October 13, 2026, so you can benchmark quotes against the live market.
Don't miss new tech stories on Google
Add Tech Insider once in the Google app and our stories appear in your news suggestions.
B200 vs. B300 vs. DGX vs. NVL72: What Each Name Means
The Blackwell lineup packs a lot of model numbers into one family, and the naming trips up plenty of buyers before pricing even enters the conversation. B200 is the standard Blackwell GPU, built with 192GB of HBM3e memory. B300 is the Blackwell Ultra variant – the same core architecture paired with 288GB of HBM3e, about 50% more – and NVIDIA positions it specifically for AI reasoning workloads with long inference chains and large context windows.
DGX is NVIDIA’s systems brand, not a chip. A DGX B200 or DGX B300 is a turnkey 8-GPU server bundling the GPUs with NVIDIA’s interconnect fabric, system memory, and software stack, while the DGX Station is the single-GPU desktop version of the same idea, pairing one B300 with a Grace CPU. Scale up further and the naming shifts to rack level: GB200 NVL72 and GB300 NVL72 each bundle 72 GPUs – standard B200s in the former, Blackwell Ultra B300s in the latter – into a single liquid-cooled rack with full NVLink switching between every GPU. In short: B200/B300 name the chip, DGX names NVIDIA’s pre-built server around it, and NVL72 names the full rack-scale version. Pricing for every tier follows below.
Blackwell Pricing at a Glance
For readers who want the headline numbers before the sourcing detail, here’s every Blackwell price point in this guide as of October 2026:
- B200 cloud rate: $5.99–$16.11 per GPU-hour across 11 named providers in the latest September 2026 cloud comparison, from Hyperbolic’s floor to Google Cloud’s hyperscaler ceiling – or as low as $5.29/hr at Lambda Labs per GridStackHub’s narrower 6+ provider index, while MorphLLM’s October 2026 read puts the specialized tier (Modal, Lambda, Runpod) at $6.25–$6.79/hr, about 1.4x an H200
- B300 cloud rate: $7.10–$17.80 per GPU-hour across five named on-demand providers, from Modal’s floor to AWS’s ceiling – reconfirmed October 8, 2026 by Telnyx, and MorphLLM’s October 2026 read puts Modal at $7.10 and Runpod at $7.89, a 14–16% premium over B200 – Nebius raised its on-demand rate 21% to $9.50/hr on October 1, 2026, and premium managed DGX B300 stacks run as high as $18.00/hr, or as low as $3.13/hr on a 48-month reservation
- DGX B200 system (8 GPUs): $280,000–$320,000 OEM anchor, or $562,154.09 for Broadberry’s named US configuration listing (September 21, 2026, via GPUPerHour)
- DGX B300 system (8 GPUs): $300,000–$350,000 OEM anchor – holding since Q1 2026 and reconfirmed October 8, 2026 by Telnyx and October 13, 2026 by IntuitionLabs – or $646,878.46 for Broadberry’s named US configuration listing (September 21, 2026, via GPUPerHour)
- DGX Station (1 GPU, desktop): $80,000–$125,000, reconfirmed October 13, 2026 (IntuitionLabs)
- Standalone B200/B300 GPU: B200 street price climbed to $45,000–$50,000 as of October 2, 2026 (IntuitionLabs), up from a $30,000–$40,000 range quoted as recently as September 2026; B300 runs about $53,000, reconfirmed October 13, 2026 (IntuitionLabs)
- GB200 NVL72 rack (72 GPUs, estimate): $2 million–$3 million
- GB300 NVL72 rack (72 GPUs, estimate): $3 million–$4 million
The rest of this guide breaks down where each figure comes from, how the tiers compare, how pricing has shifted since launch, and when renting beats buying.
Where These Prices Come From
This guide anchors every figure to named, dated sources rather than generic estimates, current as of October 2026. NVIDIA itself still hasn’t published an official per-GPU retail price for B200 or B300, or its own system price for either DGX machine – but named reseller Broadberry’s default US configuration listings, dated September 21, 2026 and reported by GPUPerHour, put a fully configured DGX B200 appliance at $562,154.09 and a fully configured DGX B300 appliance at $646,878.46, the clearest named, dated system-level prices in this guide. Every other per-GPU and system figure below remains an OEM quote or early-listing estimate, not a manufacturer list price. The most recently dated confirmations come from IntuitionLabs, which priced standalone B200 at $45,000–$50,000 on October 2, 2026 and reconfirmed the standalone B300 at about $53,000, the DGX B300 system at $300,000–$350,000, and the DGX Station at $80,000–$125,000 on October 13, 2026, and from Telnyx, which put B300 rental pricing at $7.10–$17.80 per GPU-hour and confirmed the DGX B300 system price on October 8, 2026.
- NVIDIA’s DGX B300 product page (live on nvidia.com) – system specs and positioning
- iFactory App – DGX B200-vs-B300 system pricing (Q1–April 2026), corroborated by IntuitionLabs’ September 2026 pricing guide and its October 2, 2026 and October 13, 2026 follow-ups for the DGX B300 anchor, the $53,000 B300 per-GPU price, and DGX Station pricing – and separately for a B200 per-GPU range that moved from $30,000–$50,000 to $45,000–$50,000 – and by Thunder Compute for DGX B200 MSRP
- An 11-provider cloud comparison – Hyperbolic, Hyperstack, Modal, Lambda, Runpod, Nebius, Vultr, CoreWeave, Oracle Cloud, AWS, and Google Cloud – for current B200 pricing, a broader read than GridStackHub’s narrower index of 6+ specialized providers
- Thunder Compute (September 4, 2026 pricing table, rates normalized to an 8-GPU node) – current B300 GPU-hour pricing across Modal, Runpod, Nebius, Oracle Cloud, and AWS, plus a separate index naming Spheron’s on-demand SXM6 listing and premium managed DGX B300 stacks
- MorphLLM (October 2026) – the freshest named-provider comparison in this guide, confirming current Modal, Runpod, and Lambda B200 pricing and Modal/Runpod B300 pricing, and framing the B300-over-B200 premium alongside a B200-to-H200 rate comparison
- The Motley Fool (October 1, 2026) – reports Nebius raised its on-demand B300 rate 21%, from $7.85 to $9.50 per GPU-hour, the most recent named-provider B300 price move in this guide
- Telnyx (October 8, 2026) – the most recently dated named-provider read in this guide, confirming B300 rental pricing at $7.10–$17.80 per GPU-hour and an 8-GPU DGX B300 system still at $300,000–$350,000
- GPUAAS (July 2026) – wholesale and hyperscaler B300 bands that corroborate the named-provider figures
- GPU Smith (September 2026) – B200 per-GPU pricing inside an 8-GPU HGX cluster and fully configured 8-GPU HGX B200 server pricing
- Broadberry, via GPUPerHour (September 21, 2026) – named US configuration listings for the fully configured DGX B200 and DGX B300 appliances, the clearest dated system-level prices in this guide
- The Next Platform (September 21, 2026) – confirms eight-way B200 or B300 server nodes price out at roughly $400,000–$500,000, corroborating the high end of the DGX B200 and DGX B300 system estimates
Earlier-2026 cloud pricing from Spheron Network, Verda, and GPUAAS – including a B200 floor near $2.71/hr and B300 spot pricing near $2.45/hr – has since been superseded by the tighter, higher readings referenced throughout this guide, though GPUAAS’s July 2026 wholesale and hyperscaler B300 bands still corroborate the current named-provider figures. IntuitionLabs’ own September 2026 pricing guide corroborates the B300 per-GPU price and adds a $30,000–$50,000 B200 per-GPU range alongside a $400,000–$500,000 DGX-style 8-GPU B200 system estimate, while The Next Platform, on September 21, 2026, put eight-way B200 or B300 server nodes at “maybe $400,000 to $500,000,” corroborating the top of that range. A separate September 2026 pricing guide contributes the 48-month reserved B300 rate, and a July 2026 guide contributes the lower B200 street-price estimate – corroborated by GPU Smith’s September 2026 B200 pricing page, which also contributes the B200 manufacturing-cost figure and the GB200 NVL72 rack estimate used below, while Broadberry’s September 21, 2026 named US configuration listings, reported by GPUPerHour, contribute the clearest dated DGX B200 and DGX B300 appliance prices in this guide. The GB300 NVL72 and GB200 NVL72 rack prices are sourced from broader market estimates rather than these primary named sources, so treat them as directional estimates rather than confirmed, dated quotes. Most recently, The Motley Fool reported that Nebius raised its on-demand B300 rate 21% on October 1, 2026, from $7.85 to $9.50 per GPU-hour, detailed in the B300 cloud pricing section below, while MorphLLM’s own October 2026 comparison is the freshest read on the B200 and B300 floors, confirming Modal, Runpod, and Lambda’s B200 rates and Modal and Runpod’s B300 rates, and framing B200 as roughly 1.4x an H200’s rate and B300 as a 14%-16% premium over B200. The freshest data points in this guide now come from Telnyx, which confirmed B300 rental pricing at $7.10–$17.80 per GPU-hour and the $300,000–$350,000 DGX B300 system price on October 8, 2026, and from IntuitionLabs, which priced standalone B200 at $45,000–$50,000 on October 2, 2026 and reconfirmed the standalone B300, the DGX B300 system, and the DGX Station at $80,000–$125,000 on October 13, 2026.
DGX B300 System Pricing
NVIDIA positions the DGX B300 as the Blackwell Ultra system built for AI reasoning workloads. The system began shipping in January 2026 and requires mandatory direct liquid cooling, pairing each GPU with 288GB of HBM3e memory – about 50% more than the standard B200’s 192GB. NVIDIA’s own DGX B300 product page is now live on nvidia.com, confirming these specs directly from the manufacturer. As of Q1 2026, and still holding as of October 2026 – most recently confirmed by Telnyx on October 8, 2026 and by IntuitionLabs on October 13, 2026 – an 8-GPU DGX B300 system is anchored at $300,000–$350,000, which works out to roughly $37,500–$43,750 per GPU at the system level – a baseline that hasn’t moved even as named reseller Broadberry’s default US configuration listing, dated September 21, 2026, prices a fully configured DGX B300 appliance at $646,878.46, or roughly $80,860 per GPU.
What the System Price Includes
The 8-GPU configuration bundles the GPUs with NVIDIA’s interconnect fabric, system memory, and the integrated software stack. The per-GPU figure above is an effective rate derived from the full $300,000–$350,000 system price – it is not the same as buying GPUs standalone. A single B300 GPU purchased outright runs about $53,000 (Spheron Network, July 5, 2026; reconfirmed by IntuitionLabs on October 13, 2026), now a narrower premium over the single B200‘s $45,000–$50,000 street price – up from the $30,000–$40,000 B200 band quoted as recently as September 2026, before IntuitionLabs confirmed on October 2, 2026 that October 2026 supply constraints had pushed standalone pricing higher. Broadberry’s $646,878.46 named configuration listing (September 21, 2026) sits well above that $300,000–$350,000 anchor, illustrating how much a fully configured, reseller-quoted build with support and integration options layered on can add over the base system price.
DGX B200 vs DGX B300 System Pricing
With DGX B300 shipments underway, DGX B200 8-GPU systems now have a clearer price anchor too: $280,000–$320,000 as of April 2026 (iFactory App, Thunder Compute), or roughly $35,000–$40,000 per GPU at the system level – compared to the DGX B300’s $300,000–$350,000 ($37,500–$43,750 per GPU) as of Q1 2026, an anchor still holding as of October 2026. The gap reflects the cost of the B300’s jump to 288GB HBM3e memory and the Blackwell Ultra reasoning-optimized platform. The DGX B200 figure is more source-dependent than the B300 figure, though: other 2026 pricing guides place an 8-GPU DGX B200 system as high as $370,000–$500,000, which would erase or even reverse the usual DGX B300 premium depending on configuration. This guide treats iFactory App’s $280,000–$320,000 as the primary DGX B200 anchor since it’s the most specific, consistently-dated quote, but real-world quotes well above $320,000 – up to Broadberry’s $562,154.09 configured listing – should be expected. GPU Smith’s September 2026 B200 pricing page backs up that high end from a different angle: it prices a fully configured 8-GPU HGX B200 server – the OEM reference platform DGX systems are built on – at roughly $400,000–$500,000, even though it prices standalone B200 chips inside that same 8-GPU cluster at just $30,000–$40,000 per GPU. The gap between the two GPU Smith figures underscores how much of the system-level price sits in networking, chassis, and integration rather than the silicon itself. A named reseller listing pushes the ceiling higher still: Broadberry’s default US configuration for the DGX B200, dated September 21, 2026, prices the appliance at $562,154.09 – or about $70,269 per GPU – the highest and clearest named, dated DGX B200 system figure in this guide, well above both the $280,000–$320,000 OEM anchor and the $400,000–$500,000 HGX B200 server quote. The same Broadberry listing prices the DGX B300 appliance at $646,878.46 (about $80,860 per GPU), confirming the B200 stays materially cheaper than the B300 even at this fully configured, named-reseller level. IntuitionLabs’ own September 2026 data-center GPU price table lands in the same territory independently, pricing standalone B200 chips at $30,000–$50,000 per GPU and a DGX-style 8-GPU B200 system at $400,000–$500,000 – and The Next Platform confirmed on September 21, 2026 that an eight-way server node built on either B200 or B300 runs “maybe $400,000 to $500,000,” showing the same $400,000–$500,000 band holds for B300-based systems too, not just B200.
DGX B200 and DGX B300 Appliance Pricing (Broadberry, September 2026)
For buyers who want a fully configured, named-source number rather than a bare OEM anchor, reseller Broadberry’s default US configuration listings, dated September 21, 2026 and reported by GPUPerHour, price the DGX B200 appliance at $562,154.09 (about $70,269 per GPU) and the DGX B300 appliance at $646,878.46 (about $80,860 per GPU) across their respective 8-GPU systems. These are the clearest named, dated system-level prices in this guide, sitting well above the $280,000–$320,000 DGX B200 and $300,000–$350,000 DGX B300 OEM anchors used for the breakeven and TCO math elsewhere in this guide, and above the $400,000–$500,000 fully configured HGX B200 server quote. Buyers configuring a system through a reseller like Broadberry should budget closer to these figures; the lower OEM-anchor ranges reflect base system pricing before reseller markup, support, and configuration options are added.
Why DGX B300 Commands a Premium
Blackwell Ultra is designed for AI reasoning – workloads where long chains of inference and large context windows dominate the compute profile. The DGX B300 is NVIDIA’s reference platform for that class of workload, which is why it sits at the top of the Blackwell capex stack rather than competing on raw $/FLOP with the standard B200.
DGX Station: A Desktop Entry Point
Not every DGX B300 buyer needs a rack-mount, 8-GPU system. IntuitionLabs documents a DGX Station – a desktop tower pairing a single Blackwell Ultra B300 GPU with a Grace CPU – priced at $80,000–$125,000, reconfirmed by IntuitionLabs on October 13, 2026. That’s well below the 8-GPU DGX B300’s $300,000–$350,000 system anchor, giving smaller teams and single-workstation deployments a lower-cost way to access Blackwell Ultra without buying into a full rack-mount system.
B200 Cloud GPU-Hour Pricing
For teams that prefer to rent rather than buy, the B200 is widely available on the cloud. As of September 2026, a broader named-provider comparison spanning 11 platforms puts live B200 GPU-hour rates at $5.99 to $16.11 per GPU-hour – Hyperbolic holds the $5.99 floor, Hyperstack follows at $6.00 and Modal at $6.25, and the range climbs through Lambda ($6.79), Runpod ($6.79), Nebius ($7.15), Vultr ($8.50), and CoreWeave ($8.60) before jumping to hyperscaler pricing at Oracle Cloud ($14.00), AWS ($14.24), and Google Cloud ($16.11), the ceiling. That’s a wider range than the tighter $5.29–$7.05/hr band a narrower September 2026 index of 6+ specialized providers had shown, since this fuller comparison folds in higher hyperscaler rates the narrower survey didn’t track. It’s still well above the deep spot and promotional pricing – as low as $2.71/hr – that some providers were offering earlier in 2026, reflecting a shift from launch-window promotional rates toward steadier, mainstream cloud pricing as B200 supply has matured. MorphLLM’s October 2026 provider comparison corroborates the specialized tier, putting Modal at $6.25 and both Lambda and Runpod at $6.79 per GPU-hour – about 1.4x the going rate for an H200 – confirming the floor has held steady into October rather than drifted further.
B200 Cloud Pricing Snapshot
| Metric | Value | Source |
|---|---|---|
| Absolute lowest documented rate | $5.29/hr (Lambda Labs) | GridStackHub September 2026 index |
| Floor (11-provider comparison) | $5.99/hr (Hyperbolic) | September 2026 cloud comparison |
| Ceiling (11-provider comparison) | $16.11/hr (Google Cloud) | September 2026 cloud comparison |
| Providers surveyed (fuller comparison) | 11 | September 2026 cloud comparison |
B200 Provider Price Comparison
| Provider | B200 Price (per GPU-hour) | As of |
|---|---|---|
| Hyperbolic | $5.99 | September 2026 |
| Hyperstack | $6.00 | September 2026 |
| Modal | $6.25 | October 2026 |
| Lambda | $6.79 | October 2026 |
| Runpod | $6.79 | October 2026 |
| Nebius | $7.15 | September 2026 |
| Vultr | $8.50 | September 2026 |
| CoreWeave | $8.60 | September 2026 |
| Oracle Cloud | $14.00 | September 2026 |
| AWS | $14.24 | September 2026 |
| Google Cloud | $16.11 | September 2026 |
The Full September 2026 B200 Provider Comparison
The fullest dated look at B200 cloud rates comes from an September 2026 comparison spanning 11 named providers, a broader read than GridStackHub’s narrower September 2026 index of 6+ specialized providers, which puts the market inside a tighter $5.29–$7.05 per GPU-hour band anchored by Lambda Labs’ $5.29 floor – the single lowest B200 rate documented across any named September 2026 source. The wider comparison shows the specialized-provider tier still pricing close to that narrower range – Hyperbolic, Hyperstack, Modal, Lambda, Runpod, and Nebius all fall between $5.99 and $7.15 per GPU-hour – but adds Vultr and CoreWeave at $8.50–$8.60, then a distinct hyperscaler tier at Oracle Cloud, AWS, and Google Cloud running $14.00 to $16.11 per GPU-hour. Earlier in 2026, some providers listed B200 capacity as low as $2.71 per GPU-hour, with hyperscaler-grade listings for smaller GPU configurations running as high as $27.04 per GPU-hour – a roughly 10x spread. The September 2026 comparison narrows that considerably to roughly 2.7x between its $5.99 floor and $16.11 ceiling, suggesting the steepest early promotional discounts have faded even as hyperscaler pricing keeps the top of the range elevated.
Reading the Spread
The gap between the cheapest and most expensive B200 listings depends heavily on which provider tier you compare. As of September 2026, the full 11-provider comparison’s $5.99 floor and $16.11 ceiling put the spread at roughly 2.7x – narrower than the near-10x spread seen earlier in 2026, when promotional floors and hyperscaler ceilings sat much further apart, but wider than the 1.3x spread a narrower September 2026 reading of 6+ specialized providers had shown before hyperscaler pricing entered the comparison. A single B200 running 24/7 for a month now costs about $4,313 at Hyperbolic’s $5.99/hr floor, versus roughly $11,599 at Google Cloud’s $16.11/hr ceiling – a difference of about $7,286 a month. For most buyers, that means provider tier now matters more for raw cost than it did under the narrower specialized-provider-only reading; choosing a specialized neocloud over a hyperscaler is now one of the biggest levers on B200 cost.
B300 Cloud Pricing
B300 availability on the cloud has matured quickly, and pricing is still moving. Per Thunder Compute’s September 4, 2026 pricing table, with rates normalized to an 8-GPU node, five on-demand providers ran B300 capacity: Modal at $7.10 per GPU-hour, the current floor; Runpod at $7.89; Nebius at $7.85; Oracle Cloud at $15.00; and AWS at $17.80, the ceiling among that named-provider set. MorphLLM’s October 2026 provider comparison corroborates the floor, confirming Modal at $7.10 and Runpod at $7.89 per GPU-hour and framing that as roughly a 14% to 16% premium over B200’s specialized-tier rate for the B300’s added memory capacity. That Nebius figure didn’t hold: The Motley Fool reported that Nebius raised its on-demand B300 rate 21% on October 1, 2026, from $7.85 to $9.50 per GPU-hour, pushing Nebius above Spheron in the provider ranking. A separate September 2026 index adds Spheron’s on-demand B300 SXM6 listing at $9.16 per GPU-hour, which now slots in between Runpod and Nebius, and notes that premium managed DGX B300 stacks – fully managed, turnkey deployments rather than raw on-demand instances – run $12.00 to $18.00 per GPU-hour. That puts the full B300 cloud range, current as of October 2026, at $7.10 to $18.00 per GPU-hour once managed offerings are included. Buyers willing to commit to a 48-month reservation can push the effective rate much lower still: one September 2026 pricing guide puts committed, long-term B300 capacity as low as $3.13 per GPU-hour, though that floor applies only to multi-year reserved terms, not on-demand rental. Earlier in 2026, spot and promotional B300 capacity briefly dipped as low as $2.45 per GPU-hour on providers like Spheron Network and Verda – that short-term deep-discount tier no longer shows up in on-demand named-provider readings, and if anything the on-demand floor has kept climbing, as Nebius’s October move shows. Telnyx’s own October 8, 2026 reading reconfirms the on-demand range still holds at $7.10 to $17.80 per GPU-hour, with the 8-GPU DGX B300 system remaining at $300,000–$350,000.
B300 Provider Price Comparison
| Provider | B300 Price (per GPU-hour) | As of |
|---|---|---|
| Modal | $7.10 | Sept. 4, 2026 |
| Hyperstack | $7.40 | September 2026 |
| Runpod | $7.89 | Sept. 4, 2026 |
| Spheron (on-demand SXM6) | $9.16 | September 2026 |
| Nebius | $9.50 | Oct. 1, 2026 |
| Oracle Cloud | $15.00 | Sept. 4, 2026 |
| AWS | $17.80 | Sept. 4, 2026 |
B300 Managed and Reserved Pricing
Beyond the named on-demand providers, B300 capacity is also available through two other pricing structures. Premium managed DGX B300 stacks – fully managed, turnkey deployments that bundle support and orchestration on top of the raw GPU – run $12.00 to $18.00 per GPU-hour as of September 2026, the highest tier in this guide. At the opposite end, a 48-month reservation can bring the effective rate down to as low as $3.13 per GPU-hour, though that requires committing to a long-term term rather than paying on demand.
| Tier | Price (per GPU-hour) | Commitment |
|---|---|---|
| Reserved (48-month term) | as low as $3.13 | Multi-year committed capacity |
| On-demand named providers | $7.10–$17.80 | Pay-as-you-go |
| Premium managed DGX B300 stacks | $12.00–$18.00 | Fully managed, turnkey |
What Drives the Provider Spread
Modal’s $7.10/hr floor leads the field, with Runpod close behind at $7.89/hr, within a nickel of Spheron’s on-demand B300 SXM6 listing at $9.16/hr. Nebius now prices just above that tier at $9.50/hr, after raising its on-demand rate 21% from $7.85 on October 1, 2026 – the clearest sign yet that the B300 on-demand floor is drifting upward, not down. The bigger jump still comes at the hyperscaler and managed tiers: Oracle Cloud’s $15.00/hr and AWS’s $17.80/hr both carry the managed networking, enterprise SLAs, and support overhead typical of major cloud platforms, while fully managed DGX B300 stacks push as high as $18.00/hr for turnkey deployment – roughly double to triple the specialized-provider rates. Teams that don’t need hyperscaler-grade SLAs or fully managed deployment can still capture meaningful savings by choosing Modal or Runpod over Nebius, AWS, Oracle Cloud, or a managed stack, and teams with predictable, long-term demand can go lower still with a 48-month reserved commitment at rates as low as $3.13/hr.
B300 Wholesale and Hyperscaler Pricing (GPUAAS, July 2026)
A July 2026 pricing guide from GPUAAS breaks out B300 pricing from a buyer-type angle rather than a provider angle, and it lines up closely with the named-provider and managed-stack ladder confirmed in September 2026 rather than contradicting it. GPUAAS puts wholesale or committed-volume B300 capacity at $7.00–$12.00 per GPU-hour – the same band where Modal ($7.10), Runpod ($7.89), Spheron ($9.16), and Nebius ($9.50 as of October 2026) all price – and pegs hyperscaler on-demand B300 pricing as starting above $12.00 per GPU-hour, the same territory where Oracle Cloud ($15.00), AWS ($17.80), and premium managed DGX B300 stacks (up to $18.00) land. Both figures sit inside the full $7.10–$18.00 per GPU-hour October 2026 range established above, corroborating it rather than widening it.
| Buyer Type | Price (per GPU-hour) | Source |
|---|---|---|
| Wholesale / committed volume | $7.00–$12.00 | GPUAAS, July 2026 |
| Hyperscaler on-demand | $12.00+ | GPUAAS, July 2026 |
B200 vs B300 vs DGX B300 Cost Comparison
Side-by-Side Pricing Summary
| Product | Cloud Low (per GPU-hour) | Cloud High (per GPU-hour) | System-Level Pricing | Reference Date | Source |
|---|---|---|---|---|---|
| B200 (cloud) | $5.99 (Hyperbolic) | $16.11 (Google Cloud) | – | September 2026 | 11-provider cloud comparison |
| B300 (cloud) | $7.10 (Modal) | $17.80 (AWS) | – | Oct. 8, 2026 | Thunder Compute; Telnyx |
| DGX B200 (8-GPU system) | – | – | $280,000–$320,000 (~$35,000–$40,000/GPU) | April 2026 | iFactory App, Thunder Compute |
| DGX B300 (8-GPU system) | – | – | $300,000–$350,000 (~$37,500–$43,750/GPU) | Q1 2026 – Oct. 13, 2026 | iFactory App; IntuitionLabs; Telnyx |
| DGX B200 (Broadberry configured) | – | – | $562,154.09 (~$70,269/GPU) | Sept. 21, 2026 | Broadberry |
| DGX B300 (Broadberry configured) | – | – | $646,878.46 (~$80,860/GPU) | Sept. 21, 2026 | Broadberry |
| DGX Station (desktop, 1 GPU) | – | – | $80,000–$125,000 | Oct. 13, 2026 | IntuitionLabs |
| GB200 NVL72 (72-GPU rack) | – | – | $2M–$3M (~$27,800–$41,700/GPU) | July 2026 (est.) | market estimates, unconfirmed |
| GB300 NVL72 (72-GPU rack) | – | – | $3M–$4M (~$41,700–$55,600/GPU) | July 2026 (est.) | market estimates, unconfirmed |
Some resellers separately quote fully-configured 8-GPU DGX B300 systems as high as $400,000–$500,000 – a range The Next Platform corroborated on September 21, 2026, pricing eight-way B200 or B300 server nodes at “maybe $400,000 to $500,000.” Broadberry’s own default US configuration listings, also dated September 21, 2026, price DGX B300 higher still at $646,878.46 and DGX B200 at $562,154.09 – the highest and clearest named, dated system-level prices in this guide, both now included in the table above alongside the more commonly cited $300,000–$350,000 (DGX B300) and $280,000–$320,000 (DGX B200) OEM anchors.
GB200 NVL72: The B200-Based Rack Option
Before Blackwell Ultra, NVIDIA’s rack-scale GB200 NVL72 bundled 72 standard B200 GPUs into a single liquid-cooled rack with full NVLink switching between every GPU. NVIDIA has not published an official price for this configuration either. A July 2026 pricing guide estimates a full GB200 NVL72 rack at roughly $2 million–$3 million, or about $27,800–$41,700 per GPU – below the newer GB300 NVL72’s $3 million–$4 million range, since it carries standard B200 GPUs rather than the higher-memory Blackwell Ultra B300. As with the GB300 NVL72 figure, treat this as a directional estimate rather than a confirmed cost until a named source publishes dated pricing.
GB300 NVL72: The Next Rack-Scale Tier
Above the 8-GPU DGX B300, NVIDIA’s rack-scale GB300 NVL72 bundles 72 Blackwell Ultra GPUs into a single liquid-cooled rack with full NVLink switching between every GPU. NVIDIA has not published an official price. Early market estimates place a full rack at roughly $3 million–$4 million, which works out to about $41,700–$55,600 per GPU – a premium over the DGX B300’s $37,500–$43,750 per GPU system-level rate and over the earlier, B200-based GB200 NVL72’s $2 million–$3 million estimate, reflecting the added NVLink fabric, rack-scale engineering, and the jump to Blackwell Ultra B300 GPUs. Treat this range as a directional estimate rather than a confirmed cost until a named source publishes dated pricing.
Single-GPU Purchase Pricing
Not every buyer wants a full DGX system or a cloud subscription. As of October 2026, a single B200 GPU purchased outright costs $45,000–$50,000 street price in 8-GPU volumes – confirmed by IntuitionLabs on October 2, 2026 – while a single B300 GPU runs approximately $53,000, reconfirmed by IntuitionLabs on October 13, 2026 (first reported by Spheron Network, July 5, 2026) – a narrower premium than earlier in the year, tied to the B300’s 288GB HBM3e capacity. NVIDIA has not officially published retail prices for individual B200 units; the $45,000–$50,000 figure reflects OEM quotes and current street pricing rather than a manufacturer list price. That street price has moved: GPU Smith’s September 2026 B200 pricing page had standalone B200 chips at just $30,000 to $40,000 per GPU when purchased inside an 8-GPU HGX cluster, and IntuitionLabs’ September 2026 data-center GPU price table placed the wider range at $30,000–$50,000 per GPU – but supply constraints pushed that street-level band up to meet the $45,000–$50,000 OEM-quote range by October 2, 2026, closing the gap between list and street pricing. A separate pricing guide pegs NVIDIA’s manufacturing cost for the B200 chip at only about $6,400 – a figure that still points to a wide margin between production cost and sale price even after the street-price climb. Note that a fully configured 8-GPU HGX B200 server – the OEM reference platform, not a standalone-chip purchase – still runs $400,000–$500,000 once networking, chassis, and integration are included. These standalone-chip figures exclude networking, cooling infrastructure, and installation. Pricing also varies by reseller: some markets have quoted full DGX B300 systems as high as $400,000–$500,000, and Broadberry’s September 21, 2026 named US configuration listing prices a fully configured DGX B300 higher still, at $646,878.46 – well above the $300,000–$350,000 baseline range, depending on configuration and regional markup.
How to Choose
Among the 11 named providers in the September 2026 cloud comparison, B200 is still the cheaper chip to rent on a same-provider basis: at Modal, Nebius, Runpod, Oracle Cloud, AWS, and Hyperstack – the six providers that list both chips – B300 carries a consistent per-GPU premium of roughly $0.85 to $3.56 an hour over B200. But the overall ranges now overlap. B200’s $5.99–$16.11/hr span (Hyperbolic floor to Google Cloud ceiling) reaches well into B300 territory, since hyperscaler B200 pricing from Oracle Cloud, AWS, and Google Cloud runs higher than specialized-provider B300 pricing from Modal, Runpod, and Nebius. In practice, provider choice now matters as much as chip choice: a B200 rented from AWS ($14.24/hr) costs nearly twice as much as a B300 rented from Modal ($7.10/hr). For buyers evaluating capex, the DGX B300 system price of $300,000–$350,000 remains the relevant baseline anchor, even though a named, fully configured reseller quote from Broadberry (September 21, 2026) runs as high as $646,878.46; for cloud-first teams, choosing the right provider tier – specialized neocloud versus hyperscaler versus managed – now matters at least as much as choosing between B200 and B300. Buyers who only need one or two GPUs rather than a rack-scale system still have a third option: standalone modules at $45,000–$50,000 (B200) or $53,000 (B300) – or, for a preconfigured desktop system, NVIDIA’s DGX Station at $80,000–$125,000 (IntuitionLabs, October 13, 2026). Teams scaling well beyond a single 8-GPU system have two rack-scale tiers to track: the B200-based GB200 NVL72, an early estimate of $2 million–$3 million, and the newer Blackwell Ultra GB300 NVL72, at $3 million–$4 million.
Blackwell Pricing by Budget: What Can You Actually Buy?
Translating the figures above into a straight budget-to-product lookup makes the comparison easier to act on. As of October 2026, here’s the highest tier of Blackwell hardware each budget level can realistically reach, based on the OEM and named-provider anchors used throughout this guide:
| Budget | What You Can Buy | Reference Price |
|---|---|---|
| Under $55,000 | Single B200 or B300 GPU | $45,000–$53,000 |
| $80,000–$125,000 | DGX Station (1 GPU + Grace CPU) | $80,000–$125,000 |
| $280,000–$350,000 | DGX B200 or DGX B300 (8-GPU system) | $280,000–$350,000 |
| $562,154 | Broadberry-configured DGX B200 appliance | $562,154.09 |
| $646,878 | Broadberry-configured DGX B300 appliance | $646,878.46 |
| $2 million+ | GB200 NVL72 rack (72 GPUs, est.) | $2M–$3M |
| $3 million+ | GB300 NVL72 rack (72 GPUs, est.) | $3M–$4M |
The DGX Station’s $80,000–$125,000 price sits between the single-GPU and full-system tiers, giving smaller teams a preconfigured, NVIDIA-built option without committing to a rack-scale budget. At the top end, scaling from the 8-GPU DGX B300 to the 72-GPU GB300 NVL72 – nine times the GPU count – costs roughly ten to eleven times as much, so the per-GPU rate stays broadly comparable even as total spend grows.
Renting vs. Buying: Where’s the Breakeven?
Every price above answers “how much,” but not “which option is actually cheaper for a given workload.” A simple breakeven comparison – per-GPU system cost divided by cloud GPU-hour rate, assuming continuous 24/7 use – makes the tradeoff concrete. This is a directional estimate, not a full total-cost-of-ownership model: it excludes power, networking, floor space, and staffing on the buy side.
| Comparison | Cloud Rate | Per-GPU System Cost | Approx. Breakeven (24/7 use) |
|---|---|---|---|
| B200 vs Hyperbolic floor | $5.99/hr | $35,000–$40,000 | ~8.1–9.3 months |
| B200 vs Google Cloud ceiling | $16.11/hr | $35,000–$40,000 | ~3.0–3.4 months |
| B300 vs Modal floor | $7.10/hr | $37,500–$43,750 | ~7.3–8.6 months |
| B300 vs Spheron on-demand | $9.16/hr | $37,500–$43,750 | ~5.7–6.6 months |
| B300 vs Nebius | $9.50/hr | $37,500–$43,750 | ~5.5–6.4 months |
| B300 vs managed-stack ceiling | $18.00/hr | $37,500–$43,750 | ~2.9–3.4 months |
B200: Rent or Buy?
At Hyperbolic’s $5.99/hr floor (September 2026), a rented B200 doesn’t pay back a $280,000–$320,000 DGX B200 system’s per-GPU cost until roughly 8.1–9.3 months of nonstop use – well past the useful window for most short-term projects, which favors renting. At Google Cloud’s $16.11/hr hyperscaler ceiling, that breakeven shortens dramatically to roughly 3.0–3.4 months. That’s a wide gap: unlike the narrower specialized-provider-only reading, the full 11-provider comparison shows the rent-vs-buy math now depends heavily on provider tier as much as deployment length – renting from a specialized neocloud stays cheaper for the better part of a year, while renting the same chip from a hyperscaler can justify buying within a single quarter.
B300: Rent or Buy?
The same pattern holds for B300, shifted slightly by the DGX B300’s higher per-GPU system cost. At Modal’s $7.10/hr floor, breakeven against the $300,000–$350,000 DGX B300 system lands around 7.3–8.6 months. At Spheron’s $9.16/hr on-demand rate, that narrows to about 5.7–6.6 months; at Nebius’s $9.50/hr – up 21% from $7.85 after its October 1, 2026 rate change – it narrows further still to roughly 5.5–6.4 months; and at the $18.00/hr premium managed-stack ceiling, breakeven arrives in as little as 2.9–3.4 months. Buyers with predictable, multi-year demand have a fourth path worth pricing out: a 48-month reserved commitment at rates as low as $3.13/hr, which trades upfront commitment length for a dramatically lower effective hourly cost. The practical takeaway: reserve buying outright for workloads with predictable, long-running demand that don’t fit a reservation term, use a 48-month reservation if the workload is both long-running and predictable enough to commit to, and use lower-cost on-demand named providers like Modal or Runpod for shorter or bursty rental needs – Nebius’s October price hike makes it less competitive on that front than it was in September.
3-Year Total Cost of Ownership: Buy vs. Rent
The breakeven math above answers how long it takes for renting to cost more than buying, but capex decisions usually get made against a multi-year planning horizon. Extending the same 24/7, per-GPU methodology out to a full 3-year (36-month) hold shows how large the gap becomes once a workload runs well past breakeven. This is the same directional estimate as the breakeven table above – it excludes power, networking, floor space, and staffing on the buy side, and it assumes continuous, uninterrupted use on the rental side.
| Comparison | Cloud Rate | 3-Year Rental Cost (24/7) | Per-GPU System Cost | Approx. Savings by Buying |
|---|---|---|---|---|
| B200 vs Hyperbolic floor | $5.99/hr | ~$155,261 | $35,000–$40,000 | ~$115,261–$120,261 |
| B200 vs Google Cloud ceiling | $16.11/hr | ~$417,571 | $35,000–$40,000 | ~$377,571–$382,571 |
| B300 vs Modal floor | $7.10/hr | ~$184,032 | $37,500–$43,750 | ~$140,282–$146,532 |
| B300 vs managed-stack ceiling | $18.00/hr | ~$466,560 | $37,500–$43,750 | ~$422,810–$429,060 |
Why 3-Year TCO Still Favors Buying
Because breakeven already lands well inside a single year in every scenario above – as fast as 2.9 months against the priciest managed and hyperscaler rates, and no slower than 9.3 months even at the cheapest specialized-provider floor – a full 3-year hold pushes the savings from buying into the hundreds of thousands of dollars per GPU. At Hyperbolic’s $5.99/hr B200 floor, a 36-month rental runs approximately $155,261, versus $35,000–$40,000 to buy the same per-GPU capacity in a DGX B200 system – a savings of roughly $115,261–$120,261 per GPU. The gap widens further against hyperscaler and managed rates: renting B300 at the $18.00/hr managed-stack ceiling for 3 years costs about $466,560, against $37,500–$43,750 to buy into a DGX B300 system – a difference of roughly $422,810–$429,060 per GPU.
The Case for Renting Anyway
Raw TCO math favors buying for any workload that genuinely runs 24/7 for years, but few teams commit to a single GPU generation for a full 3-year term. NVIDIA’s architecture cadence means owned hardware carries real obsolescence risk that a rental doesn’t: a 3-year-old system may be several generations behind by the time it’s fully depreciated, while cloud capacity can be swapped for next-generation GPUs as soon as a provider lists them. Buying makes the most sense for predictable, long-running workloads where the 24/7 utilization assumption actually holds; renting – including via the 48-month reservation discussed above – remains the safer default for workloads with uncertain duration, variable utilization, or a need to stay current with each new Blackwell generation. NVIDIA has already named Blackwell’s successor platform, Vera Rubin, and signaled it will begin arriving in 2026, so anyone weighing a multi-year purchase is effectively betting that Blackwell-class hardware stays competitive through the early part of that transition.
Blackwell Pricing Trends: Then vs. Now
Zooming out from the current snapshot, Blackwell pricing has moved in two different directions depending on where you look. Cloud GPU-hour rates have climbed off their launch-window lows as supply has matured and named providers have replaced promotional pricing with steadier, mainstream rates. System-level DGX pricing, by contrast, has held remarkably steady since each system’s anchor was first established – even as named reseller configurations now price well above that anchor.
Early in 2026, B200 capacity was briefly available for as little as $2.71 per GPU-hour, and B300 spot pricing dipped as low as $2.45 per GPU-hour on providers like Spheron Network and Verda. Neither floor shows up in the named-provider data anymore. As of September 2026, the cheapest documented B200 rate among 11 named providers is Hyperbolic’s $5.99 per GPU-hour, and the cheapest B300 rate is Modal’s $7.10 per GPU-hour – both more than double the early-2026 promotional floors, reflecting a shift from launch-window discounting toward sustained, mainstream cloud pricing. That upward drift hasn’t stopped: on October 1, 2026, Nebius raised its B300 on-demand rate 21%, from $7.85 to $9.50 per GPU-hour, per The Motley Fool – the clearest sign yet that B300 cloud pricing is still climbing well past its launch-window floor, even nine months after the chip went on sale.
DGX system pricing tells a steadier story at the anchor level, even as reseller quotes climb. The DGX B300 launched in January 2026 with an 8-GPU system anchored at $300,000–$350,000, and that anchor held from Q1 2026 all the way through October 2026 – reconfirmed as recently as October 8, 2026 by Telnyx and October 13, 2026 by IntuitionLabs, nine-plus months after launch – per IntuitionLabs, Spheron, and NVIDIA’s own DGX B300 product page. The DGX B200’s $280,000–$320,000 anchor, established in April 2026, has likewise held steady through October 2026. But named reseller pricing has moved well past those anchors: Broadberry’s default US configuration listings, dated September 21, 2026 and reported by GPUPerHour, price a fully configured DGX B200 appliance at $562,154.09 and a fully configured DGX B300 appliance at $646,878.46 – the first named, dated system-level prices for both machines to appear in this dataset at that fully configured tier, and well above the underlying OEM anchors. Standalone chip pricing hasn’t been nearly as steady: B200’s street-level price, quoted as low as $30,000–$40,000 per GPU in GPU Smith’s September 2026 figures, climbed to $45,000–$50,000 by October 2, 2026, per IntuitionLabs, as supply constraints took hold – the one figure in this guide that moved meaningfully higher at the component level even while system-level anchors held flat.
| Metric | Early 2026 | October 2026 |
|---|---|---|
| B200 cloud floor | ~$2.71/hr (promotional) | $5.99/hr (Hyperbolic) |
| B300 cloud floor | ~$2.45/hr (spot) | $7.10/hr (Modal) |
| B200 standalone GPU (street price) | $30,000–$40,000/GPU (Sept. 2026) | $45,000–$50,000/GPU (IntuitionLabs, Oct. 2, 2026) |
| B300 standalone GPU (street price) | $53,000/GPU (Jul. 2026, Spheron) | $53,000/GPU (IntuitionLabs, Oct. 13, 2026) |
| DGX B300 system (8 GPUs) | $300,000–$350,000 (Q1 2026 anchor) | $300,000–$350,000 (Telnyx/IntuitionLabs, Oct. 8–13, 2026) |
| DGX Station (1 GPU) | $80,000–$125,000 (Sept. 2026) | $80,000–$125,000 (IntuitionLabs, Oct. 13, 2026) |
| B300 cloud on-demand range | – | $7.10–$17.80/hr (Telnyx, Oct. 8, 2026) |
| DGX B200 system (8 GPUs) | $280,000–$320,000 (April 2026 anchor) | $280,000–$320,000 (holding steady) |
| Named reseller DGX B200 listing | Not published | $562,154.09 (Broadberry, Sept. 21, 2026) |
| Named reseller DGX B300 listing | Not published | $646,878.46 (Broadberry, Sept. 21, 2026) |
| Nebius B300 on-demand rate | Not applicable | $9.50/hr (up from $7.85, Oct. 1, 2026) |
For cloud buyers, the takeaway is that waiting for rates to fall back to early-2026 promotional levels isn’t a realistic strategy – that pricing tier has left the market, and Nebius’s October increase suggests the drift is still upward. For system-level buyers, the DGX B200 and DGX B300 anchors have now held for the better part of a year, making them a more reliable budgeting baseline than cloud rates – though buyers should expect a fully configured, named-reseller quote to land well above that baseline, as Broadberry’s September 21, 2026 listings show. Buyers shopping for standalone chips face the same upward drift: B200’s street price has risen to $45,000–$50,000 under October 2026 supply constraints – a climb IntuitionLabs confirmed on October 2, 2026 – erasing the discount that was available as recently as September.
Frequently Asked Questions
How much does a DGX B300 system cost?
As of Q1 2026, an 8-GPU DGX B300 system is anchored at $300,000–$350,000, which implies roughly $37,500–$43,750 per GPU at the system level. That anchor held from Q1 2026 through October 2026, most recently reconfirmed by Telnyx on October 8, 2026 and by IntuitionLabs on October 13, 2026, and corroborated separately by Spheron’s figures and by NVIDIA’s own DGX B300 product page, now live on nvidia.com, even though NVIDIA itself still doesn’t publish a system price. A named reseller listing shows the fully configured price running well above that anchor: Broadberry’s default US configuration, dated September 21, 2026, prices the DGX B300 appliance at $646,878.46. The DGX B300 began shipping in January 2026.
How much does the DGX Station cost?
IntuitionLabs documents NVIDIA’s DGX Station – a desktop tower pairing one Blackwell Ultra B300 GPU with a Grace CPU – at $80,000–$125,000, reconfirmed on October 13, 2026, well below the 8-GPU DGX B300’s $300,000–$350,000 system price. It’s the lowest-cost way to get a preconfigured Blackwell Ultra system directly from NVIDIA’s DGX line.
How much does a DGX B200 system cost compared to DGX B300?
An 8-GPU DGX B200 system is priced at $280,000–$320,000 as of April 2026 (iFactory App, Thunder Compute), while the DGX B300 commands $300,000–$350,000 as of Q1 2026, an anchor that held through October 2026. Other 2026 pricing guides place DGX B200 higher still, around $370,000–$500,000, so treat the exact DGX B200 figure as source-dependent rather than fixed – GPU Smith’s September 2026 B200 pricing page lands in that same band, pricing a fully configured 8-GPU HGX B200 server at $400,000–$500,000 even though it priced standalone B200 chips inside that cluster at just $30,000–$40,000 per GPU that month – a figure October 2026 supply constraints have since pushed to $45,000–$50,000 per GPU. Some resellers have quoted full DGX B300 configurations as high as $400,000–$500,000 depending on market and configuration – a figure The Next Platform corroborated on September 21, 2026, noting that eight-way server nodes built on either B200 or B300 run “maybe $400,000 to $500,000.” IntuitionLabs’ September 2026 data-center GPU price table independently lands in the same band, citing a DGX-style 8-GPU B200 system at $400,000–$500,000. The clearest named, dated system prices come from reseller Broadberry: its default US configuration listings, dated September 21, 2026, price the DGX B200 appliance at $562,154.09 (about $70,269 per GPU) and the DGX B300 appliance at $646,878.46 (about $80,860 per GPU) – confirming B200 still runs materially cheaper than B300 even at this fully configured, named-reseller level.
What’s the difference between the NVIDIA B200 and B300 GPUs?
The B300 is NVIDIA’s Blackwell Ultra variant of the standard B200. Both share the same core Blackwell architecture, but the B300 steps up to 288GB of HBM3e memory versus the B200’s 192GB – roughly 50% more – and its DGX form requires mandatory direct liquid cooling. NVIDIA positions the B300 for AI reasoning workloads with long inference chains and large context windows, which is also why it carries a price premium over the B200 at every tier as of October 2026: standalone chips run $53,000 versus $45,000–$50,000, cloud rentals run roughly $0.85 to $3.56 more per GPU-hour on a same-provider basis, and full 8-GPU DGX systems anchor at $300,000–$350,000 versus $280,000–$320,000.
How much does the GB200 NVL72 cost?
NVIDIA has not published an official price for the GB200 NVL72, the B200-based predecessor to the newer Blackwell Ultra GB300 NVL72. A July 2026 pricing guide estimates a full rack of 72 standard B200 GPUs at roughly $2 million–$3 million, or about $27,800–$41,700 per GPU – below the GB300 NVL72’s $3 million–$4 million estimate. Treat this as a directional estimate until a named source confirms dated pricing.
How much does the GB300 NVL72 cost?
NVIDIA has not published an official price for the rack-scale GB300 NVL72, which bundles 72 Blackwell Ultra GPUs into a single liquid-cooled rack. Early market estimates put a full rack at roughly $3 million–$4 million, or about $41,700–$55,600 per GPU – a premium over the DGX B300’s per-GPU system rate and over the earlier B200-based GB200 NVL72 rack, estimated at $2 million–$3 million. Treat this as a directional estimate until a named source confirms dated pricing.
How much does a single B200 or B300 GPU cost to buy?
As of October 2026, a standalone B200 GPU costs $45,000–$50,000 street price in 8-GPU volumes – confirmed by IntuitionLabs on October 2, 2026 – while a standalone B300 GPU runs approximately $53,000, reconfirmed by IntuitionLabs on October 13, 2026 (first reported by Spheron Network, July 5, 2026), reflecting its larger 288GB HBM3e memory. NVIDIA has not officially published either figure as a list price. B200’s street price has climbed to meet that range: GPU Smith’s September 2026 B200 pricing page had it at just $30,000–$40,000 per GPU inside an 8-GPU HGX cluster, and IntuitionLabs’ September 2026 pricing guide placed the overall range at $30,000–$50,000 per GPU, but October 2026 supply constraints have since pushed street pricing up to $45,000–$50,000, the same band as OEM quotes. A separate pricing guide estimates NVIDIA’s manufacturing cost at about $6,400 per GPU – a figure that hints at wide margins even after the street-price climb. These prices exclude networking, cooling, and installation.
What is the cheapest B200 cloud rate as of September 2026?
As of September 2026, the single lowest documented B200 GPU-hour rate is $5.29 per GPU-hour at Lambda Labs, per GridStackHub’s September 2026 index of 6+ specialized providers, which puts the narrower market band at $5.29–$7.05 per GPU-hour. Among the fuller 11-named-provider comparison, Hyperbolic holds the floor at $5.99 per GPU-hour, with Hyperstack ($6.00) and Modal ($6.25) close behind; that comparison’s ceiling reaches $16.11 per GPU-hour at Google Cloud, for a $5.99–$16.11 range once hyperscaler pricing from Oracle Cloud, AWS, and Google Cloud is included. That’s still a higher range than earlier in 2026, when some providers offered B200 capacity as low as $2.71 per GPU-hour during launch-window promotions. MorphLLM’s October 2026 comparison confirms that specialized tier has held steady into the new month, with Modal at $6.25 and Lambda and Runpod both at $6.79 per GPU-hour.
How cheap can B300 cloud pricing get?
It depends on the commitment level and when you’re checking. Among on-demand named providers, the lowest documented B300 rate is Modal at $7.10 per GPU-hour, per Thunder Compute’s September 4, 2026 pricing table. Runpod follows at $7.89, Spheron’s on-demand SXM6 listing at $9.16, and Nebius at $9.50 – up 21% from $7.85 after an October 1, 2026 rate change reported by The Motley Fool – then Oracle Cloud at $15.00, and AWS at $17.80, with premium managed DGX B300 stacks running as high as $18.00 at the very top. That puts the full B300 cloud range, current as of October 2026, at $7.10 to $18.00 per GPU-hour for on-demand and managed capacity. Buyers can go lower by committing to a 48-month reservation, which one September 2026 pricing guide puts as low as $3.13 per GPU-hour – but that rate applies only to multi-year committed capacity, not on-demand rental. GPUAAS’s July 2026 wholesale ($7.00–$12.00/hr) and hyperscaler ($12.00+/hr) bands corroborate this range rather than widening it. MorphLLM’s October 2026 comparison corroborates the floor too, confirming Modal and Runpod at $7.10 and $7.89 and framing that as a 14%–16% premium over B200 for the B300’s added memory. Telnyx’s October 8, 2026 read reconfirms the on-demand named-provider range at $7.10 to $17.80 per GPU-hour, the freshest named-provider check on B300 cloud pricing in this guide.
Did Nebius raise its B300 cloud pricing in October 2026?
Yes. The Motley Fool reported that Nebius raised its on-demand B300 GPU-hour rate 21% on October 1, 2026, from $7.85 to $9.50 per GPU-hour. That moves Nebius from pricing near the floor to above Runpod ($7.89) and Spheron’s on-demand SXM6 listing ($9.16), though it still undercuts Oracle Cloud ($15.00), AWS ($17.80), and premium managed DGX B300 stacks (up to $18.00). The on-demand floor remains Modal at $7.10 per GPU-hour, unaffected by Nebius’s increase.
Is B300 always more expensive than B200 in the cloud?
It depends on how you compare. On a same-provider basis, yes: among the six providers that list both chips – Modal, Nebius, Runpod, Oracle Cloud, AWS, and Hyperstack – B300 costs more than B200 every time, by roughly $0.85 to $3.56 per GPU-hour as of October 2026, with Nebius’s premium widening to about $2.35 after its October 1 rate increase. But the two overall ranges still overlap: B200’s $5.99–$16.11/hr span (Hyperbolic to Google Cloud) reaches above B300’s $7.10/hr floor (Modal), since hyperscaler B200 pricing has climbed higher than specialized-provider B300 pricing. A B200 rented from AWS ($14.24/hr) or Google Cloud ($16.11/hr) still costs more than a B300 rented from Modal ($7.10/hr) or Runpod ($7.89/hr). That’s a shift from a narrower September 2026 reading that had B300 sitting entirely above B200 with no overlap – the picture changed once hyperscaler pricing entered the fuller comparison.
Is it cheaper to rent or buy a Blackwell B200 or B300 GPU?
It depends on how long the workload runs and which provider tier you use. Against the lowest September 2026 cloud rates – Hyperbolic’s $5.99/hr for B200, Modal’s $7.10/hr for B300 – renting stays cheaper than buying a DGX system for roughly 7.3–9.3 months of continuous use. Against the highest rates – Google Cloud’s $16.11/hr for B200, or premium managed DGX B300 stacks at up to $18.00/hr – buying pays for itself in as little as 2.9–3.4 months. Workloads predictable and long-running enough to commit to a 48-month reservation can push B300’s effective rate as low as $3.13/hr, changing the math further in favor of renting. As a rule of thumb, short or bursty workloads favor renting from lower-cost specialized providers, hyperscaler or managed rentals make the most sense for short bursts where buying isn’t practical, and predictable, long-running workloads favor either buying a DGX system outright or locking in a multi-year reservation.
When did the DGX B300 start shipping?
The DGX B300 began shipping in January 2026. It requires mandatory direct liquid cooling and pairs each GPU with 288GB of HBM3e memory, about 50% more than the B200’s 192GB capacity.
What workloads is the DGX B300 designed for?
NVIDIA describes the DGX B300 as the Blackwell Ultra system for AI reasoning – workloads characterized by long inference chains and large context windows where the integrated DGX reference architecture provides the most value.
Has Blackwell GPU pricing dropped since launch?
No – if anything, named-provider cloud rates have kept rising past the deep promotional floors seen earlier in 2026. B200 spot pricing as low as $2.71/hr and B300 spot pricing as low as $2.45/hr have given way to steadier named-provider floors of $5.99/hr (Hyperbolic) for B200 and $7.10/hr (Modal) for B300, and the climb hasn’t stopped: Nebius raised its B300 on-demand rate 21% on October 1, 2026, from $7.85 to $9.50 per GPU-hour, as launch-window discounts keep fading and mainstream cloud pricing takes over. System-level pricing has been more stable at the OEM-anchor level: the DGX B300’s $300,000–$350,000 anchor, set in Q1 2026, and the DGX B200’s $280,000–$320,000 anchor, set in April 2026, have both held steady through October 2026 – even as named reseller Broadberry prices a fully configured system at $562,154.09 (DGX B200) and $646,878.46 (DGX B300) as of September 21, 2026, confirming it’s reseller quotes, not the baseline anchor, that are climbing. Standalone chip pricing has climbed too: B200’s street price, as low as $30,000–$40,000 per GPU in September 2026, rose to $45,000–$50,000 by October 2, 2026 under supply constraints, per IntuitionLabs – so across cloud, reseller, and standalone-chip pricing alike, the trend since launch has been upward, not downward.
Is it cheaper to buy or rent a Blackwell GPU over 3 years?
For workloads that genuinely run 24/7, yes – buying wins by a wide margin. A 36-month rental of a B200 at Hyperbolic’s $5.99/hr floor costs about $155,261, versus $35,000–$40,000 to buy the same per-GPU capacity, a savings of roughly $115,261–$120,261 per GPU. For B300, renting 3 years at the $18.00/hr managed-stack ceiling costs about $466,560, against $37,500–$43,750 to buy into a DGX B300 system – a savings of roughly $422,810–$429,060 per GPU. The tradeoff is obsolescence risk: a 3-year hold locks in one Blackwell generation, while renting lets teams move to newer hardware as soon as it ships.
