TheAINVIDIA B300 and B200 pricing.
Whole 8-GPU bare-metal nodes in three lots. B300 in Kuala Lumpur, available now: one 8-node block at $9.00/GPU·hr for 3 months, full term paid up front. B300 in Iowa: $4.70/GPU·hr on 36 months, $4.50 on 60, 30% at signing, ready within 12 weeks after the deposit, from a single node. B200 in Los Angeles available from November 2026: $8.40/GPU·hr month-to-month. No egress fees. Written quote within 24 hours; billing starts only after acceptance benchmarks pass.
Written answer in 24h · no payment until you accept · or email cloud@theai.com
Rate card
| Lot and term | $ / GPU·hr | $ / node·hr (8 GPUs) | ≈ $ / node·month (730 h) | Term total and payment |
|---|---|---|---|---|
| B300 · Kuala Lumpur · 3 monthsavailable now · one 8-node block (64 GPUs) | $9.00 | $72.00 | $52,560 | $1,261,440whole block, 3 months · full term paid up front |
| B300 · Iowa · 36 monthsready within 12 weeks after the deposit | $4.70 | $37.60 | $27,448 | $988,128per node, 36 months, take-or-pay · 30% = $296,438 |
| B300 · Iowa · 60 monthsready within 12 weeks after the deposit · −4% vs 36 months, our lowest rate | $4.50 | $36.00 | $26,280 | $1,576,800per node, 60 months, take-or-pay · 30% = $473,040 |
| B200 · Los Angeles · month-to-monthavailable from November 2026 · 17 nodes, up to 8 more on request | $8.40 | $67.20 | $49,056 | rolling monthlybilled monthly |
Prices in USD per GPU-hour; a node is 8 GPUs (B300 SXM 288GB or B200 180GB). The 36- and 60-month terms are take-or-pay: 30% on acceptance of the quote (wire transfer), the remainder invoiced per the capacity agreement, monthly is standard. The Kuala Lumpur block is paid for the full term up front. Machine-readable: GET /api/v1/pricing.
B300 in Kuala Lumpur, Malaysia — available now
One block of 8 nodes (64 GPUs), offered as a whole, on a 3-month term at $9.00/GPU·hr — $1,261,440 for the block, full term paid up front.
- GPUs: 8× NVIDIA B300 SXM · 288 GB HBM3e per GPU · NVLink 5 / NVSwitch
- CPU: 2× Intel Xeon 6767P
- Memory: 4 TB RAM
- Storage: 4× 7.68 TB NVMe local
- Network: 8× NVIDIA ConnectX-8 800 Gb/s per node (one per GPU)
- Access: bare metal · public IP per server · no network restrictions
- SLA: 99.975% monthly uptime · service credits 10% / 25% / 100% below 99.975% / 99% / 95%
- Location: Kuala Lumpur, Malaysia
Request the Kuala Lumpur block →
B300 in Iowa, United States — 36 or 60 months
Dedicated HGX B300 nodes in a Tier III data center, up to 32 nodes, ready for service within 12 weeks after the deposit is paid. $4.70/GPU·hr on 36 months, $4.50 on 60; take-or-pay with 30% at signing; from a single node.
- Server: HGX B300 NVL8, 10U, air-cooled
- GPUs: 8× NVIDIA B300 · 288 GB HBM3e per GPU (2.3 TB per node)
- GPU interconnect: NVLink 5 via NVSwitch, 1.8 TB/s per GPU
- CPU: 2× Intel Xeon 6, 64 cores each (128 cores / 256 threads); model (6767P, 6776P or 6768P) confirmed with the order
- Memory: 2 TB DDR5
- Storage: 8× 3.84 TB NVMe for data (~30.7 TB) + 2× 960 GB M.2 NVMe for the OS
- Compute network: 8× NVIDIA ConnectX-8 800G, one per GPU — 6.4 Tb/s per node; rail-optimised, non-blocking within the pod
- Storage and external network: 2× NVIDIA BlueField-3 200G DPU; 400G storage network and a 100G backup external uplink
- Location: Tier III data center, Iowa, United States
Reserved B300 pricing vs published reserved rates
| Provider | On-demand | 1 year | 24 months | 36 months | 60 months |
|---|---|---|---|---|---|
| TheAI Cloud Iowa · ready within 12 weeks after the deposit | — | — | — | $4.7030% at signing | $4.5030% at signing |
| Verdapublished reserved tiers · source | $8.285 | $7.62 | $6.21full prepayment | — | — |
| DigitalOceanGPU Droplets, HGX B300 · source | — | $7.94 | — | — | — |
Why 36 or 60 months in Iowa. We pass the commitment through at a flat rate for the whole period: $4.70/GPU·hr on 36 months, $4.50 on 60 — the longer term is 4% lower because the later years fall on a softer market, and the customer shares that risk with us. For a shorter commitment, see the Kuala Lumpur block above.
Published price lists, verified Sep 25, 2026; B300 per GPU-hour; provider names link to the source. Hyperstack and Nebius publish on-demand only. CoreWeave and Lambda quote reserved via sales. AWS Capacity Blocks list at $14.04/GPU·hr. Verda: published 1-year/2-year reserved tiers (full prepayment); on-demand/spot not compared.
Five-year reserved at $4.50 is our lowest published rate, with 30% at signing. Market context: what ~50 providers quoted us for B300.
B200 in Los Angeles, United States — available from November 2026
Whole 8× B200 180GB nodes, month-to-month at $8.40/GPU·hr — $67.20/node·hr, about $49,056 per node per month, billed monthly. 17 nodes (136 GPUs), up to 8 more on request. Networking is 2× 200 Gb/s BlueField-3 per node (400 Gb/s): suited to inference and single-node workloads, not multi-node training.
- GPUs: 8× NVIDIA B200 · 180 GB HBM3e per GPU (~1.43 TB per node)
- CPU: 2× AMD EPYC 9575F (224 vCPU)
- Memory: 3 TB RAM
- Storage: 960 GB OS SSD + 2× 3.5 TB NVMe
- Network: 2× 200 Gb/s NVIDIA BlueField-3 (400 Gb/s per node) — for inference and single-node workloads, not multi-node training
- Confidential computing: AMD SEV supported
- Location: Los Angeles, United States
- Allocation: whole node · month-to-month
How a reservation works
- Request a configuration — lot, nodes, term and target start. Free, no obligation.
- Written quote within 24 hours: site, delivery window, exact spec, rate.
- You accept → KYC, then MSA and order form → payment locks nodes and rate: 30% on 36- and 60-month terms, the full term up front for the Kuala Lumpur block, monthly invoices for B200.
- Delivery → acceptance benchmarks (NCCL over NVLink and the cluster fabric, DCGM diag, burn-in) → billing starts.
Pricing FAQ
Which terms do you offer?
Each lot has its own term. B300 in Kuala Lumpur: 3 months for the whole 8-node block, paid up front. B300 in Iowa: 36 or 60 months from a single node, take-or-pay with 30% at signing; the 60-month term carries the lower rate. B200 in Los Angeles: month-to-month. The rates are on Pricing; the written quote confirms the delivery window and the exact node configuration.
Is the reservation paid or binding?
Requesting a configuration is free and creates no obligation. Your quote — site, delivery window, spec and rate — comes within 24 hours. Payment then locks your nodes and rate: 30% at signing on 36- and 60-month terms (take-or-pay from that point), the full term up front for the Kuala Lumpur block.
How do I lock my slot and rate?
Request a configuration → written quote within 24 hours → you accept → KYC, then MSA and order form → payment invoice → nodes and rate locked. On 36- and 60-month terms the invoice is a 30% prepayment, take-or-pay, with the remaining 70% invoiced monthly per the agreement; the Kuala Lumpur block is invoiced for the full term up front; B200 month-to-month is invoiced monthly.
Can I cancel a locked reservation?
A locked reservation is a rental contract, not a hold: the 30% prepayment on 36- and 60-month terms is credited to your term and is not refunded if you cancel. For the other lots, cancellation terms are set in the agreement. Until you accept a quote, nothing is owed.
What if the delivery window slips?
If the delivery window in your agreement is missed, you choose: a rate reduction for the delay, or a full refund of the prepayment if the delay is material. Billing never starts before acceptance, and your rate does not move.
What if the cluster doesn't pass acceptance?
Acceptance criteria are written into your agreement before delivery. If the cluster doesn't pass them, the prepayment is refunded in full. The criteria are spec-based until first delivery: nccl-tests all_reduce inside the node over NVLink and across nodes over the cluster fabric, against the bandwidth thresholds named in the agreement; DCGM diagnostics at level 3; and a burn-in run.
How does billing work?
36- and 60-month terms: a 30% prepayment invoice after you accept the quote, then monthly invoices per the capacity agreement. The Kuala Lumpur block: one invoice for the full 3-month term up front. B200 month-to-month: monthly invoices. Billing starts only after acceptance benchmarks pass. No egress fees.
Ready to reserve?
Written quote in 24h — no payment until you accept.
or email cloud@theai.com · full terms: Terms · Refund policy