Hyperstack (a NexGen Cloud brand) is a UK-headquartered GPU cloud with capacity in Canada, Norway and Europe. It has a clean per-GPU rate card, reservation discounts, free egress, per-minute billing and a reputation for responsive support. Its lineup runs from RTX A4000 to HGX B200 and B300, and it lists the RTX Pro 6000 Server Edition, which makes it one of the few direct comparators for packet.ai's most popular SKU.
packet.ai serves the US and EU and reaches lower prices on the same silicon through its scheduler rather than through long reservations. This comparison is card-for-card, with the reservation discounts Hyperstack offers included so the gap is fair.
Platform overview: packet.ai vs Hyperstack
What Hyperstack offers
Hyperstack's published on-demand rates: RTX Pro 6000 SE (96GB) $1.85/hr, H200 SXM $3.99/hr, H100 SXM $3.20/hr, H100 PCIe $2.50/hr (NVLink variant $2.60/hr), A100 SXM 80GB $1.60/hr, L40 (48GB) $1.00/hr, RTX A6000 $0.50/hr, RTX A4000 $0.15/hr. HGX B200 is listed around $6.00/hr. Reservations start at $1.30/hr for the RTX Pro 6000, $1.75/hr for H100 PCIe and $2.79/hr for H200.
- Per-minute billing, free egress, hibernation to pause instances without losing state
- Reservation pricing on 1–3 year terms brings most cards down 20–30%
- Regions in Canada, Norway and mainland Europe; no US on-demand capacity listed
- Also runs an AI Studio for per-token LLM inference and fine-tuning
What packet.ai delivers
Hyperstack's on-demand rates are already sharp, and its reservation tiers push several cards lower still. The gap between the two platforms is narrower here than against most other comparisons on this list. What separates them is geography: Hyperstack has no listed US on-demand capacity, running out of Canada, Norway and mainland Europe. packet.ai runs both US and EU on-demand.
On the two cards that overlap most directly, the RTX Pro 6000 and B200, the comparison depends on which Hyperstack tier you use. Against Hyperstack's on-demand RTX Pro 6000 SE at $1.85/hr, packet.ai's RTX 6000 Pro Dynamic at $0.66/hr is 64% lower. Against Hyperstack's reserved rate of $1.30/hr, the gap is still 49%. On B200, Hyperstack's roughly $6.00/hr on-demand compares to packet.ai's $3.75/hr Dynamic, a 38% gap.
- Dynamic: scheduler-enforced multi-tenant, US and EU on-demand. RTX 6000 Pro at $0.66/hr, B200 at $3.75/hr. No reservation required to get the lower rate.
- Dedicated: a whole card, single-tenant, 99.99% SLA. RTX 4090 at $0.39/hr, L40S at $0.92/hr, A100 80GB at $1.43/hr, B200 at $6.99/hr.
What Hyperstack has that packet.ai does not: a broader entry-level card lineup (RTX A6000, RTX A4000), an AI Studio for per-token inference already live rather than on a waitlist, and reservation pricing that closes much of the gap for teams that can commit to a term. If EU-only capacity works for your workload and you can reserve, Hyperstack's committed rates are genuinely competitive. If you need US on-demand at the lower rate without committing to a term, that is where packet.ai sits.
Billing is hourly with monthly commits up to 20% off. There are no platform fees and no ingress charges; egress is $0.04/GB. Up to 2 TB of local NVMe per node is included. Dynamic instances are SSH-ready in under five minutes, Dedicated in five to ten. Capacity is live in the US (California, Virginia, Texas, Oregon) and Europe (Frankfurt, Amsterdam, Paris, London, Dublin). H100 SXM, H200 and RTX 5090 are on the notify list.
Against Hyperstack, packet.ai's on-demand RTX Pro 6000 is already below Hyperstack's best reservation price, and its Dynamic B200 is around 40% under Hyperstack's on-demand B200. On the A100 80GB the gap is narrower but still in packet.ai's favour, and packet.ai has US capacity where Hyperstack does not.
packet.ai vs Hyperstack at a glance
Published starting rates in USD per GPU-hour, on-demand unless noted. Percentages compare the first price in each cell.
| Category | packet.aiYOU | Hyperstack |
|---|---|---|
| RTX 6000 Pro (96GB) | $0.66/hr Dynamic−64% | $1.85/hr on-demand · from $1.30/hr reserved |
| A100 80GB | $1.43/hr Dedicated−11% | $1.60/hr SXM · from $1.36/hr reserved |
| B200 (180–192GB) | $3.75/hr Dynamic · $6.99/hr Dedicated | ~$6.00/hr HGX B200 |
| L40S / L40 (48GB) | L40S $0.92/hr Dedicated | L40 $1.00/hr |
| RTX 4090 (24GB) | $0.39/hr Dedicated | Not offered (RTX A6000 $0.50/hr, A4000 $0.15/hr instead) |
| H100 / H200 | Launching soon (notify list) | H100 PCIe $2.50/hr · H100 SXM $3.20/hr · H200 $3.99/hr |
| Billing | Hourly; monthly commits up to 20% off | Per minute; 1–3 year reservations |
| Data transfer | No ingress fee; egress $0.04/GB | Free egress |
| Regions | US (CA, VA, TX, OR) + EU (Frankfurt, Amsterdam, Paris, London, Dublin) | Canada, Norway, mainland Europe |
| Tenancy | Dynamic (shared, isolated) or Dedicated (99.99% SLA) | Dedicated VMs; hibernation available |
Detailed comparison
RTX Pro 6000: the headline gap
This is the cleanest like-for-like on any comparison page: same 96GB GDDR7 Blackwell card, two published rates. Hyperstack asks $1.85/hr on-demand and $1.30/hr on a multi-year reservation. packet.ai asks $0.66/hr on Dynamic with no commitment, which is 64% below Hyperstack's on-demand price and 49% below its reserved floor.
Hyperstack gets to its price with reservations; packet.ai gets to its price by scheduling multiple isolated workloads onto the card and passing the utilisation back. Both are legitimate. Only one of them lets you leave at the end of the hour.
A100, L40 and B200
The A100 80GB is closer: $1.43/hr Dedicated on packet.ai against $1.60/hr on Hyperstack, with Hyperstack's reservation floor at $1.36/hr just under packet.ai's on-demand rate. If you will genuinely run an A100 for three years, Hyperstack's reserved price wins by seven cents. If you might not, packet.ai wins.
On the 48GB Ada class the two are within a dime (L40S $0.92 versus L40 $1.00; note the L40S is the faster part). On the B200 packet.ai's Dynamic rate of $3.75/hr is roughly 37% under Hyperstack's ~$6.00/hr, and Dedicated at $6.99/hr is slightly above it.
Regions and residency
Hyperstack's strength is European and Canadian data residency on renewable-powered sites; there is no US on-demand region on its rate card. packet.ai runs both US and EU sites, so a team serving North American users has a local option, and a team with GDPR constraints can pin to Frankfurt or Dublin. If your requirement is specifically Norway or Canada, Hyperstack is the one that has it.
Billing granularity and hibernation
Hyperstack bills per minute and offers hibernation so a stopped VM keeps its state at storage-only cost. packet.ai bills hourly and expects you to spin Dynamic instances down between sessions. For interactive development with frequent pauses Hyperstack's model is more forgiving; for anything running continuously the difference disappears.
Who should choose which
Choose packet.ai if
- You want the RTX 6000 Pro at the lowest published rate anywhere, no reservation
- You need US capacity as well as EU
- You want Dynamic pricing for dev and Dedicated with a 99.99% SLA for production
- You are moving to B200 and want the lowest non-spot hour
Choose Hyperstack if
- You need Canadian or Norwegian data residency
- You are certain you will run the same card for 1–3 years and want a reservation
- You want per-minute billing and hibernation for stop-start development
- You need H100 or H200 today
Conclusion
Hyperstack is a well-run European neocloud with fair prices and good support. packet.ai is simply cheaper on the cards they share, dramatically so on the RTX Pro 6000, and it has US capacity Hyperstack lacks. Choose Hyperstack for Canada/Nordic residency or Hopper today; choose packet.ai for Blackwell at on-demand prices below Hyperstack's reservations.
Same silicon. Smarter economics.
Every rate on this page is a published starting price you can deploy against today. Dynamic launches in under five minutes; no credit card to start.
Additional resources
- packet.ai GPU pricing: the full Dynamic, Dedicated and Clusters rate card.
- packet.ai documentation: API, CLI and platform guides.
- How intelligent scheduling works: why Dynamic keeps peak performance at a lower price.
- Clusters and wholesale quotes: multi-node InfiniBand capacity at ~30% below retail.
- Rent GPUs on packet.ai: specs, pricing and workload fit for every card.