Hyperstack ranks #3 of 10 in our gpu cloud providers testing. $2.50 on-demand with free egress and per-minute billing — and a policy that deletes your machines if the balance runs dry..
86/100
the best on-demand rate here from a conventional provider, with genuinely free bandwidth — spoiled slightly by unclear sku labelling and a hard funding rule.
why people look for an alternative
−resources terminated a week after the balance runs out
−pcie vs sxm not distinguished at the headline rate
−the $2.00 tier isn't clearly labelled as interruptible
−no idle-billing policy published
stay with Hyperstack if $2.50 on-demand — best conventional-provider rate here is the thing you care about most — nothing below beats it on that.
#1 in gpu cloud providers · cheapest verified h100 of the ten, billed by the second, with free egress and one storage trap worth knowing.
91/100
verdictthe best combination of price, billing granularity and free bandwidth in the category — just don't shut pods down to save money, because that costs more.
RunPod vs Hyperstack
Hyperstack
RunPod
price
$2.50/gpu/hr
$2.69/gpu/hr
free tier
no
no
h100 $/gpu/hr
$2.50
$2.69-2.99
spot
$2.00 (tier unclear)
$1.99 (pcie community)
min unit
not stated
1 gpu
egress
free
free
sla
linked, terms not published
none published
switch formost workloads — it is the cheapest honest rate here and the billing granularity actually rewards short jobs.
pros
+cheapest verified on-demand and spot h100 rates here
+per-second billing on compute and pod storage
+no ingress or egress charges of any kind
+b200 available on-demand rather than reservation-only
cons
−stopped volumes cost double the running rate
−no published sla for standard pods
−network volumes bill hourly while compute bills per second
#4 in gpu cloud providers · the only provider here that publishes an uptime number — and its own blog quotes a different price than its pricing page.
83/100
verdictalone in this category in promising 99.9% uptime in writing, with storage bundled and no lock-in — held back by a decentralised model and a price its own pages disagree on.
Spheron vs Hyperstack
Hyperstack
Spheron
price
$2.50/gpu/hr
$2.54/gpu/hr
free tier
no
no
h100 $/gpu/hr
$2.50
$2.54
spot
$2.00 (tier unclear)
$1.67
min unit
not stated
1 gpu
egress
free
not itemised
sla
linked, terms not published
99.9%
switch forshort bursts where deployment speed and a stated uptime figure matter more than a fixed vendor relationship.
pros
+the only published uptime sla in this ranking
+storage bundled into the hourly rate
+per-minute billing with no lock-in or minimum
+$1.67 spot is among the cheapest here
cons
−pricing page and vendor blog quote different rates
−decentralised network — specs and performance vary by host
#5 in gpu cloud providers · free egress, free iops, free internal transfer — behind a headline number that is eight times what most readers think it is.
81/100
verdictthe most generous data terms in the category attached to the most misread price — read it as a node rate, divide by eight, and the spot tier is genuinely competitive.
CoreWeave vs Hyperstack
Hyperstack
CoreWeave
price
$2.50/gpu/hr
$6.16/gpu/hr
free tier
no
no
h100 $/gpu/hr
$2.50
$6.16 (derived)
spot
$2.00 (tier unclear)
$2.46 (derived)
min unit
not stated
8-gpu node
egress
free
free
sla
linked, terms not published
none published
switch forlarge-scale training where you were going to rent whole nodes anyway and data movement is constant.
pros
+egress, iops and internal transfer all free
+cheapest cold object storage here at $0.015/gb/mo
+spot works out to $2.46/gpu — genuinely competitive
+purpose-built for large-scale multi-node training
cons
−prices per 8-gpu node, routinely misread as per-gpu
−on-demand at $6.16/gpu is expensive
−no published sla or billing increment
−reserved discounts require sales with no published terms
#6 in gpu cloud providers · the lowest numbers in the category, on a peer marketplace where the number is a floor rather than a price.
78/100
verdictgenuinely the cheapest way to get an h100 if you treat it as a market rather than a vendor — the trade is that nothing about the price, host or bandwidth is fixed.
Vast.ai vs Hyperstack
Hyperstack
Vast.ai
price
$2.50/gpu/hr
~$1.92/gpu/hr
free tier
no
no
h100 $/gpu/hr
$2.50
~$1.72-1.92 floor
spot
$2.00 (tier unclear)
50%+ below on-demand
min unit
not stated
1 gpu
egress
free
varies by host
sla
linked, terms not published
none published
switch forbudget-constrained batch work where you can shop hosts and tolerate variability.
pros
+lowest advertised h100 rates in the category
+per-second billing, no minimums, no rounding
+68+ gpu types across many independent hosts
+reserved tier up to 50% off for 1-6 month terms
cons
−marketplace floor, not a guaranteed price
−bandwidth billed separately by some hosts, rate unpublished
#7 in gpu cloud providers · european capacity at a fair rate, on a pricing page that answers the hourly question and no others.
76/100
verdicta credible european option with published preemptible and b200 rates — undercut by a pricing page that omits egress, storage, idle billing and sla entirely.
Nebius vs Hyperstack
Hyperstack
Nebius
price
$2.50/gpu/hr
$3.85/gpu/hr
free tier
no
no
h100 $/gpu/hr
$2.50
$3.85
spot
$2.00 (tier unclear)
$2.15 preemptible
min unit
not stated
not stated
egress
free
not stated
sla
linked, terms not published
none published
switch forteams with european data-residency requirements who want blackwell access without a hyperscaler contract.
pros
+clear per-gpu on-demand and preemptible rates
+b200 published at both on-demand and preemptible rates
+european provider for data-residency needs
cons
−no egress, storage-for-compute, idle or sla detail published
#9 in gpu cloud providers · single-gpu rental with billing that stops when you power off — at a headline price the vendor itself calls a promo.
70/100
verdictthe friendliest operational model here — rent one gpu, power it off, stop paying — attached to the second-most expensive rate and a price flagged as temporary by its own documentation.
Paperspace (DigitalOcean) vs Hyperstack
Hyperstack
Paperspace (DigitalOcean)
price
$2.50/gpu/hr
$5.95/gpu/hr
free tier
no
no
h100 $/gpu/hr
$2.50
$5.95 (promo)
spot
$2.00 (tier unclear)
not offered
min unit
not stated
1 gpu
egress
free
free
sla
linked, terms not published
none published
switch forstart-stop experimental work inside an existing digitalocean account.
pros
+single-gpu rental, not node-only
+billing stops when the machine is powered off
+free ingress and egress
+integrated with an existing digitalocean account
cons
−headline rate is self-labelled a promo with no stated list price
#10 in gpu cloud providers · the only provider selling a guaranteed, dated block of h100 capacity — in units of eight, at a price you have to reserve to get.
64/100
verdictexpensive, sold only in eights, and the one place you can book h100s for a known window — which for some buyers is the entire decision.
AWS vs Hyperstack
Hyperstack
AWS
price
$2.50/gpu/hr
$5.19/gpu/hr
free tier
no
no
h100 $/gpu/hr
$2.50
$5.19 (capacity blocks)
spot
$2.00 (tier unclear)
not verified
min unit
not stated
8-gpu instance
egress
free
tiered, charged
sla
linked, terms not published
not captured this pass
switch forenterprises that need a contractual capacity guarantee on a specific date more than they need a low rate.
pros
+the only dated, guaranteed capacity reservation in the category
+per-second ec2 billing with linux at no licence cost
+enterprise support, compliance and ecosystem
cons
−no single-gpu option — fixed 8x h100 instances only
−$5.19/gpu/hr, among the most expensive here
−classic on-demand pricing not machine-readable
−on-demand capacity often unavailable without reserving
every tool on this page went through the same test as Hyperstack — same tasks, same order, scored the same way. the comparison tables are the figures from that testing, not vendor spec sheets.