A structural difference in deployment model, tool coverage, and what you can actually run — before you sign up.
Every claim below is sourced. See notes at the bottom of this page.
| Capability | Clusterra | CryoCloud |
|---|---|---|
| Compute in your AWS account | Yes — full BYOC | No — CryoCloud's infrastructure |
| cryoSPARC (ab initio, heterogeneous refinement) | Yes — BYO license, runs natively | No — prohibited by cryoSPARC non-commercial license (fee-for-service clause) |
| RELION-5 (Bayesian polishing, multi-GPU) | Yes | Yes |
| cryoSPARC → RELION hybrid pipeline | Yes (cs2star/pyem in catalog) | No (no cryoSPARC) |
| Cryo-ET subtomogram averaging (STA) | Yes — validated: 3.99 Å on EMPIAR-10164 | Partial (IMOD/PyTOM primitives, no validated end-to-end STA) |
| AreTomo2 (tilt-series alignment + CTF) | Yes | Not documented |
| Warp / M (frame alignment, CTF refinement) | Yes | Not documented |
| IsoNet (denoising / missing-wedge correction) | Yes | Not documented |
| MemBrain (membrane particle picking) | Yes | Not documented |
| Data egress to retrieve results | None — data stays in your S3 | ~$0.09/GB (~$90/TB) AWS standard egress |
| GPU scale (max concurrent) | Karpenter auto-scale; spot + on-demand mix; no hard cap | Fixed CryoCloud capacity; queue behind other users |
| Your data leaves your environment | Never | Yes — transmitted to and stored on CryoCloud servers |
| Existing AWS BAA / compliance posture | Inherits automatically (same VPC, same IAM, same KMS) | New data processor agreement required |
| Published all-in pricing | Plans from $99/mo + your own AWS spot; no GPU-hr cap | 100–300 GPU-hr/yr included, then metered + ~$90/TB egress |
Cryo-EM SPA: RELION-5 3D classification + refinement on a public EMPIAR dataset, multi-GPU on A10G spot. Full pipeline from micrographs to reconstructed map. Read the case study →
Cryo-ET STA: Immature HIV-1 Gag (EMPIAR-10164), AreTomo2 tilt-series alignment → RELION-tomo gold-standard refinement. 3.99 Å FSC on a $2.30 spot run. CryoCloud has no documented end-to-end STA pipeline at gold-standard resolution. Read the case study →
One number that includes everything you actually pay. Clusterra = the plan fee plus your own AWS spot; nothing hidden, no egress.
A 2-structure/month lab on the Researcher plan ($99/mo, 250 GPU-hr included) plus ~$100 of your own AWS spot lands around ~$200/mo all-in. One cryo-EM structure is ~100–140 GPU-hr; a realistic project runs ~$150–300 in your own A10G/g5 spot. There is no annual GPU-hr ceiling — you can scale a busy month and pay only your own AWS compute over the included hours ($0.40/GPU-hr).
Cryo-EM facility labs get their first structure free and subsidized rates through our facility channel.
CryoCloud bundles only 100–300 GPU-hr per year — a 2-structure/month lab burns the entire annual allowance in 1–2 months, then pays metered overage on top. Add ~$90/TB egress every time you pull processed maps back out. The realistic all-in for that same lab lands around ~$575–910/mo, and the GPU-hr ceiling is a hard wall, not a soft overage.
With Clusterra (BYOC), egress is $0 — data never leaves your S3.
Not feature gaps — architectural choices that don’t change with a pricing tier.
cryoSPARC's non-commercial license prohibits “fee-for-service data processing” and “third-party use of Licensee's installation.” Any multi-tenant SaaS hosting cryoSPARC for others violates this. CryoCloud cannot offer cryoSPARC regardless of how it prices or configures its service.
Clusterra installs cryoSPARC inside your AWS account, under your license. The license terms are satisfied — you are the licensee and the operator.
Source: cryosparc.com/terms
A standard cryo-ET dataset is 1–5 TB of raw tilt series plus reconstruction intermediates. At AWS standard egress rates (~$0.09/GB), retrieving 2 TB of processed data costs ~$180 per dataset — before any CryoCloud platform fee.
At 10 datasets/month that's $1,800/mo in egress alone, in addition to the platform cost. With Clusterra, data stays in your S3 from the moment the microscope writes it. Zero egress.
Source: aws.amazon.com/ec2/pricing/on-demand (data transfer pricing)
cryoSPARC's ab-initio reconstruction and heterogeneous 3D classification are now standard for initial model building and particle cleanup. RELION's Bayesian polishing and CTF refinement then push resolution. The two-tool pipeline has been the field norm since RELION-3 / cryoSPARC-2.
Switching from .cs to .star format (pyem / cs2star) is a documented friction point. Clusterra ships pyem pre-integrated. With CryoCloud, you have no cryoSPARC step to convert from.
Source: RELION/cryoSPARC interop discussed in nature methods, cryoSPARC forums since 2018
We’d rather you pick the right tool.
Clusterra deploys into your own AWS account and manages everything — you do need an AWS account and you pay AWS directly for compute. If you have no AWS account and want a single bundled invoice with a subsidised academic free tier, CryoCloud's SaaS may be simpler to start. That said, our $99/mo Researcher tier and the facility channel's free first structure now cover the low-volume academic case too — and without the per-year GPU-hr cap.
If cryoSPARC is genuinely not in your workflow, your data has no confidentiality requirements, and egress costs at your dataset volume are acceptable — CryoCloud's RELION environment covers the basics. The tradeoffs (SaaS data handling, no cryo-ET) don't bite until you scale.
For industry teams, core facilities, or anyone running cryoSPARC, cryo-ET, or datasets with IP sensitivity — the calculus is different. Clusterra handles all the AWS infrastructure; the only thing you bring is an AWS account.
Free 30-day pilot. cryoSPARC + RELION + your tilt series, in your account, before you commit — no charge.
Book a 30-min pilot call →