Last updated: July 2026.
Amazon Nimble Studio is gone. AWS wound the service down through 2024, and the console has been dark since October of that year. If you went looking for an announcement, there wasn’t much of one. The product just moved onto AWS’s “Services in Full Shutdown” list.
The bigger story is that Nimble wasn’t alone. HP Anyware, the former Teradici PCoIP and the long-time standard for remote VFX workstations, got its own end-of-life notice in May 2026. The two products that defined cloud studios for this industry died within two years of each other, and neither was replaced one-for-one.
I have history with this stack. In 2021 I built a post-production studio in the cloud by hand on EC2, then ran client work on Nimble Studio once it matured. Some of my older articles now recommend a dead product, so consider this both the correction and the current map: what to actually use in 2026, whether you need a render farm, a fleet of cloud workstations, or just frames back by morning.
What died, and when
Nimble Studio was the one first-party product that bundled both halves of a cloud studio: virtual artist workstations and a Deadline render farm, deployed from a single console. AWS shut it down in stages. Existing customer workstations became unavailable after June 19, 2024, official end of support landed on June 30, 2024, and the console and everything behind it went fully inaccessible on October 22, 2024. If you need a single date, use June 30, 2024, but in practice the wind-down ran most of that year.
It wasn’t the only service to go, either. NICE EnginFrame reached end of support in September 2025, and Amazon WorkDocs went in April 2025. AWS has been pruning this corner of its catalog for a while now.
HP followed. On May 7, 2026, it stopped selling HP Anyware, Trusted Zero Clients, and the services around them. Existing multi-year contracts keep maintenance and support until October 31, 2029, but the Anyware Trust Center and Trusted Zero Client support end October 31, 2026. If your studio runs on PCoIP, you now have a migration deadline instead of a roadmap.
The practical consequence: Amazon DCV (renamed from NICE DCV in 2024) is now the last major remote-display protocol standing for this kind of work. It costs nothing extra on EC2, and it shapes every option in the workstation section below.
The render half: AWS Deadline Cloud
The official successor to Nimble’s render farm is AWS Deadline Cloud, announced April 2, 2024, almost exactly as Nimble entered its shutdown year. I covered the launch back then; this article replaces that piece because the service has moved on since. One thing to be clear about up front: Deadline Cloud is a fully managed render farm and nothing else. There is no workstation component. AWS replaced half of Nimble and left the other half to the market.
The pricing comes in two models. With service-managed fleets, AWS runs the render nodes and you pay the underlying EC2 cost; AWS’s own pricing example works out to about $0.19 per hour for a c5.4xlarge spot instance. Customer-managed fleets connect your own machines, including on-prem hardware, at $0.015 per connected worker hour. There’s also a “Wait and Save” option that discounts jobs that can wait for a flexible start time. Most overnight renders qualify.
Usage-based licensing rents DCC and renderer licenses by the hour on your AWS bill: Autodesk Arnold runs $0.66 per hour, and Houdini Engine and Karma carry tiered rates that fall as usage grows. Out of the box, the service integrates with Maya, 3ds Max, Arnold, Nuke, KeyShot, Houdini, Blender, V-Ray, and Unreal Engine. The main 2026 addition is a practical one: Deadline Cloud now uses AI to read failed job logs, identify the root cause, and recommend a fix. Anyone who has grepped render logs at 2am understands the value.
The honest downsides. Your bill is EC2 plus EBS plus S3 job attachments plus licensing line items, which is much harder to predict than a per-frame quote from a managed farm. And getting real value out of it assumes AWS fluency. This is a tool for studios with a pipeline, not a drop-off window for frames.
Deadline 10 and OpenCue: the free options
The render manager Nimble was built around, Thinkbox Deadline 10, is still free. AWS made every Thinkbox product free back in 2022 (license-free since version 10.1.23), and on November 7, 2025 it moved Deadline 10 into maintenance mode: security updates and critical fixes only, no new features. Downloads, AWS Portal, the Spot Event Plugin, and usage-based licensing (now routed through Deadline Cloud UBL) all keep working. For a studio render wall that already runs Deadline, maintenance mode is fine for years. Just know the direction of travel: AWS would clearly prefer you end up on Deadline Cloud eventually.
If you want an actively developed open-source manager instead, OpenCue (the Academy Software Foundation project that started at Sony Pictures Imageworks) is having a good run in 2026: recent releases added a Rust-based distributed scheduler, a browser UI called CueWeb with LDAP auth, and a workstation tray app, CueNIMBY. It demands real pipeline engineering resources, but no vendor can deprecate it out from under you.
The workstation half nobody rebuilt
AWS never shipped a managed replacement for Nimble’s virtual workstations, and two years on, nothing suggests it plans to. If you need artists on cloud machines in 2026, you’re assembling the studio yourself from one of four parts bins.
DIY: EC2 G6e plus Amazon DCV
This is the architecture I built by hand in 2021, before Nimble was stable enough to trust, and later documented as a full multi-region studio build with FSx and Resilio. That architecture still works; the parts have just gotten better: the current G6e instances carry NVIDIA L40S GPUs (48GB class) and deliver up to 2.5x the graphics performance of the G5 generation in that article, and a g6e.xlarge (4 vCPUs, 32 GiB RAM, one L40S) runs about $1.86 per hour on-demand in us-east-1, roughly $1.66 on spot. Amazon DCV streams the desktop, free on EC2, with Wacom Bridge tablet support fully integrated for paint and roto work.
Two caveats from experience. First, you are the IT department: directory services, AMIs, storage, and cost hygiene are all yours (the multi-region article is a realistic picture of what that involves). Second, don’t build the budget around spot pricing. Spot discounts on GPU instances have compressed to roughly 10 percent under on-demand for the g6e.xlarge, which tells you how tight GPU supply is. Shutting down idle machines now saves you more than any spot strategy will.
Managed desktops: WorkSpaces and Arch Platform
If DIY sounds miserable, AWS’s own middle option is Amazon WorkSpaces. The old GraphicsPro bundle retired on October 31, 2025, and in February 2026 AWS launched new Graphics G6, Gr6, and G6f bundles on NVIDIA L4 GPUs, including fractional-GPU G6f sizes for lighter work. It’s a managed desktop product rather than a studio product, but if an artist needs a GPU desktop without a pipeline attached, it covers the gap. The fractional sizes are also a better fit than renting a whole L40S you won’t saturate.
The name that comes up most often as a true Nimble successor is Arch Platform Technologies, an orchestration layer that runs cloud studio infrastructure for you. Trade coverage positions it, alongside Amazon DCV, as the leading migration path for displaced Anyware users. It’s worth evaluating, with one caveat: much of that coverage traces back to the company’s own press materials, so treat it as a vendor pitch to verify against your own testing, not an industry consensus.
Zero IT: Vagon
For a solo artist, the honest answer is probably none of the above. Vagon sells cloud workstations by the minute with no AWS account, no AMIs, and no directory service to set up: T4-class machines start at $0.99 per hour, A10G tiers run $1.67 to $3.57, a quad-A10G machine is $11.97, and persistent storage starts at $7.99 per month for 75GB. It’s AWS underneath with the setup pain removed, and you pay for that convenience: heavy daily use favors DIY, the GPU generations lag (no L40S tier), and storage caps at 525GB. For an occasional heavy rig, it’s exactly right.
Two quick footnotes on this tier. Shadow PC Pro is still alive as a flat-monthly cloud PC if you want a persistent machine, though its roots are in gaming rather than production. And Paperspace, which used to be the standard rec here, was absorbed into DigitalOcean.
Managed farms: when you just need frames
Everything above assumes you want infrastructure. Plenty of people just want frames. The good news is that the managed render farm business survived the platform churn intact: RebusFarm, Fox Renderfarm, GarageFarm, Ranch Computing, and GridMarkets all made it from 2024 into 2026.
Pricing has converged to near parity. On a comparable Blender and Cinema 4D test scene, Drop & Render, GridMarkets, Fox, and GarageFarm all landed within roughly EUR 190 to 230 of each other. Typical managed-farm cost runs $0.05 to $0.25 per 1080p frame, with complex V-Ray archviz work reaching $0.30 to $0.50. The real differences are workflow: which DCCs are supported, how good the support humans are, and how much the priority tiers cost over economy.
| Service | Best fit | Pricing | Worth knowing |
|---|---|---|---|
| GarageFarm | First farm, indie and small-studio jobs | Credits; $25 minimum top-up | 24/7 human support; priced within a few percent of the big names on a standard test scene |
| GridMarkets | Houdini pipelines | Credits; same test-scene band as Fox and GarageFarm | The Houdini specialist of the group |
| Fox Renderfarm, RebusFarm, Ranch Computing | General CPU and GPU jobs | Typical range $0.05 to $0.25 per 1080p frame | All survived from the 2024 era and are still operating |
| Conductor (CoreWeave) | GPU-bound studio bursts | Instance-hour; H100-class around $2.70 per GPU-hour | No egress fees; native Blender plug-in |
| AWS Deadline Cloud | Studios comfortable in AWS | EC2 rates plus hourly licensing (Arnold $0.66/hr) | Deepest control, least hand-holding |
Conductor deserves its own paragraph because it’s the studio-grade option here. It now runs on CoreWeave infrastructure: pay-as-you-go, always-on-demand capacity, no egress fees, and a native Blender plug-in added along the way. CoreWeave is the company that went public in March 2025, was first to deploy NVIDIA’s Blackwell Ultra systems commercially, and took a further $2 billion NVIDIA investment in January 2026. That fleet is why Conductor is the right call when the job is GPU-bound and bursty, and it has expanded beyond rendering into AI and LoRA training workloads. The fair question is long-term attention: CoreWeave’s business is overwhelmingly AI now, and VFX rendering is a side lane. I’d still use it. I’d just keep a second farm account as a hedge.
Why the cloud math still works
If you’re wondering whether to skip all of this and buy render nodes instead, look at GPU prices first. The RTX 5090 launched at $1,999 MSRP and has sold for $2,500 to $3,800 on the street through 2026, driven by GDDR7 shortages and AI demand. Across the tiers, cards are running 15 to 32 percent over MSRP. (I went deeper on the shortage math in the GPU guide.) Cloud GPU pricing, meanwhile, has fallen over the same period.
Put those together and the burst case is straightforward. A thousand hours of g6e.xlarge time costs about $1,860, which is less than one street-priced 5090 before you’ve paid for the machine around it, the power, or the depreciation. Owned hardware still wins for steady, year-round render load; nothing beats a paid-off box that renders every night. But capacity you need three weeks per quarter is exactly what the cloud is for, and 2026 GPU pricing has pushed that crossover point further out than it has ever been.
Who should use what
- Freelancer or indie artist who needs frames rendered: a managed farm. GarageFarm is the cheapest entry ($25 gets you started); GridMarkets if your pipeline is Houdini.
- Small studio with an existing pipeline that wants elastic AWS rendering: Deadline Cloud with a service-managed fleet and usage-based licensing. Already running Deadline 10 on your own hardware? Keep it. Maintenance mode is fine for years, and customer-managed fleets at $0.015 per worker hour let you burst to AWS from the same manager.
- Studio that needs full cloud workstations: EC2 G6e plus Amazon DCV if you have, or can hire, the AWS skills. Arch Platform if you’d rather pay someone to run it, verified against your own testing.
- Solo artist who wants zero IT: Vagon, by the minute.
- GPU-heavy or AI-adjacent pipelines: Conductor on CoreWeave.
In conclusion
Step back and there’s a clear pattern. Every managed product in this story is gone: Nimble Studio, EnginFrame, HP Anyware, the GraphicsPro bundles. The pieces underneath them all survived: EC2, Amazon DCV, Deadline the software, the render farms. The studio I assembled by hand in 2021 outlived the polished product AWS built to replace it, which was not the lesson I expected to take from any of this. So if you’re making a five-year infrastructure decision, build on pieces you can migrate off of, and assume any managed studio product can be shut down on the vendor’s timeline rather than yours.
If you’re mid-migration off Nimble or Anyware and hit something this article doesn’t cover, leave a comment and I’ll dig into it. And if you want the full detail on the DIY workstation build, the multi-region AWS article is still the reference; I’ve flagged the parts that have changed since it was written.
