Dispersed

On-demand GPU compute for all builders. No vendor lock-in, no cloud bottlenecks. Built by the @rendernetwork team.

[Dispersed @Dispersed_ai]

@robinhodalphaX In what way - what are you thinking?

[Dispersed @Dispersed_ai]

So, how does https://t.co/soHNAJ8Lkf actually work? Here’s the process in four steps:

1/ You submit a job and specify the hardware you need (GPU, VRAM, RAM) and your container.

2/ The network matches your job to an available node from a global pool of independently run GPUs.

3/ Your job runs in an isolated container.

4/ Pick BATCH or PERSISTENT depending on your workload.

BATCH is billed for execution time and runs to completion.
PERSISTENT is billed hourly and runs until you stop it.

Try it: https://t.co/blXwZH7eCG

[Dispersed @Dispersed_ai]

@klq1q84 We'll get there 🙌

[Dispersed @Dispersed_ai]

Most cloud GPU providers give you a black box: an API that logs, gates, and restricts how you use it.

@dispersed_ai is a composable, portable protocol for compute instead. Match your job to the right hardware and control your stack. Don’t rent every cycle from a hyperscaler.

https://t.co/Vw1WLVonfg

[Dispersed @Dispersed_ai]

Dispersed is live for two kinds of people:

1/ Builders who need affordable, on-demand GPU compute without cloud lock-in: https://t.co/qNsOhgyCsr

2/ GPU owners who want to earn RENDER by putting idle hardware to work: https://t.co/pjJTgypidx

Build with the compute you need, without relying on a single centralized provider.

[Spenser Dickerson @SpenserFX]

With everything pushing toward faster throughput, I’ve been exploring Gaussian Splats as a real-time exchange format for heavy 3D assets.

For this test: 3 Big Cat assets, 900 renders, 90 minutes.

I used the @rendernetwork API to push the datasets directly from @Blender then reconstructed each asset as a high-fidelity 3DGS.

The result is a fraction of the original overhead, while retaining an incredible amount of visual detail. I can drop these cats into new 3D scenes, run them directly on web and mobile, or bring them back into @OTOY Octane for real-time performance, relighting and final rendering.

Heavy 3D assets → Rendered datasets → 3DGS → Real-time anywhere.

Check out the Big Cats in real-time:

Tiger
https://t.co/yThCqMjP8H

Jaguar
https://t.co/iG5kNvgQ3y

Lion
https://t.co/hFq9cwNaxJ

[Dispersed @Dispersed_ai]

https://t.co/WZWw5JRs6d

[Dispersed @Dispersed_ai]

Dispersed isn’t a hypothetical. Real workloads are already running on it:
- Continuous, auto-deploying 3D rendering
- Document analysis at scale
- Risk analytics
- Personal finance management tools
- Agentic AI infrastructure

Different problems but same compute layer.

Check out how https://t.co/rq9YSDglEh, for example, is automating analysis of scientific studies in order to break through the replicability problem (aiming to save $20B annually):

[Dispersed @Dispersed_ai]

https://t.co/pmnO4RbWoX

[Dispersed @Dispersed_ai]

The GPU market is projected to hit $353B by 2030.
Meanwhile, an estimated 40% of global GPU capacity sits completely idle.

That gap is the entire reason @Dispersed_ai exists.

[Dispersed @Dispersed_ai]

https://t.co/hbTSUjtBtf

[Dispersed @Dispersed_ai]

Every day since Jan. 1, 2026, a Bitcoin block turns into a unique 3D sculpture on Bitmap by @MhxAlt. It’s fully automated and there is no standing server.

The cost on traditional cloud: $8
The cost on Dispersed: a few cents

Same pipeline but a different pricing model. Here’s how:

[Dispersed @Dispersed_ai]

@ebravick @ManifestNetwork 💪

[Eric Bravick @ebravick]

Happy to be supporting Dispersed with @ManifestNetwork

[Dispersed @Dispersed_ai]

https://t.co/z5yUFZr8nf

[Dispersed @Dispersed_ai]

Dispersed wasn’t built from scratch to chase an AI trend.

Since 2017, @rendernetwork has proven that idle GPUs can be coordinated to unleash creative work at scale, powering over 80 million rendered frames along the way.

@Dispersed_ai extends that same proven model beyond rendering, for builders working on AI and general compute.

[The Render Network @rendernetwork]

.@rendernetwork just crossed 80 million frames rendered.

On July 19, the network hit 77 million. As of Sept 10, that number is now 80 million.

3 million more frames rendered by the distributed network in under two months.

More to come.

[Dispersed @Dispersed_ai]

A few things that actually got demoed:

- Pocket Pets: virtual pets with no database
- Night Shift: LLM observability
- Johnny API: cartoon band auditions powered by Stable Audio

Everyone left with credits to keep building. If you want to be in the room for the next event, or want to start now: [email protected]

[Dispersed @Dispersed_ai]

On Aug 26, we sponsored a Power Hour in Denver right before Let’s Vibe, giving builders about 40 minutes to deploy something real on @Dispersed_ai.

Starter recipes included Ollama + Open WebUI, ComfyUI, and LightX2V, or builders could bring their own stack.

[The Render Network @rendernetwork]

AI agents need compute. Who powers it is becoming one of the biggest questions in AI.

Render Network Foundation’s @SunnyOsahn is joining the “Agentic AI and the Battle for Compute Infrastructure” panel at @conf3rence on Sept 16 at 11:10 CEST in Dortmund.

He’ll be joined by Thomas Lüke (@googlecloud) and André Liesenfeld (@Microsoft) with Daniel Heinen moderating.

Learn more: https://t.co/gVlQqtNUdy