Joined May 2026
- Tweets 67
- Following 5
- Followers 753
- Likes 15
17 Photos and videos
Pinned Tweet
Jun 12
The website is now live and GPU rentals are open.
Pay with $ICPX tokens and access enterprise grade compute at a fraction of the market rate. H100s, H200s, RTX 4090s, available right now.
Stake $ICPX to unlock even deeper discounts and make your compute costs go even lower over time.
HdeAPoHivsm9MZfeY5tW7apJEprc8Fs594bWmnzfpump
icpx.one/
36
27
151
42,434
ICPX Protocol retweeted
Well, it is still early but personalized self-hosted AI models are gonna win.
1
8
334
We’ve been heads down building and tonight we’re ready to share what’s been in the works.
Major updates dropping later tonight. Infrastructure, partnerships, and what comes next for $ICPX.
If you’ve been watching from the sidelines, tonight is a good time to pay attention.
24
4
46
3,530
$ICPX was built to tear that barrier down.
A decentralized GPU marketplace on @Solana where providers compete for your job, prices drop as the network grows, and you pay per job down to the second with zero idle billing. No minimum spend, no enterprise contracts, no gatekeeping. An H100 for $0.34/hr. An H200 for $0.90/hr. An RTX 4090 for $0.08/hr. Available right now to anyone.
But this was never just about cheaper GPUs. The creator marketplace, the inference endpoints, the one-click app templates, the managed fine-tuning jobs, every piece of this is designed to give builders what the cloud giants never wanted them to have. Full access. Fair prices. And infrastructure that gets cheaper the more people use it.
27
3
36
3,894
The numbers speak for themselves.
H100 PCIe at $0.34/hr. H100 SXM5 at $0.63/hr. H200 SXM5 at $0.90/hr. A100 SXM4 at $0.27/hr. A100 PCIe at $0.14/hr. RTX 4090 at $0.08/hr.
Every single GPU on the $ICPX network is priced at a fraction of what you would pay anywhere else. And these are not promotional rates. This is what compute costs when you remove the middleman, let providers compete for your job, and settle everything on-chain through @Solana.
Stake $ICPX and these prices drop even further. The more committed you are to the network, the cheaper your compute gets. For developers, AI teams, and anyone running serious workloads, this changes the entire cost structure of what you can build and how fast you can move.
19
5
53
4,381
The goal has always been to make GPU compute accessible to everyone, whether you are running a full AI pipeline or just spinning up a quick job to test something out. No minimums, no commitments, just compute when you need it.
20h
We'll having payment streaming on devnet by tonight. Mainnet on Monday.
After successful testing will make the program immutable and opensource.
Any comments, feedback or else drop them below.
We should be able to rent a GPU even for small jobs or just to play around.
9
3
44
3,769
ICPX Protocol retweeted
20h
We'll having payment streaming on devnet by tonight. Mainnet on Monday.
After successful testing will make the program immutable and opensource.
Any comments, feedback or else drop them below.
We should be able to rent a GPU even for small jobs or just to play around.
Jun 13
Introducing the next layer of solana:HdeAPoHivsm9MZfeY5tW7apJEprc8Fs594bWmnzfpump.
GPU rental was just the entry point.
solana:HdeAPoHivsm9MZfeY5tW7apJEprc8Fs594bWmnzfpump is now a full compute platform. One-click app templates mean you can spin up ComfyUI, Stable Diffusion, Ollama, Jupyter, and full training rigs instantly, no setup, no configuration, just deploy and run. Inference endpoints let you deploy a model and get a live pay-per-call API URL, meaning developers can build products directly on top of solana:HdeAPoHivsm9MZfeY5tW7apJEprc8Fs594bWmnzfpump compute and only pay per token, per second.
Persistent storage volumes attach directly to your instances so your data lives with your compute. Reserved and spot tiers give you flexibility, lock in a GPU weekly or monthly at a discount, or run cheaper interruptible spot compute when you don't need guaranteed uptime. And managed fine-tuning jobs make training your own models as simple as uploading a dataset, picking a base model, and letting the network handle the rest.
On top of all of this sits the Creator Marketplace. List your fine-tuned models, LoRA checkpoints, curated datasets, ComfyUI workflows, and prompt packs. Buyers pay in SOL or solana:HdeAPoHivsm9MZfeY5tW7apJEprc8Fs594bWmnzfpump and you earn solana:HdeAPoHivsm9MZfeY5tW7apJEprc8Fs594bWmnzfpump on every single sale. Buy a template and launch it straight into a live GPU rental in one click.
This is no longer just a place to rent a GPU. It's an entire AI economy with compute at the bottom and a marketplace at the top, and solana:HdeAPoHivsm9MZfeY5tW7apJEprc8Fs594bWmnzfpump is the currency that powers all of it.
Build. List. Earn. solana:HdeAPoHivsm9MZfeY5tW7apJEprc8Fs594bWmnzfpump
icpx.one/marketplace
7
4
38
7,761
Jun 13
Every layer of $ICPX is being actively upgraded. The rental infrastructure is being optimized for faster deployments and lower latency across all 5 global regions. The payment protocol on @Solana is being tightened so streaming payments settle even faster and per-job billing becomes even more precise. Node performance is being benchmarked against real AI workloads so every GPU you rent is delivering exactly what it promises.
19
10
79
5,080
Jun 13
The creator marketplace is being refined. Faster load times, smoother checkout, better discovery for models, datasets and workflows. The one-click app templates are being expanded. Inference endpoints are being stress tested. The entire platform is being pushed harder so when the next wave of users arrives, the experience is seamless.
3
5
33
2,717
Jun 13
The creator marketplace is being refined. Faster load times, smoother checkout, better discovery for models, datasets and workflows. The one-click app templates are being expanded. Inference endpoints are being stress tested. The entire platform is being pushed harder so when the next wave of users arrives, the experience is seamless.
2
4
31
2,273
We are getting ready to push a major update for @ICPXProtocol
7
6
36
6,022
Jun 13
Tomorrow is going to be interesting. 👀 solana:HdeAPoHivsm9MZfeY5tW7apJEprc8Fs594bWmnzfpump
15
8
42
4,286
Jun 13
Hey @Solana,
A GPU marketplace built natively on your network. Streaming payments, per-job billing settled on-chain, and now a full AI creator marketplace where models, datasets and workflows are bought and sold in $ICPX.
42 active nodes. 6,992 jobs completed.
The network is live and growing. 👀
17
10
69
4,017
Jun 13
Introducing the next layer of solana:HdeAPoHivsm9MZfeY5tW7apJEprc8Fs594bWmnzfpump.
GPU rental was just the entry point.
solana:HdeAPoHivsm9MZfeY5tW7apJEprc8Fs594bWmnzfpump is now a full compute platform. One-click app templates mean you can spin up ComfyUI, Stable Diffusion, Ollama, Jupyter, and full training rigs instantly, no setup, no configuration, just deploy and run. Inference endpoints let you deploy a model and get a live pay-per-call API URL, meaning developers can build products directly on top of solana:HdeAPoHivsm9MZfeY5tW7apJEprc8Fs594bWmnzfpump compute and only pay per token, per second.
Persistent storage volumes attach directly to your instances so your data lives with your compute. Reserved and spot tiers give you flexibility, lock in a GPU weekly or monthly at a discount, or run cheaper interruptible spot compute when you don't need guaranteed uptime. And managed fine-tuning jobs make training your own models as simple as uploading a dataset, picking a base model, and letting the network handle the rest.
On top of all of this sits the Creator Marketplace. List your fine-tuned models, LoRA checkpoints, curated datasets, ComfyUI workflows, and prompt packs. Buyers pay in SOL or solana:HdeAPoHivsm9MZfeY5tW7apJEprc8Fs594bWmnzfpump and you earn solana:HdeAPoHivsm9MZfeY5tW7apJEprc8Fs594bWmnzfpump on every single sale. Buy a template and launch it straight into a live GPU rental in one click.
This is no longer just a place to rent a GPU. It's an entire AI economy with compute at the bottom and a marketplace at the top, and solana:HdeAPoHivsm9MZfeY5tW7apJEprc8Fs594bWmnzfpump is the currency that powers all of it.
Build. List. Earn. solana:HdeAPoHivsm9MZfeY5tW7apJEprc8Fs594bWmnzfpump
icpx.one/marketplace
16
18
67
7,541
Jun 12
The more $ICPX is used, the cheaper it gets. That's not a promise, that's the economics of the protocol.
At 1x usage you're paying $2.50 per GPU hour. At 10x usage that drops to $0.95. As more GPU providers join the network and compete for jobs, supply increases, prices fall, and the marketplace becomes more efficient for everyone. This is the complete opposite of AWS and Google Cloud, where scaling your usage means scaling your bill exponentially.
We're continuing to build on top of this model. Every update, every infrastructure improvement, every new node that comes online pushes the price curve lower. The 91% cost saving at scale isn't a target we're chasing, it's the natural outcome of a decentralized marketplace where supply and demand set the price and no single provider controls access.
The foundation is live on @Solana. The upgrades are in motion. And the more the network grows, the more unstoppable the economics become
Jun 6
The Cost of Scaling on Centralized Cloud vs. $ICPX
The data is straightforward. Every time you scale on AWS or Google Cloud, your cost per GPU hour goes up. By 10x usage you are paying 340% more than when you started. $ICPX flips that entirely, the more the network is used, the more providers compete, and the cheaper compute gets. At 10x scale $ICPX costs $0.95 per GPU hour against AWS at $11.20. That is not a marginal difference. That is a fundamentally better model.
9
12
44
5,375
Jun 12
The foundation was always solid. Now we're making it better.
The original $ICPX protocol on @Solana is still at the core of everything we're building. Streaming payments, per job billing, on-chain settlement. That's not changing.
What we're doing now is layering on top of it, tightening the infrastructure, improving the developer experience, and preparing the network for the next wave of demand.
Same vision. Better execution. The OG tech is getting an upgrade
7
11
42
2,631
Jun 12
The GPU marketplace is just the beginning.
The next layer of $ICPX is a full ecosystem where anyone can build, sell, and monetize on top of the compute network. Imagine a marketplace where AI creators sell their fine-tuned models and LoRA checkpoints directly to other users, where curated training datasets are bought and sold on-chain, and where ComfyUI workflows, prompt packs, and environment recipes are published and traded freely.
Every sale, every transaction, every exchange flows through the $ICPX ecosystem. The token isn't just for renting GPUs anymore. It becomes the currency of an entire AI economy built on top of decentralized compute.
This transforms $ICPX from a cost saving tool into a platform. Users don't just rent. They build, sell, and earn. The network grows stronger with every creator that joins, every model that gets listed, every workflow that gets shared.
We're not just building infrastructure for AI. We're building the marketplace where the AI economy actually happens.
8
12
55
3,918
And while we're hardening the infrastructure, don't forget what's being built underneath it all.
$ICPX runs on Solana with streaming payments, meaning you pay per job, not per hour. No wasted spend, no idle billing. You use exactly what you need and pay exactly for that. What you'll be accessing is a protocol built to be as efficient as it is unstoppable.
8
9
43
3,096
Jun 12
AI agents are the next wave and they're going to need compute to run.
Right now, every major AI lab and developer is racing to build autonomous agents systems that can think, plan, and execute tasks without human input. But there's a problem nobody is talking about loudly enough: agents need GPU power to operate, and that power is expensive and locked behind centralized providers.
This is where $ICPX is heading.
The vision let AI agents autonomously rent GPU compute directly through $ICPX. No human in the loop, no credit card, no AWS account. An agent spins up, identifies it needs compute, pays with $ICPX tokens, runs its workload, and moves on. Fully autonomous, fully on-chain.
As the agent economy scales and it will the demand for cheap, programmable compute will explode. Agents don't negotiate prices, they don't shop around, they just need reliable infrastructure at the lowest cost possible. $ICPX is being built to be exactly that layer.
We're not just building a GPU marketplace. We're building the compute backbone for the autonomous agent economy and we're just getting started.
5
16
41
7,153
Jun 12
We're monitoring everything in real time, these attacks are hitting hard but our team hasn't stepped away for a second. $ICPX isn't going anywhere.
We're currently setting up enterprise-grade DDoS protection to make sure this never happens again. The attacks have been relentless, so our defense will be too.
4
5
36
2,503