The DePIN Revolution in AI Art Generation
In January 2025, Sogni launched their testnet with 3,500 users and an ambitious goal: reach 10,000 users by March 31st for their first airdrop snapshot. What happened next exceeded even their wildest expectations. Within 90 days, Sogni’s user base exploded to over 340,000 active creators, generating a staggering 1.75 million AI images per day at peak demand.
This isn’t just another growth story. It’s a testament to how decentralized physical infrastructure networks (DePIN) can democratize access to AI computing power—and how Sogni partnered with Salad to make it happen.
Building the “Supernet”: A Network for the People, Powered by the People
Founded in late 2023, Sogni set out with a bold vision: pull creative AI out of the corporate cloud and back in the hands of creators! Private, permissionless, and user-centric. Unlike traditional AI services that rely on centralized infrastructure, Sogni built what they call the “Sogni Supernet,” a two-tier network architecture that’s both revolutionary and practical.
“Our product has always been about empowering creatives,” explains Mark Ledford, Sogni’s CTO. “To do that at scale without the corporate clouds and credit card fees, we knew we had to build a grassroots community GPU network, a DePIN network. Thanks to partners like Salad, that network is thriving and more accessible than ever for artists to create and for worker nodes to earn.”
The Supernet consists of two tiers:
- The Relax Network: Powered primarily by Mac computers, handling lighter workloads
- The Fast Network: High-performance Nvidia GPUs (averaging RTX 4090s) for rapid image generation
This dual approach allows Sogni to serve different user needs while maintaining the decentralized ethos at its core.
The Challenge: When Viral Growth Meets Infrastructure Reality
As Sogni’s testnet opened to the public, the team faced a challenge that would make or break their vision. Thousands of new users were signing up literally overnight, creating massive demand spikes averaging 1.5 million images per day.
Initially, Sogni tried to scale with their own bare metal servers. “We bought a handful of gaming 4090 machines off Amazon as our first nodes,” Ledford recalls. “But we needed a surefire way of scaling to meet any kind of viral demand.”
The limitations became clear:
- Physical space constraints (“I’m hosting two boxes at home—I’m not trying to have my own server farm”)
- Inability to scale down during low-demand periods
- High upfront capital requirements for hardware
- No flexibility for demand fluctuations
Traditional cloud providers weren’t the answer either. AWS doesn’t offer RTX 4090s, a consumer-only GPU, and enterprise GPU providers only offered enterprise GPUs that were overkill for our inference needs and would sink the startup financially. “The power, popularity, and cost of recent Nvidia gaming GPUs was our sweet spot,” Ledford notes. “An RTX 4090 has 90% of the performance of a $10-20k enterprise GPU for our needs at a fraction of the cost. Network a few hundred together, and you will have parallel processing capabilities that can rival corporate GPU colos without the corporation.”
Why Salad: The Perfect Match for DePIN Infrastructure
When evaluating cloud partners, Sogni needed more than just competitive pricing. They needed a provider that understood the unique requirements of a decentralized inference network and could scale with their explosive growth.
What Made Salad Stand Out:
- Unbeatable Economics: “Salad not only had the lowest prices in the market for image generation but also offered us incredible scalability,” says Ledford. The team secured enterprise discounts and the flexibility to pay in stablecoins, crucial for a crypto-native startup.
- Container Groups Built for Scale: Unlike other providers offering one-off GPU rentals, Salad’s Container Groups provided a professional, intuitive interface. “I have not had a need to poke around the API just because we don’t spin up and down instances that frequently. I can just use your interface, and it’s so intuitive,” Ledford explains.
- Real Support from Real Engineers: “With Salad, I can directly connect with the engineering team and get the kind of support we wouldn’t from a bigger cloud provider,” Ledford emphasizes. This direct line to technical expertise proved invaluable during integration.
- Proven at Scale: Knowing that Civitai, a massive player in the AI art space, successfully scaled with Salad gave Sogni confidence. “We were aware that Civitai was able to scale their on-site generator with you guys. That gave us the assurance that you were the right folks to deal with.”
- Community Power Ethos: Salad’s DePIN model perfectly matched Sogni’s ethos and brand. Leveraging powerful consumer GPUs that would otherwise be sitting idle is cost-effective, environmentally friendly, and helps Sogni stay decentralized. Sogni Supernet is a “decentralized network of networks” and has never experienced an across-the-board GPU outage, as there is no central provider.
The Technical Implementation: Smart Architecture for Massive Scale
Sogni’s implementation on Salad showcases how thoughtful architecture can maximize distributed computing resources:
Dynamic Model Management
With over 100 models on the platform, Sogni built a system where workers dynamically download models based on network demand. “The more models that a worker has on it, the faster it can switch between them to meet demand without having to dedicate a worker node to each model,” Ledford explains. This approach works particularly well with Salad’s infrastructure, where longer-lived workers become more efficient over time.
Concurrency: The Addictive Differentiator
While competitors focused on single-image generation speed, Sogni leveraged their distributed architecture for massive concurrency. “We can do 16 images in the same four or five seconds, which is incredible,” Ledford notes. “That’s the addictive part. Where other platforms offer up to 4 images per batch Sogni also offers up to 512.”
Democratizing Network Participation Through Salad Recipes
A key innovation in Sogni’s ecosystem is the ability for anyone to become a network contributor through Salad’s Recipe system. Users who don’t own high-end GPUs can still participate as Sogni Workers by renting compute power through Salad. This creates an inclusive model where aspiring contributors can test the waters before investing in hardware or simply earn rewards without the upfront capital expense. “It’s very convenient for them to, with the least technical knowledge, get up and running with a rental through Salad,” Ledford explains. This accessibility has been crucial in building Sogni’s distributed network while giving more people the opportunity to earn from the AI revolution.
Results: Empowering Creators and Workers Alike
For Sogni:
- Dramatically reduced infrastructure needs while handling increased traffic
- Cost savings compared to enterprise GPU providers
- Flexibility to handle demand spikes during contests and campaigns
- Reliable infrastructure supporting millions of daily image generations
For Creators:
- Affordable access to AI art generation
- Fast generation times with 16-image concurrency, 512 image batches
- Reliable network performance ensures consistent quality and uptime for creative workflows
For Workers:
- Earnings up to $350/month per RTX 4090
- Low barrier to entry via Salad GPU rentals
- Dual participation as both creators and compute providers
“There are quite a lot of users who want to participate who don’t have their own machines,” Ledford explains. “It’s very convenient for them to get up and running with the least technical knowledge with a rental through Salad.”
The Power of Aligned Missions
The success of Sogni on Salad illustrates a fundamental truth about the future of AI infrastructure: democratization isn’t just an ideal—it’s a competitive advantage. By leveraging Salad’s distributed GPU network, Sogni transformed what could have been a crippling infrastructure challenge into a sustainable growth engine.
“We chose Salad because we share a common mission,” reflects Ledford. “Salad is an OG in the distributed compute space. You guys gave us a great rate that was even cheaper than your already rock-bottom industry prices.”
As Sogni continues to grow, rendering over 100 million images in their testnet, their partnership with Salad proves that the future of AI doesn’t have to be controlled by a few massive corporations. Instead, it can be truly decentralized, accessible, and powered by the community it serves.
For startups looking to scale AI workloads without breaking the bank, Sogni’s journey offers a clear lesson: sometimes the best infrastructure isn’t the most expensive—it’s the most aligned with your mission.
Ready to scale your AI workloads on distributed infrastructure? Contact Salad to learn how our GPU network can power your next breakthrough. To see Salad in action generating art through a Sogni demo app, visit Sogni’s Photobooth app. Visit this link to try out Sogni with 100 free tokens.
