Chutes (Subnet 64 on Bittensor) just released a major update showing they’re not just another crypto project, and they’re generating real revenue and building products people actually use. In their recent Novelty Search (NS) presentation, the team revealed they made over $1.3 million in the last 90 days. They’re now consistently hitting $9,000 to $20,000 per day, with a peak of $22,000.
But the bigger story is their strategic shift: moving from GPU rentals to competing on intelligence itself, while launching privacy-focused consumer products that centralized AI companies can’t match.
Intelligence Arena: From Renting GPUs to Selling Intelligence
For the past year, Chutes focused on being a marketplace where you could rent powerful computers (GPUs) for AI work. You needed computing power; they connected you with someone who had it, and everyone got paid through Bittensor’s network.
But the team realized something important: renting GPUs is a race to the bottom. Eventually, prices drop so low that margins disappear. Everyone with a GPU starts competing on price alone, and nobody makes meaningful money.
So they’re shifting to something different. Instead of renting you a computer, they’re selling you the answer. You don’t care which GPU solves your problem; you just want the best result. This is what they’re calling the “Intelligence Arena.”
Here’s how it works now. You send a question or task. Multiple miners compete to give you the best answer using whatever AI models or tools they’ve built. The best answer wins. You get better results, and miners are rewarded for the quality of their intelligence rather than just for hardware.
To make this competitive, they created a prize pool that currently sits at over 47 thousand dollars. If a miner can beat the current best-performing system, they win the entire pool. Then someone else tries to beat them. It’s king of the hill, but for AI performance. This forces constant improvement without anyone telling miners what to do.
Revenue Growth
During the NS episode, Chutes shared financial records tracking their rise from a few hundred dollars a day in early 2025 to a consistent run rate of $9,000 to $20,000 daily. With a recent peak of $22,000 in a single day, the subnet has generated $1,355,176 in the last quarter alone.
What matters is where that money comes from. It’s split between regular consumers using their services ($441,287), businesses and enterprises ($378,409), and traditional money payments rather than crypto ($819,696).
To show where this money is coming from, they noted that Affine (SN120) is a key client, contributing approximately 10% of Chutes’ total revenue by utilizing their infrastructure.
Affine is far from the only major player relying on this infrastructure. Chutes has effectively positioned itself as the cloud backbone for a significant portion of the Bittensor ecosystem, powering a diverse roster of top-tier subnets. The list of active users includes heavy hitters like Numinous (SN6) and Gradients (SN56), alongside specialized projects such as Trishool (SN23), ReadyAI (SN33), Aurelius (SN37), etc.
They also track something they call the sustainability ratio, which is how much of their daily costs they cover with revenue instead of relying on token emissions. Currently, they’re covering 30 to 40 percent. Their goal is 100 percent, which would make the system completely self-sustaining without needing blockchain rewards to keep miners participating.
Two Products Anyone Can Use
To prove their infrastructure works for real applications, Chutes launched two consumer products that compete directly with major tech companies.
Chutes Search is live right now at chutes-search.com. It works like Perplexity: you ask questions, it searches the web, and gives you answers with sources cited. The interesting part is their strategy. They’re open-sourcing the entire thing. Any developer can take the code and build their own search engine, as long as they use Chutes for the computing power. This creates guaranteed future demand for their network.
Fictio.ai is their bigger bet on mainstream adoption. It’s a roleplay platform similar to Character.ai, where you chat with AI personalities; anime characters, role-playing scenarios, personal assistants, whatever you want.
The difference is privacy. On normal platforms like Character.ai, the company can technically read everything you type. On Fictio, they use special hardware protections called Trusted Execution Environments combined with end-to-end encryption. Your conversation is encrypted on your device and only decrypted inside the processor doing the AI work. Nobody else can see it, not the person running the computer, not Chutes, nobody.
Making It Easy for Developers
Chutes understands that getting developers to switch providers is hard. If they have to rewrite their entire app, they won’t bother.
So Chutes made their system completely compatible with OpenAI’s API. If you built something that uses ChatGPT, switching to Chutes requires changing literally one line of code, which is the web address where requests get sent. Everything else works exactly the same.
They’re also integrating with tools developers already use. Vercel for hosting websites, OpenClaw for autonomous AI agents, and Claude Code for programming assistants. The goal is to become the invisible backend that powers all sorts of AI applications without developers needing to think about it.
Harvard Research Collaboration
One announcement that adds serious credibility was that Chutes is now officially collaborating with Harvard University on research about how to run large language models more efficiently.
This isn’t just nice to have. It signals that what Chutes is building is legitimate computer science innovation, not just a crypto project. Academic institutions don’t partner with random startups unless the technology is genuinely interesting.
It also means the improvements happening in their Intelligence Arena, where miners compete to build better AI systems, will be studied and published by real researchers.
The Efficiency Argument
Chutes makes a bold claim about efficiency compared to big tech companies. They calculate something they call tokens per dollar of valuation, basically, how much actual AI output you generate relative to your company’s worth.
By their math, Chutes scores 384 while OpenAI, Anthropic, and others score below 5.5. They argue this proves decentralized systems can be exponentially more efficient than centralized giants.
The logic is simple. Big AI companies build massive data centers that cost billions. Chutes uses “idle compute”, which are GPUs that would otherwise sit unused in data centers around the world. They rent this stranded hardware for pennies on the dollar. The GPU exists anyway, so letting Chutes use it when it’s idle costs almost nothing.
OpenAI reportedly spends 1.4 trillion to make 20 billion in revenue. Chutes spends far less because they’re using hardware that already exists rather than building everything from scratch.
What This Actually Proves
The Chutes update matters because it shows decentralized AI infrastructure can work as a real business, not just a crypto experiment.
They have actual revenue. They’re covering a meaningful percentage of their costs without relying entirely on token emissions. They launched consumer products that real people can use. They partnered with a top university. They’re making it easy for developers to switch from centralized providers.
This addresses the main criticism of decentralized projects, that they’re all talk and no execution. Chutes is executing. The $1.3 million in revenue proves people are willing to pay for what they’re offering.
Whether they can scale this to compete with giants like OpenAI and Google remains to be seen. Coordinating thousands of independent GPU providers is complex. Maintaining quality and reliability across a distributed network is hard. Building consumer trust in decentralized services takes time.
Try Chutes Search: chutes-search.com
Check out Fictio: fictio.ai
Website: chutes.ai
Follow @chutes_ai on X
Watch the Novelty Search on Youtube
