io.net (@ionet) header
io.net (@ionet) profile picture

The intelligent stack for powering AI workloads | https://xitter.cf/t.co/hIYFLxle8l: decentralized GPUs | io.intelligence: inference & agents | https://xitter.cf/t.co/EinR91I0wl

Manhattan, New York
Joined May 2018
Photo by io.net (@ionet)Photo by io.net (@ionet)Photo by io.net (@ionet)Photo by io.net (@ionet)Photo by io.net (@ionet)Photo by io.net (@ionet)Photo by io.net (@ionet)Photo by io.net (@ionet)Photo by io.net (@ionet)Photo by io.net (@ionet)
Pinned Tweet
io.net just passed $30 million in total network earnings. Proof that AI compute doesn’t have to be controlled by a handful of giants. AI infrastructure built for the many, not the few.
26
31
16
230
15,595
More than $1B in the pipeline. The demand is there. But the hyperscaler queue costs you weeks of waiting and close to 3x the price. io.net CEO @Gaurav_ionet explains why the fix is a marketplace, not more data centers.
More Than A Billion In The Pipeline At @ionet The compute crunch is real, and it comes down to access, not demand. At AWS that means weeks, sometimes months, of waiting at almost 3x the price. @ionet's answer is a marketplace that pulls in GPUs from data centers and individuals. Gaurav Sharma (@Gaurav_ionet) breaks it down.
3
10
4
92
6,922
#1 in DePIN by revenue. $16.07M, 30-day annualized. This is what real workloads, real customers and real revenue look like. Infrastructure demand is the metric that compounds.
TOP 10 DePIN PROJECTS BY REVENUE Which decentralized infrastructure networks are generating the most revenue? Here are the top 10 by 30-day annualized revenue: $IO: $16.07M $GEOD: $8.71M $RENDER: $6.60M $AKT: $2.89M Dabba: $2.60M $HNT: $1.87M $FIL: $1.35M $LPT: $297.7K $XNET : $158.4K Althea: $143.6K One additional growth signal: Render recorded +28% revenue growth over the previous 30 days in the latest comparable DeFiLlama-derived snapshot. DePIN is increasingly about real infrastructure demand, not just token incentives. Which network has the strongest business model?
12
23
8
147
10,570
364 bytes. That’s what one ~4-byte token was costing us. @ilkh0m, @rajaguru95 and @vikashkaushik01 found it, fixed it and shipped it with no change to time to first token. 48x less bandwidth. Zero drama. Great infrastructure is built by engineers who sweat the bytes.
9
12
6
105
8,497
8% to 76% in a year. That’s the share of @OpenRouter tokens used by AI agents. At Breakpoint, @solana counted 120+ teams building AI on its network. Their agents need wallets and compute. io.net gives them GPUs on demand. Payments move value. Compute does the work.
13
13
4
116
8,310
12,696 GPU hours. Zero failures. UC Berkeley’s RAIL Lab trained a robot navigation model on io.net with real-world data from Frodobots. 85.7% navigation success vs. 33.3% for the baseline. Open compute helped take this research from the lab to real-world.
13
19
7
140
8,910
$9.9 billion in one quarter. That’s what Anthropic spent on compute. More than all its other operating costs combined. Its answer is to spend over $900 billion in long-term commitments to secure capacity it controls. That’s a scale almost no company on Earth can match. Most just need GPUs they can use now, and stop paying for when the job ends. That’s what io.net was built for.
3
9
5
97
8,711
Hyperscalers sell you a promise. Then you get a service credit when it breaks. Hardware-level isolation, cryptographic attestation and failover within 60 seconds are not promises. They’re proof. We broke down how secure, reliable AI compute actually works.
6
12
5
115
9,044
GPU prices didn’t just spike. According to the @WSJ, they were propped up. Nvidia financed its own customers, then hit pause when the chips piled up. Turns out the shortage had a sponsor. More data centres won’t fix that. Better use of the compute that already exists will. That’s what io.net was built for.
15
15
4
107
9,984
3x lower compute costs. That’s what @vistaralabs saw moving inference to IO Intelligence. While generating 5,600 apps in 2 months with zero infrastructure failures. Affordable at 200 runs a day. Still affordable at thousands. That’s what io.net makes possible. io.net/blog/vistara-labs-cas…
13
12
8
137
12,248
60% of organisations overspent on AI last year. Half approved AI spend without knowing what it would add up to. You can’t budget for what you can’t see. And most companies can’t afford the surprise. Compute pricing should be transparent. Contracts should flex as needs change. That’s what io.net was built for. shorturl.at/tTesk
6
6
6
100
10,504
The @Microsoft outage didn’t just take down email. It hit Copilot, AI connectors, and stopped businesses from operating. That’s the risk when your AI stack runs through one provider. Their downtime is your downtime. Spread the supply, and you spread the risk. That’s what io.net was built for.
10
8
5
113
10,347
30 to 45 weeks to get new chips. About 95% of enterprise GPU capacity sitting idle. AI has an access problem. @ionet is changing the equation. 35M+ compute hours delivered. 1M+ added every month. No lock-in. No contract tiers. Thanks to @fluence_project for including us in their new report.
The same GPU costs about 2x more at a hyperscaler than at a GPU cloud. The four largest hyperscalers plan about $730B of capex in 2026. Who’s winning the race for GPUs? We put prices, capex and supply side by side in The AI Compute Race Report, with @near_ai @gensynai @Compute_Labs @fortytwo @Theta_Network @PhalaNetwork @BoundlessHQ @gaib_ai @ionet 👇
7
12
4
116
9,739