Prop 315: PIP3.5 — GPU Capacity Maintenance via Provider Incentive Program
PIP3.5 — GPU Capacity Maintenance via Provider Incentive Program
1. Summary
This is a funding proposal attached to PIP03, specifically a request to shift funds allocated to B200s to validated GPUs on the Akash Network.
We request funding from the Community Pool to extend contracts for the network’s most utilized resources—H100s, H200s, and A100s—preventing a drop in network capacity and ensuring continuity for ongoing AI workloads.
2. Context
2.1 Background
- Akash launched prior GPU incentive pilots (PIP01 and PIP02) to attract high-end GPU providers and validate demand. To date, through the PIP programs, Overclock Labs has brought on over 1,100 high end GPUs to support AI workloads.
- These GPUs have provided accessible compute to develop highly utilized resources on Akash, such as AkashML.
2.2 Results of Prior Pilots
- PIP01 successfully introduced high-grade GPUs and attracted large-scale workloads.
- PIP02 expanded deployment further, including H200s, H100s, and A100s, with high utilization across clusters.
- PIP03 is currently validating demand through pro6000s and 5090s on the network.
2.3 Current Situation: Contract Sunset
Several PIP01 and PIP02 GPU contracts are now approaching expiration. Without follow-up action, in demand GPU capacity will leave the network. Furthermore, Overclock Labs believes it is in the network's best interest to shift from the more expensive B200s, as originally outlined in PIP03, and focus on maintaining the current GPUs on the network.
B200s were originally included in PIP03 to start validating demand for next-generation GPUs on Akash and to get ahead of where the AI compute market is going. The idea was the same pattern we’ve followed with earlier PIPs: introduce new hardware in a limited way, learn from real usage, and then scale once the network and market are ready.
Since PIP03 was approved, a few things became clear: while interest in B200s absolutely exists, availability is still limited and pricing remains high. At the same time, the network already has strong, proven demand for H100s, H200s, and A100s that are actively running production workloads today. Many of those contracts are now approaching expiration. Given that reality, we believe it’s a better use of community funds to keep highly utilized GPUs online rather than allocate capital toward hardware that isn’t yet ready to be deployed at scale on Akash.
This isn’t a step back from B200s, it’s a timing decision. B200s remain a clear target for future expansion. For now, this proposal focuses on protecting the capacity that users are already relying on every day.
3. Validated Demand
- Average utilization across H200s, H100s, and A100s has stayed consistently over 50% with peak usage cresting 95%.
- Ongoing interest from model-training and inference users further confirms continued growth.
3.1 Demand of GPUs
Over the course of the PIP02 contract timelines, from December 2024 to November 2025, average utilization of H200s, H100s, and A100s were as follows:
| GPU Model | GPU Usage % |
|---|---|
| H100 | 59.30% |
| H200 | 57.34% |
| A100 | 61.61% |
| RTX 4090 | 56.07% |
Overclock Labs believes that a proper retooling will positively impact both the Akash Network as a whole and the community whose funds are being used to renew key contracts referenced above.
4. Proposal
4.1 Quantities & Budget
Total Request Amount: $789,768.13
New GPU budget:
| GPU | Total GPUs | GPU Hourly Rate | Contract Term (months) | Extended Cost |
|---|---|---|---|---|
| H100 | 24 | $ 1.50 | 12 | $ 315,360.00 |
| H200 | 64 | $ 2.00 | 12 | $ 1,121,280.00 |
| A100 | 32 | $ 0.82 | 12 | $ 229,862.40 |
| Total GPUs | 120 | Total Cost | $ 1,666,502.40 | |
| Less: Funds released from PIP03 | $ (1,050,624.00) | |||
| Plus: OCL impact due to token price change** | $ 41,658.97 | |||
| Net Request | $ 657,537.37 | |||
| Vol Buffer %* | 20.11% | |||
| Vol Buffer Amount | $ 132,230.76 | |||
| Total Request | $ 789,768.13 |
*This buffer accounts for the historical daily volatility of AKT measured over a period of 56 days. By providing a more substantial buffer against potential downswings in AKT, we mitigate the need to request any budget shortfalls through subsequent proposals. In the event of excess funds above the US dollar amount for labor, taxes, and overage, all remaining AKT will be promptly returned to the community.
**1,861,865.60 AKT liquidated to $902,419.49. B200s represent 70.697% of the original PIP03 budget. Due to downward price movements, $637,990 was left to cover $1,050,624 in original B200 estimated costs. $370,975 (70.697% * $524,735) of the volatility buffer was attributed/absorbed by the B200 budget line item, leaving exposed losses to OCL of $41,658.97.
Amended historical cost for PIP03 GPUs:
| GPU | Total GPUs | GPU Hourly Rate | Contract Term (months) | Extended Cost |
|---|---|---|---|---|
| B200* | 0 | $ - | 0 | $0.00 |
| RTX PRO 6000 | 24 | $ 1.50 | 12 | $311,040.00 |
| RTX 5090** | 8 | $ 0.60 | 12 | $41,472.00 |
| Total Historical Cost | $352,512.00 |
*Originally requested funds for 32 B200s for a total of $1,050,624.
**Due to limited 5090 demand, budget allocations only reflect the single node of 5090s that Overclock brought on to Akash Network as of 1/20/2026. Overclock Labs will continue to review demand and use additional PIP03 funds left over for 5090s as needed.
Limited Market Impact & Transparent Reporting
Limited Market Impact
Overclock Labs will custody the requested funds in a distinct wallet so that funds from any other source are not commingled.
All funds will be liquidated and managed in a manner that ensures minimal impact on the market. These funds will be handled with the same care and attention as all previous Community Funding Proposals, with liquidations done in a fashion that will not adversely affect the market. In practice, the effort of this liquidation will add depth to the AKT market for buyers looking to enter.
In some cases, the Overclock Labs will use structured market-making (MM) activity to manage funds responsibly. Conversions will take place gradually over defined trading windows rather than through large sell transactions. Spreading activity over longer windows helps ensure that each trade reduces the likelihood of noticeable price movements and helps maintain a stable market for all participants. This measured approach also allows the Overclock Labs to meet operational needs while minimizing market impact and supporting steady liquidity within the AKT ecosystem.
Please feel free to share any thoughts, questions or other feedback in this discussion thread.
Transparent Reporting
The Provider Console will display current GPUs on the network along with new GPUs brought on through the PIP.
All costs and records will be made publicly available through reports to ensure maximum transparency and accountability.