BarakahGrid Dubai Pilot - 15 Node. Dual Track (Nvidia +RISC-V)

[BarakahGrid]

Key updates from original version:

1. Deployment Model

Before: 15-node mixed cluster (5 NVIDIA + 10 RISC-V) from day one.

Now: Phased approach — start with 10 NVIDIA nodes, add 3 RISC-V nodes as a secondary pilot, then expand into heterogeneous South Asia clusters.

2. Risk Management

Before: Running NVIDIA + RISC-V simultaneously (higher complexity).

Now: De-risked by proving yield on NVIDIA first, then layering RISC-V benchmarking side-by-side.

3. Funding Narrative

Before: $600k to fund a single dual-track pilot.

Now: $600k spread across phases (NVIDIA → RISC-V → heterogeneous), with clear sequencing.

4. KPI Structure

Before: Single cluster KPI (uptime, yield, solar integration).

Now: Distinct KPI sets per phase (NVIDIA yield, RISC-V performance per watt, heterogeneous yield vs CapEx).

5. Expansion Roadmap

Before: 15-node → 100-node cluster scaling.

Now: Phased 10 NVIDIA → add 3 RISC-V → heterogeneous clusters in Pakistan/Bangladesh → blueprint for 100+ nodes.

✅ Why this update:

Keeps the sovereignty vision alive (RISC-V still central),

but ensures credibility + immediate yield by leading with NVIDIA.

[brewsterdrinkwater]

Thank you very much for taking the time to write this. It is great to see new people participating in the Akash open framework.

I have a number of questions.

1. Can you introduce the team and the company to the Akash Community?
2. Have you all built and managed a provider on Akash?
3. For "Nvidia nodes live" Q1 goal, what resources were you thinking about bringing onto the Akash Network?
4. What kind of user(s) would you all see deploying on the provider you want to pilot?
5. What tier datacenter would the provider be at?
6. After the pilot ended, what is next?

It would be great if you all could come to sig-providers next week to meet some other folks in the community. Check out the calendar for special interest groups and working groups.

[BarakahGrid]

Thanks for your questions and for engaging with the proposal, I really appreciate it.

1. Team intro: Right now BarakahGrid is small — I’m leading it, and I have a partner who helps with local logistics and ops. We’re at the early stage of building, but committed to scaling properly.

2. Experience on Akash: I haven’t personally deployed a provider on Akash yet — this pilot would be the first step, and part of the reason I’m putting it forward is to learn and document as I go with community support.

3. NVIDIA node resources: For Phase 1 I’m now focusing on 10 NVIDIA nodes (A100s) in Dubai to prove yield. This keeps it simple and in line with Akash’s current demand.

4. Users: I expect demand mainly from AI developers/researchers training and running models, and smaller labs priced out of Big Cloud.

5. Datacenter: The plan is to start in a Dubai facility that meets Tier III+ standards — I’ll confirm the exact site as soon as we lock hardware and hosting agreements.

6. After pilot: Once NVIDIA is proven and cash-flowing, we’ll add a small RISC-V side pilot to benchmark open hardware. Then expand into heterogeneous clusters in Pakistan/Bangladesh using consumer GPUs (4090s/3090s) to show accessibility for the Global South.

Thanks again for the thoughtful points. I’ll keep updating the proposal text to reflect this phased approach so it’s clearer for everyone.

[shimpa1]

hi @BarakahGrid ,

Thank you for the proposal.
$600k is a large amount of money., so I would request specifics regarding the proposal.

Is this planned to be built in a data center, an office or in someones home?
If it's a data center, what tier is it? Are there any certifications to attest the datacenter tier?

5 nVidia nodes is very vague.
Which exact nodes?
How many GPUs?
Which GPUs?
PCIe or SXM?
Used or new?

How about the network?
Ethernet speed?
Infiniband?
Network topology?
Do you have an AS?
What ASN is it?
How many peers does your AS have?
Which ASN are you peering with?
Upstream capacity?
Upstream redundancy?

RISC-V is not supported on Akash.
How do you plan on achieving support for the RISC-V part of the code?
Will you be maintaining the code?
If not, who is?

Any warranty?
Any collateral you're willing to offer?

• many many other questions.
  The proposal is very very vague in it's current state. I would suggest that whoever wrote it tries to answer all of those questions in as much detail as possible. If we need to ask every single question, that gets tiring very quickly.

thanks

[BarakahGrid]

Thanks for your detailed feedback — I really appreciate the way you’ve broken the questions down.

To be transparent, I’m still new to this space and don’t have all the answers on datacenter certifications, networking topology, or ASNs yet. That’s part of why I’ve put this proposal forward — so I can refine it with input from people more experienced in the ecosystem.

Based on feedback so far, I’ve simplified the plan into phases:

Phase 1: 10 NVIDIA nodes in a certified Dubai datacenter (Tier III+). Focus will be on proving yield and operational stability.

Phase 2: Add a small RISC-V pilot, side-by-side with NVIDIA, purely for benchmarking sovereign hardware.

Phase 3: Expansion into heterogeneous nodes in Pakistan/Bangladesh (consumer GPUs), once Phase 1 is stable.

For the early pilot, I’ll keep the setup straightforward:

NVIDIA A100s, PCIe-based, hosted in a known Tier III+ Dubai facility.

Standard datacenter networking (ethernet) — nothing exotic until later stages.

RISC-V nodes will only come in as an experiment after NVIDIA nodes are already up and running.

I’ll document each step in as much detail as possible so the community can review, and I’m happy to take guidance from those with more technical experience along the way.

Appreciate you pushing me to tighten the details — I’ll keep iterating the proposal with clearer specifics as things lock in.

[BarakahGrid]

We’ve updated the proposal after community feedback. The revised plan de-risks Phase 1 (NVIDIA-only), then layers in RISC-V benchmarking as Phase 2. Changelog below for clarity.

[BarakahGrid]

Update – Withdrawing Current Proposal

After reflecting on community discussions and the current priorities of the Akash ecosystem, I have decided to withdraw this grant request.

I recognize the valid concerns raised around treasury spending and the need for measurable ROI. My original proposal was too broad in scope. Instead, I am preparing a refined, low-cost, evidence-driven pilot grant focused on:

Deploying 2 Akash GPU nodes in the UK (FlameVilla site)

Measuring uptime, revenue, and power draw

Capturing and documenting surplus heat reuse (heating + hot water)

Publishing a full Pilot Case Study Report for the community

This refined proposal (≈$12k) will deliver real-world data, transparent reporting, and a replicable blueprint for sustainable Akash node deployments.

Thank you to the community for the feedback that helped shape this direction. I will be resubmitting shortly with the updated proposal.