GPU Compute Servies for Network of Optimal Dynamic Energy Signatures
Overview
Buyer
Place of Performance
NAICS
PSC
Set Aside
Original Source
Timeline
Qualification Details
Fit reasons
- NAICS alignment with historical contract wins in similar service areas.
- Scope strongly matches core technical capabilities and delivery model.
Risks
- Past performance thresholds may require one additional teaming partner.
- Potential clarification needed on staffing minimums before bid/no-bid.
Next steps
Validate eligibility requirements, assign capture owner, and schedule partner outreach to confirm teaming strategy before submission planning.
Quick Summary
The Defense Advanced Research Projects Agency (DARPA) is soliciting proposals for GPU Compute Services to support its Network of Optimal Dynamic Energy Signatures (NODES) program. This opportunity seeks a dedicated, managed high-performance computing (HPC) solution, including advanced GPU hardware, storage, networking, robust security, and comprehensive user support. Proposals are due May 5, 2026, at 5:00 PM EST.
Scope of Work
DARPA requires a dedicated, private HPC solution featuring fifty (50) NVIDIA HGX H100 80GB 700W SXM5 8X GPUs (or equivalent/better, such as Blackwell B200), with an additional one hundred (100) identical or better GPUs available for redundancy. The solution must include 550TB of Network File System (NFS) storage, dedicated 100G Internet Access, and 24/7/365 customer and scientific software support for up to 100 users. All resources must be operational by May 15, 2026. The contractor will not have data rights to the program's data.
Contract & Timeline
- Contract Type: Firm-Fixed-Price (FFP)
- Estimated Value: Up to $40,000,000.00
- Period of Performance: A 5-month base period followed by a 5-month option period (total 10 months). A new CLIN 0003 for additional services has a PoP from May 15, 2026, to March 14, 2027.
- Set-Aside: Unrestricted
- NAICS: 518210 (Data Processing, Hosting, and Related Services) with a $40M size standard.
- Proposal Due: May 5, 2026, 5:00 PM EST
- Published: April 27, 2026 (Amendment 0001)
Key Requirements & Evaluation
Proposals will undergo a two-step evaluation. Step 1 (Pass/Fail) assesses Technical Capability, requiring:
- Availability of all 50 GPUs by May 15, 2026.
- Current SOC 2 Type II certification (subcontractor cert acceptable if covering entire infrastructure).
- Data centers not located in adversarial nations, with a preference for US-hosted solutions.
- Provision of specified GPU hardware and storage.
Step 2 (Best Value) evaluates passing quotes based on:
- Technical Approach: (Significantly more important than Price) Assessing system reliability, network performance, support, data center location, and implementation plan. Superior technical approaches (e.g., NVIDIA Blackwell architecture) are highly valued.
- Past Performance: (Significantly more important than Price) Relevance and quality of up to three contracts from the past three years.
- Price: Evaluated for fairness and reasonableness.
The government may award to a higher-rated technical quote with a higher price if technical superiority offers a justifiable benefit and reserves the right to award without discussions.
Additional Notes
Offerors must comply with CMMC Level 1 (Self) or higher cybersecurity requirements. While Controlled Unclassified Information (CUI) is not anticipated for this period, future phases of the NODES program may require a FEDRAMP-certified environment. All proposals must acknowledge receipt of Amendment 0001. Contact: BAA Coordinator at NODES@darpa.mil.