GPU Compute Servies for Network of Optimal Dynamic Energy Signatures

Solicitation / RFP / RFQ

This Request for Quotation (RFQ) is for GPU compute commercial services. The government intends to award a single Firm-Fixed-Price (FFP) contract with a 5-month base period and a 5-month option period. This procurement is conducted under Federal Acquisition Regulation (FAR) Part 12, Acquisition of Commercial Products and Commercial Services.

• Technical Capability (Go/No-Go Requirements):

  • Availability of 50 GPUs by May 15, 2026.
  • Current SOC 2 Type II certification for data centers, located in non-adversarial nations.
  • Provision of at least fifty (50) NVIDIA HGX H100 80GB 700W SXM5 8X GPUs, or equivalent/better (e.g., NVIDIA Blackwell architecture), with at least one hundred (100) additional identical GPUs in reserve.
  • Provision of at least 550TB of Network File System (NFS) storage.

• Technical Approach:

  • GPU Hardware: Emphasis on advanced technology like NVIDIA Blackwell architecture.
  • System Reliability & Redundancy: Approach to maintaining uptime and redundancy.
  • Network Performance: Description of network architecture and interconnects.
  • Support & Administration: Tiered support structure (24/7/365) and Service Level Agreement (SLA).
  • Data Center Location: Identification of data center countries (preference for US-hosted solutions).
  • Implementation & Onboarding: High-level plan to ensure operational readiness by May 15, 2026.

• Past Performance:

  • List up to three (3) relevant contracts from the past three years (similar scope, size, and complexity).
  • For each reference: Contract Number, Client Name, Point of Contact, Total Contract Value, Brief Description of Services Provided.

• Contract Type: Firm-Fixed-Price (FFP).

• Period of Performance: 5-month base period and 5-month option period.

• Quotation Submission: Submit electronically to NODES@darpa.mil in Adobe PDF or Microsoft Office compatible formats. Quotes must be valid for a minimum of 90 days.

• Quotation Format: Separated into three volumes:

  • Volume I: Technical (10-page limit).
  • Volume II: Price (No Limit).
  • Volume III: Administrative (No limit).

• Questions Deadline: Submit electronically to NODES@darpa.mil no later than April 24, 2026, at 5:00 PM EST.

• Amendments: Quoters must acknowledge all solicitation amendments.

• Evaluation: The government will review documents for completeness and determine quoter responsibility and eligibility for award. Preference will be given to solutions hosted entirely within the United States.

Not specified in this document.

• FAR provision 52.212-1 applies, with amendments. References to FAR 52.212-3 and FAR 52.212-5 are removed and not applicable.
• Quotes must adhere to US letter size (8 1/2" x 11") pages, with charts/tables not exceeding 11" x 14".
• Text font: Times New Roman, 11 point minimum.
• Pages exceeding limits will not be evaluated.
• Pricing must be a single, all-inclusive per-GPU-hour rate.

Solicitation / RFP / RFQ

This document outlines the evaluation criteria for awarding a contract, emphasizing a "best value" approach where the Government selects the most advantageous quote considering price and other factors.

The evaluation is conducted in two steps:

• Step 1: Technical Capability (Pass/Fail) - Quotes are assessed against requirements in Section L. Meeting all requirements results in a "Pass"; failure to meet one or more results in a "Fail" and disqualification from further evaluation.
• Step 2: Best Value Evaluation - Only "Pass" rated quotes are further evaluated through a comparative analysis to determine the most advantageous to the Government.

1. Factor 1: Technical Approach (Significantly more important than Price) - Evaluates the quality and thoroughness of the proposed technical approach, focusing on system reliability/redundancy, network performance, support/administration, data center location, and implementation. Superior approaches that reduce risk or enhance capability (e.g., providing equivalent or superior GPUs) will receive higher ratings.
2. Factor 2: Past Performance (Significantly more important than Price) - Assesses the relevance and quality of the quoter's recent past performance on similar contracts, using provided references and other sources.
3. Factor 3: Price - Calculated based on the estimated usage of 50 GPUs operating 24/7 for a 10-month term (base plus option). Price will be evaluated for fairness and reasonableness.

The contract will be awarded to the responsible quoter whose quote is determined to be the most advantageous to the Government, considering price and other factors (best value). The Government may award to a higher-rated technical quote with a higher price if the technical superiority offers a worthwhile benefit. The Government may award without discussions.

Solicitation / RFP / RFQ

This solicitation (HR001126QE003) is for GPU compute services, providing a dedicated, managed high-performance computing (HPC) solution. This includes compute hardware, network-attached storage, high-speed networking, robust security controls, and comprehensive user support for the "Network of Optimal Dynamic Energy Signatures Program."

• GPU compute services

• Compute hardware

• Network-attached storage

• High-speed networking

• Robust security controls

• Comprehensive user support

• Detailed technical specifications are provided in Attachment 1.

• The estimated contract value is USD 40,000,000.00.

• The contract type is Firm Fixed Price.

• The period of performance for the base period is 5 months (May 15, 2026 - October 14, 2026).

• An option period is available for an additional 5 months (October 15, 2026 - March 14, 2027).

• Offer Due Date: May 05, 2026, 05:00 PM.

• Evaluation will be based on price and other factors, as detailed in Attachment 3 (Section M - Evaluation Factors for Award).

• Instructions, conditions, and notices to quoters are in Attachment 2 (Section L).

• The acquisition is unrestricted.

• NAICS Code: 518210, with a size standard of USD 40,000,000.00.

• Offerors must comply with cybersecurity requirements, including CMMC Level 1 (Self) or higher, as detailed in DFARS 252.204-7021 and 252.204-7019.

• Payment requests and receiving reports must be submitted via Wide Area WorkFlow (WAWF).

• The solicitation incorporates FAR clauses 52.212-1, 52.212-4, and 52.212-5 by reference, with attachments provided.

Statement of Work (SOW) / Performance Work Statement (PWS)

To provide GPU compute services for the Defense Advanced Research Projects Agency (DARPA) Biological Technologies Office (BTO) to support the Network of Optimal Dynamic Energy Signatures (NODES) Program. The program aims to enable distributed, data-driven biological discovery through large-scale data integration, AI-enabled analysis, and iterative experimentation.

• Provide a dedicated, managed high-performance computing (HPC) solution.

• Include compute hardware, network-attached storage, high-speed networking, robust security controls, and comprehensive user support.

• Offer large-scale availability of AI-specific hardware to meet urgent government needs.

• Compute Resources: Access to fifty (50) dedicated, private NVIDIA HGX H100 80GB 700W SXM5 8X GPUs (or equivalent) for model training and inference. Hardware must allow direct access without intermediate layers and feature low-latency systems for multi-GPU AI training.

• Availability: GPU availability required by mid-May 2026.

• Storage: 550TB of NFS storage for data and computational models.

• Networking: Dedicated Internet Access on a 100G Port.

• Redundancy: Additional HGX H100 server available for outages.

• Support: 24/7/365 customer and scientific software support for up to 100 users.

• Security: Isolated Virtual Private Cloud (VPC) deployments with default security. Contractor must hold SOC 2 type II certification.

• Reporting: Monthly reports on resource utilization and system performance, plus ad-hoc government-requested metrics.

• 5-month base period.

• 5-month Option 1.

• Contractor shall provide all necessary personnel, equipment, supplies, facilities, transportation, tools, materials, supervision, and non-personal services.

• Contractor shall not have any data rights to the data.

• Resources must be optimized for large-scale, high-performance scientific computation and multi-GPU training.