Lambda Scalar Server and Support

This is a Notice of Proposed Contract Action under FAR Part 12 Acquisition of Commercial Items and FAR Part 13 Simplified Acquisition Procedures. This notice is for information only; a solicitation will not be issued. THIS IS NOT A REQUEST FOR QUOTATION. No contract will be awarded on the basis of offers received in response to this notice. The associated North American Industry Classification System (NAICS) code is 334111 with a small business size standard of 1250 employees.

The Government requires the purchase of one (1) Lambda Scalar server to support ongoing research and development in artificial intelligence (AI) and machine learning (ML). The required server system is comprised of the following components:

Operating System: Ubuntu 22.04 - Includes Lambda Stack for managing TensorFlow, PyTorch, CUDA, cuDNN, etc. Processor: 2 x AMD EPYC 9554 (64 cores 3.10`3.75GHz, 256MB cache, PCIe 5.0)

GPU: 8 x NVIDIA L40S Memory: 768GB | 24 x Hynix DDR5-4800 RDIMM - 32GB Operating System Drive: 1 x Samsung PM9A3 NVMe Gen4 M.2 SSD - 1.92TB Scratch Drive: 2 x Samsung PM9A3 NVMe Gen4 U.2 SSD - 7.68TB TDP/BTU/PSU: External TPD = 4,323W | Internal TPD = 4,040W | Heat = 13,790BTU/h = 1.149 Tons of cooling | PSU config=(2+2) | PSU Wattage = 2,000 per PSU @208V+ | Ships with 4x C13-C14 power cords Native Network: Dual 10G RJ45 & 1G IPMI Lambda Scalar Hardware Support - 3 Year Standard: Three year hardware warranty with Lambda technical support covering hardware issues Lambda Scalar System Support - 3 Year Standard: Three year Lambda technical support covering software issues including Lambda Stack, ML frameworks, drivers, OS and BIOS The Government intends to negotiate a sole source contract with Lambda Labs, Inc., 2510 Zanker Road, San Jose, CA 95131 The Government has determined that other companies' products do not meet, or cannot be modified to meet the Government's needs.

The Lambda Scalar Server includes Lambda Stack, a proprietary artificial intelligence (AI) software which provides a one line installation and managed upgrade path for PyTorch, TensorFlow, CUDA, cuDNN, and NVIDIA drives. The Lambda stack is designed to work seamlessly with the NVIDIA GPUs to optimize performance over using out-of the box NVIDIA drivers.

This unique and proprietary technology, combined with the L40S server's optimized design, results in unparalleled performance in the continued research and development of AI and machine learning (ML) tasks that other providers cannot match. Furthermore, AFRL/RI currently possesses a significant number of existing servers that work seamlessly with the Lambda Stack and Lambda products. The required Scalar Server is compatible with this previously procured equipment.

This feature will allow simplicity of installation and provide interoperability with AFRL/RI servers currently being used for AI/ML training, and inference operations. Use of another source's products would cause AFRL to experience a duplication of costs not expected to be recovered through competition. Additionally, Lambda provides a team of Machine Learning Engineers to provide Machine Learning support to customers working on building AI/ML solutions.

AFRL/RI is a Government Leader in the Large Language Model space building production level systems and requires this level of support from the Lambda support team in order to meet mission requirements. Their team of experts provide comprehensive guidance and support throughout the project ensuring seamless integration into our existing infrastructure and system configuration for optimal performance.

This is a notice of proposed contract action and NOT a request for competitive proposals; however, all responsible sources may submit a capability statement or proposal, which shall be considered by the agency. All inquiries should be sent by email to Jennifer.Calandra@us.af.mil. Responses must be submitted by the date/time listed on the notice. Any response to this notice must show clear and convincing evidence that competition would be advantageous to the Government in future procurements.

Responses received will be evaluated; however a determination by the Government not to compete the proposed procurement based upon responses to this notice is solely within the discretion of the Government. The Government anticipates an award date on or before 31 January 2024. A Sole Source Justification will be attached to the subsequent Notice of Award in accordance with applicable regulation.