RFI_On-Premises Next-Gen AI Compute Data Center_PNNL.pdf

Telephone (509) 375-2529 Email: Samantha.Ernst@pnnl.gov

February 25, 2026

REQUEST FOR INFORMATION (RFI)

On-Premises Next-Gen AI Compute Data Center

Solution, Design, Infrastructure, and Implementation

Introduction:

The Pacific Northwest Division of Battelle Memorial Institute (Battelle), Management and

Operating Contractor of the Pacific Northwest National Laboratory (PNNL), on behalf of the

Department of Energy (DOE), is issuing this request for information (RFI) to gather industry input on the design, construction, infrastructure, and operation of a scalable, high-performance data center optimized for AI and compute workloads. Through this RFI, Battelle seeks industry input to inform future procurement strategies and technical requirements for the design and delivery of an on-premises, secure, scalable data center with an advanced AI compute infrastructure capable of training and operationalization of multi-data Large Language Models (LLMs) for scientific and national governmental missions. Initial production readiness for ~2 MW of compute is targeted for 2028, with potential future expansion to ~40 MW. The facility must support mixed AI training/inference workloads with stringent security, compliance, and availability requirements appropriate for a DOE national laboratory.

THERE IS NO SOLICITATION AT THIS TIME.

This request for information does not constitute a solicitation/request for proposals and submission of any information in response to this market survey is purely voluntary. This RFI is released pursuant to FAR 5.201(e), Exchanges with

Industry, and is issued for the purpose of conducting market research as described in FAR

10.002. As such, nothing in this request shall be

interpreted as a commitment on the part of

Battelle to enter into a contract with any respondent. Battelle assumes no financial responsibility for any costs incurred.

Battelle will evaluate industry and market information to ascertain potential market capacity, potential solution options, design/requirement considerations, and potential interested contractors. Telephone inquiries will not be accepted or acknowledged, and no feedback or evaluations will be provided to companies regarding their submissions.

BATTELLE IS NOT REQUESTING PROPOSALS AT THIS TIME AND INTERESTED PARTIES

SHALL NOT SUBMIT PROPOSALS IN RESPONSE TO THE REQUEST FOR INFORMATION.

Submission Instructions:

Interested parties who consider themselves experienced, knowledgeable, and well-versed in high-performance, advanced compute data centers are invited to submit a response and capability statement.

Responses are to be submitted no later than 4:00pm Pacific Time, March 27, 2026. The response and capability statements shall be provided in electronic format. All responses and comments must mailto:Samantha.Ernst@pnnl.gov be emailed to Samantha.Ernst@pnnl.gov.

At a minimum, responses to this RFI request shall contain the following information:

1) Organization name, DUNS number, address, email address, web site address, and telephone number.

2) Company's Business Size (Large or Small) and socioeconomic type (i.e., HUBZone, Service-

Disabled Veteran Owned, 8(a), Woman-Owned, etc.) based on the following NAICS code:

a. 518210 “Computing Infrastructure Providers, Data Processing, Web Hosting, and

Related Services” with a size standard of $40,000,000.

3) Tailored capability, qualification, and experience statements.

4) Responses to the following questions:

A. Data Center Facility Infrastructure Requirements

A.1 Tier, Structure, Security, Compliance

1. What data center Tier level (I–IV) is recommended for AI workload criticality, and what are the trade-offs (e.g., cost, complexity, availability)?

2. What are best practices for physical security and access control in high performance compute (HPC) / AI data center environments?

3. What structural requirements (e.g., floor loading, ceiling height, slab vs.

raised floor) are necessary to support high density AI-centric racks or clusters or other high density compute equipment?

4. What federal and state regulatory compliance requirements (e.g., NIST, FISMA, WA State energy codes) must be considered?

A.2 Power Infrastructure & Redundancy

1. What are the total facility power requirements and redundancy strategies

(e.g., N+1, 2N) for a scalable AI data center initially at 2 MW of compute and scalable up to 40 MW?

2. What are the recommended specifications for backup generators and automatic transfer switches to ensure reliable failover?

3. What are the recommendations for supplemental on-site power if utility distribution is limited (e.g., on-site generation, storage, microgrids)?

4. What is the recommended UPS configuration and required characteristics to avoid equipment failure during brief (less than 15 seconds) power outages?

A.3 Cooling Infrastructure, Fire/Life Safety, Monitoring

1. What cooling methods are recommended for current and future AI compute loads (e.g., air, direct to chip liquid, immersion), including operational challenges, procedures, and emerging solutions?

2. What are the capital and operational cost estimates for each cooling method?

3. What are the reliability and maintainability considerations for cooling systems?

4. How can free cooling or economization strategies be integrated effectively in the Southeast Washington State climate (desert/arid environment)?

5. What fire suppression systems are suitable for liquid cooled environments and high-density compute?

6. What monitoring solutions and preventive measures are available to detect and mitigate coolant leaks in liquid-cooled environments?

A.4 Facility Layout & Scalability

1. Assuming a 2 MW AI data center that is scalable/expandable up to 40 MW of compute, what are the recommended data center aisle layouts and airflow management strategies?

2. What hot aisle/cold aisle containment layouts are optimal for airflow management and thermal control?

3. What infrastructure is required for liquid cooling distribution (e.g., Cooling

Distribution Units, piping)?

4. Can you provide sample floor plans or layout diagrams and projected footprint (sqft) for a single-story facility (no stacking), assuming an initial

2 MW AI data center scalable/expandable up to 40 MW of compute?

5. What design recommendations support flexibility and future escalation of facility requirements as power density and cooling needs increase?

6. What design principles support adaptive and future proof data center layouts?

7. What power and cooling design strategies support future proofing and adaptability to evolving technologies?

B. Compute, Network, and Solution Options

B.1 Cluster Architecture, Density, and Growth

1. What are the recommended cluster density and layout configurations for a facility initially supporting AI compute clusters that consume up to 2 MW of power, with scalability to larger deployments up to 40 MW?

2. What is the estimated power consumption per rack and/or per cluster for AI compute workloads projected for 2028, and how should designs accommodate future growth in rack density and total capacity?

3. What strategies and design considerations support scalable expansion of compute clusters as AI workloads and hardware (e.g., GPUs, accelerators) evolve?

4. What considerations should be made for mixed cooling types (e.g., liquid cooled compute vs. air cooled storage) within the same facility?

5. What are the pros and cons of AC vs. DC power and what considerations should Battelle keep in mind?

B.2 Network & Connectivity

1. What are the recommended network and fiber connectivity configurations to support AI workloads that require high throughput and low latency (e.g., topology, oversubscription ratios, internal vs. external connectivity)?

2. What are the implications of prohibiting RF and wireless technologies (e.g., Bluetooth, Wi Fi) within the data center on operations, monitoring, and device management?

B.3 Compute Solution Options & Lead Times

1. Are there any all-encompassing, turnkey AI compute solution options available for consideration (including hardware, software stack, and integration services)?

2. What non-GPU alternatives or hybrid options/solutions exist for AI compute? What are the benefits, challenges/limitations, and considerations specific to non-GPU options?

3. What would be the expected lead time for 2 MW of AI compute capacity, and what options exist for reducing lead time and expediting delivery?

4. What integration, configuration, and start-up support options are available, and what considerations should Battelle keep in mind (e.g., staff skills, tooling, support models)?

5. Please provide a Rough Order of Magnitude (ROM) for the procurement of

2 MW of AI compute, broken down into price elements such as design, equipment, integration/support, and other major cost categories.

6. Are there additional considerations for an AI data center housed in a containerized or modular structure rather than in a traditional brick-and-mortar facility?

B.4 Reference Architectures

1. Can you provide reference architectures for AI data centers ranging from

2 MW to 40 MW of compute, including:

• Scalability of compute clusters and interconnects

• Cooling architectures

• Power distribution (utility, UPS, generators, PDUs/busways, redundancy schemes)

C. Sustainability and Resource Efficiency

1. What are the target Power Usage Effectiveness (PUE) and Water Usage

Effectiveness (WUE) metrics for AI data centers in a Southeast Washington

State (desert/arid) climate?

2. What sustainable energy integration options (e.g., Power Purchase

Agreements, on-site solar, nuclear, other low-carbon sources) are recommended?

3. What opportunities exist for waste heat reuse in AI data centers (e.g., district heating, industrial processes)?

4. What monitoring and automation systems are available for tracking energy and resource usage (e.g., PUE, WUE, carbon intensity)?

5. What design strategies support long-term sustainability and resource efficiency?

D. AI-Powered Operations, Monitoring, and Automation

1. What AI powered tools are available to optimize site metrics such as PUE, WUE, and carbon footprint?

2. What systems exist to monitor and dynamically adjust cooling based on workload, hardware temperature, and environmental conditions?

3. What predictive maintenance tools are available to detect and prevent hardware failures (IT and/or facility equipment)?

4. What AI driven solutions optimize cooling efficiency in liquid cooled environments?

5. What automation and analytics tools support future proofing of data center operations?

Respondents should clearly label and answer each question individually and to the best of their ability

It is understood that Respondents may not have the knowledge and expertise to answer every question. Supporting documentation, diagrams, and references are encouraged.

Battelle reserves the right to use any and all information submitted by, or obtained from, respondents in any manner Battelle determines is appropriate, including, but not limited to, the development of requirements/specifications, creation of future RFIs/Sources Sought/Expression of Interest requests, competitive solicitations/Requests for Proposals, etc. Respondents/interested parties should avoid including any classified, business confidential, and/or proprietary information in the response.

However, if a respondent must submit such information, the information shall be clearly marked, and the respondent must provide sufficient justification as to why such information is business confidential and/or proprietary. Battelle will review the information and safeguard it appropriately.

Not responding to this RFI does not preclude participation in any future solicitation(s), should any be issued. If solicitations are released, they may be posted on https://ebs.pnnl.gov/advertised.aspx or to

SAM.gov. It is the responsibility of interested parties and potential offerors to monitor these sites for additional information pertaining to related upcoming opportunities.

Battelle looks forward to hearing from you regarding this request.

Sincerely, Samantha Ernst

Senior Contract Specialist https://ebs.pnnl.gov/advertised.aspx https://sam.gov/