Foundational technology for what AI does next.

Stellar develops and supplies foundational technology, working with third-party partners on chips and hardware. Real AI workloads guide research towards technology and products that can be validated and integrated.

Conceptual illustration of chip and system collaboration
01

Foundational R&D

Start with application problems and study technical approaches and system interactions.

02

Third-party collaboration

Work with partners on engineering and integration within an agreed project scope.

03

Phased releases

Advance through research and validation, introducing results as they mature.

AI efficiency also depends on how data moves.

Our AI storage exploration provides an example: training requires a continuous supply of data, plus saving and restoring progress. When the data path cannot keep pace, computation may have to wait.

Dataset

Read and prepare the data required by a task.

Storage & supply

Keep data moving to the processing stage.

Model workload

Run training or other AI tasks.

Save & restore

Write state so interrupted work can resume.

An illustrative AI storage workload; the actual data path depends on the system.

Research shaped by the workload.

Storage exploration is a public entry point into our research. We consider components as part of an AI system and evaluate their value in actual tasks.

01

Data supply

Study sustained reads, access latency and data preparation to understand where a workflow waits.

02

Sustained workloads

Examine continuous writes, saving and recovery, with attention to performance and stability over time.

03

System interaction

Consider components, software and operating conditions together, connecting research to integration and practical use.

AskAIs Mini SSD research project imageRESEARCH MATERIAL / HISTORICAL PROJECT

AskAIs Mini SSD. An AI storage exploration.

The Mini SSD material records how model workloads led us to examine data supply, checkpoints and sustained access. It is a starting point for understanding this research path.

Research question
How storage fits model training and data movement.
Published material
Workloads, interfaces, sustained performance and engineering evaluation.
Where to learn more
The original research article remains available. Contact the team for current specifications and supply arrangements.
Read more

Shared progress. Clear responsibilities.

Stellar’s role is foundational R&D and technology supply. Third-party collaboration connects engineering capabilities, with responsibilities agreed for each project.

ParticipantRoleScope to discuss
StellarResearch & technology supplyApplication requirements, technical direction and integration of research results.
Third-party partnersEngineering & product integrationEngineering implementation, system integration and delivery responsibilities within the project.
Joint workValidation & releaseTest conditions, result documentation, release and ongoing support arrangements.

Real value. Real workloads.

AI storage and hardware research requires more than a short peak-performance test. Evaluation needs a stated environment, task and method to make results meaningful.

01

Sustained performance

Throughput and latency across sustained access and task execution.

02

Stability & durability

Behavior under prolonged workloads, writing demands and exceptions.

03

Power & thermals

How operating conditions affect performance and deployment.

04

Compatibility & integration

Interaction with hosts, software and system data paths.

These are research and evaluation dimensions. Product-specific metrics and results will accompany the relevant published materials.

From research to useful products.

Collaborations progress according to actual results. Each stage needs evidence that explains the application, the result and the next step.

  1. 01 / TECHNICAL RESEARCH

    Define the problem

    Align workloads, user needs and technical direction around an approach that can be evaluated.

  2. 02 / ENGINEERING

    Validate in a system

    Agree responsibilities with partners and progress implementation, integration and testing.

  3. 03 / RELEASE

    Introduce complete results

    Publish technology, specifications, evaluation conditions and supply arrangements as results mature.

LET’S BUILD / HARDWARE

Bring a real problem. Build the next step.

Developing hardware, integrating systems or working with an AI workload? Discuss your needs with us. Collaboration starts with the application, technical scope and capabilities each party can bring.

  1. 01Application and the problem to improve
  2. 02Existing hardware, software and operating environment
  3. 03Proposed scope and engineering responsibilities
Discuss hardware collaboration

Questions, answered.

Is Stellar a chip manufacturer?

Our role in this business is foundational technology research and supply, in collaboration with third parties. Engineering, integration and delivery responsibilities are defined for each project.

Can I buy Mini SSD today?

This page preserves Mini SSD as research material. Contact the team for current models, specifications and supply status. Plans in historical articles do not represent current availability.

What can a technology partnership include?

Start with your needs and existing systems to define the appropriate research, technology supply or engineering collaboration. Deliverables, delivery methods and support are agreed for each project.

Where will future results be announced?

We will publish research, product and partnership updates on the website and Blog as results develop. Subscribe through RSS to follow them.

Read the Stellar business update
Chips & AI Hardware Research and Technology Partnerships | Stellar AGI Labs