Hardware we can walk up to
The DGX sits in the Patras building we run the operation from. Partner data stays on hardware we can walk up to and unplug, which is a sentence we can say because we own the machine.
The stack · Layer 02, on-prem
NVIDIA DGX. Our own iron, in Patras.
The on-prem layer is hardware we bought and run ourselves. It exists so that the work which should never leave the building does not have to, and so a brand can be told exactly where its data sits.
What we run
//01
//02
//03
DGX in the building. What that means in practice.
For the work that should not leave
Not everything belongs in a cloud. The on-prem layer takes the work that should never leave the building, and the elastic layer takes the work whose volume moves.
Bought, not rented
This layer is capital we spent, not a subscription we renew. That is why it is the answer to a question about data location rather than a line in a vendor agreement.
Why it is in the stack
//01
//02
Layer 02. On-prem.
A brand can be told where its data is
A manufacturer handing us its catalogue can be told the machine, the building and the country. That answer is only available because the iron is ours.
It sets the boundary we operate to
The stack has three layers and the boundary between them is ours to draw. We decide what runs on our own hardware, and we do not have to renegotiate it with anyone.
Questions
What people ask. Answered plainly.
Does running NVIDIA hardware make TPL an NVIDIA partner?
No. We bought the hardware and we operate it on our own account. The memberships and partnerships we do hold are published in the row at the foot of every page, and NVIDIA is not among them.
Where is the hardware?
In Patras, in the building the operation runs from. That is the point of the layer: the machine is somewhere we can reach.
NVIDIA and DGX are trademarks of their owner. TPL S.A. bought and operates this hardware on its own account, and is neither an NVIDIA partner nor its agent.