Moving AI data off-site has to do more than break even

When I consult with a company that's deciding whether to move its AI application's data to a separate, off-site storage provider, my test is practical and simple. Will the move increase profits enough that the margins are attractive? If moving the data will increase revenue or productivity, even as a long-run projection we can determine with certainty, and that margin yields a profit over the cost of the move and the expenses that come with it, then yes.

Breaking even isn't the bar. Nobody wants to be in the red. Being in the black is okay, but that's not really what we're in business for. We want to be in the green. The move has to produce a net profit margin.

Being in the black is okay, but that's not really what we're in business for. We want to be in the green.

Why companies keep it close

The first thing I look at is reliability. A lot of companies keep their AI data centers on site because they think having something at home, something domestic, will be more efficient. They think they'll have more operational control and more direct access to everything.

In some cases, proximity does deliver that. It provides some speed of access. But it doesn't always turn out to be as efficient as they expect, because someone has to keep all of it running.

What on-site upkeep really takes

Keeping data on site means maintaining your own storage, and then contracting a company or hiring employees directly to do the maintenance. That's often a huge part of the operational cost. And maintenance isn't just keeping the hardware and the software going. A lot of what these data sites need falls into the category of janitorial: keeping the area clean, making sure dust isn't accumulating, and many other things. A lot of that requires additional personnel.

A separate storage provider usually has its own technicians and maintenance crews, people who monitor, manage, update and repair the systems and reroute when needed. When the data moves off-site, maintaining the actual hardware becomes their responsibility. That's what the companies I consult for are paying for.

Speed and efficiency can also tip the decision. If moving the data off-site will speed up operations locally, where the company's people are, that's one reason to do it. So is no longer having to manage technicians and other staff on site.

After the move

The operational costs sometimes go up at first after the move. Then they balance out or even get lower, because the company no longer has to contract a separate firm or employ people to do that work. A lot of the time, handing that work to the provider outweighs the cost.

It doesn't always translate immediately into net profit, or even a revenue increase or a better projection. But a lot of times it increases productivity among staff and on site in other ways, peripheral ones that make it hard to see right away why moving off-site was the better option.

When the maintenance work goes to the provider, the people who did it on site can sometimes be repurposed to do something else. Sometimes, unfortunately, they have to be let go. I don't do that myself, but I've consulted with a lot of companies where we've had to let go of employees because they were no longer viable on site as technicians, as maintenance staff and as engineers.