Breakpoint AI, a San-Francisco based company, came to us looking for a data center that could accommodate its very specific requirements. The company specialises in generative AI for image datasets, and was looking for a fully managed service that wouldn’t require constant journeys to and from the data center.
The client in question had previously been searching in California, but found itself faced with a real lack of capacity in the nearby area. This led to the search being extended, and before long it had found TRG through Google.
The problem: Sky high energy costs and a lack of space nearby
The client discussed its needs with our team, and soon realised the savings that could be made by opting for a remote setup. The cost of real estate and the cost to implement in California is so much higher than it is here in Texas, for a number of reasons.
Take the cost of Pacific Gas & Electric (PG&E), for example. In California it’d set a company like this back about 24 cents per KWH. That’s approximately six times what we’re paying in Texas. Add a lack of space to the equation and it’s easy to see why it’s so punitive for California-based companies to buy their own GPUs and operate them nearby.
Working fully remotely can be a challenge, but with our fully managed service it doesn’t have to be. The client was soon reassured by our team’s responsiveness, understanding and familiarity with the GPUs. They believed that we would make the process easy for them – and of course that’s exactly what we did.
The solution: Fully managed colocation for a stress-free experience
Our answer in this case was a fully integrated solution, entirely managed by our team here at TRG. The client simply had to send us its servers. Once they arrived, we took care of cabling and labelling, rack and stack, and the PDUs.
The fully managed service that we offer allows clients to simply send servers over in a box, and our team takes care of everything from that point. In this case, we got straight to work putting servers into the rack for them, then cabled them up, popped their branded baseball cap on the top and sent them a photo of the server in its new home.
In this case, the setup itself was relatively simple. But it did require special PDUs and of course a good level of understanding was needed in terms of how these servers had to be cabled for high density power.
Like anything, the project wasn’t without its challenges. One of the servers sent via the manufacturer turned out to be faulty, so it was up to our team to package it back up and return it to the manufacturer. We managed the back and forth to remedy the situation, and with our experienced technicians on hand it was soon resolved.
The end result: A flexible and future-proof setup from our team of experts
The client was delighted with the fully managed package that we offered. It freed up their team to concentrate on pressing issues in its San Francisco HQ, and meant that no one needed to waste time travelling back and forth to the data center – even when hiccoughs did inevitably arise along the way.
We’re continuing to integrate and provide assistance as the client moves forward towards its vision to scale the company. Future plans include the addition of more and more GPUs as the company grows, and our team will be here to help make that happen.
This was a brilliantly fun project to be a part of, and we’re still working alongside the client to keep everything running smoothly as its capabilities increase.
