PETER DYKES EXA Q&A
Talking about capacity leads us to how generative AI is driving demand. Is this an issue for you? AI is a real trend that really does create demand in our in our industry and is here to stay,
the hyperscalers and some have said they expect the impact of AI to be a five to 10 times increase in capacity between their core data centres, and that they are in the process of very quickly looking to procure more capacity between their datacentres, and we’re talking many, many pairs of fibres between centres. What I’m also hearing is we’re very much on the back end of this in Europe, and the US is a year or so ahead, so actually, the impact in the US is far more significant than anything we’ve yet seen in Europe.
PD
SR
hopefully for good reasons. Everyone is talking about it, but what does it actually mean for us? I think for lots of companies that build and operate datacentres, it’s more impactful than for the telecom operators right now, particularly when there is a need for megawatt datacentres with 100 kilowatts per rack, just for training AI large language models. That doesn’t seem to be driving a lot of increase in telecoms bandwidth requirements, because our understanding is that the training is done once, in one place. That takes a long time to do but once the model is launched, it switches to inference. We are however seeing more demand for services to the Nordics, from startup companies or ones that have only been around for about a year, where a lot of AI workloads are going, because it’s cold and you get a better PUE [Power Usage Effectiveness] when you’ve got a low ambient temperature. But I think it’s also because energy is cheap in the Nordics, and is generally very green, if not 100% renewable. The uptick in capacity in the Nordics is not in itself transformational, however. For us, what is interesting is that we deal with
Steve Roberts, SVP Strategic Investments & Product Management, EXA Infrastructure.
What is your strategy going forward?
PD
our radar. There’s a lot of datacentres being built in Atlanta right now, some of which are fuelled by AI, and some by hyperscale. We also think there’s need for more diversity across the Atlantic, so there make possibly be another trans- Atlantic cable in the future. In Europe, you may well see us making some additional investments heading East and Southeast, certainly extending further across the Mediterranean. We’re not planning to enter the Middle East or North Africa, but we probably do want to explore how we can take a bit more demand from there. So, for example, creating a PoP in Egypt, to take some subsea traffic that’s landing there or traffic that’s coming from Africa.
SR I think our strategy has really been over the last couple of years has been catch up. We’ve been a little bit under invested when we were under previous ownership, and I think some of the other private equity owned companies had invested more than we had. So now we have 37,000 kilometres in construction, and we’ve been looking at infilling areas where we were a little bit weak, but we’re also looking at what we believe the next growth areas are, and what are the adjacent geographies we can connect to that will help drive more traffic onto our network. We’re focused a lot on Southern Europe and trans-Atlantic markets, but we will be expanding further west and south in the US but not into Latin America. I certainly I think Miami and Atlanta would be on
Thank you.
PD
EXA Infrastructure network Image: EXA
www.opticalconnectionsnews.com
9
ISSUE 36 | Q1 2024
Made with FlippingBook flipbook maker