Any prediction of what to expect in the new year will mention artificial intelligence (AI), and rightly so. It’s the most transformative technology since the dawn of the internet itself, and generative AI made what was once a utopian ideal suddenly accessible to anyone with a web browser. Yet it still feels like we’re just at the precipice of the technology – and much will occur in 2024 to help it realize its potential, particularly at the network level.
But it’s not just about AI. If we remove the understandably myopic focus on AI for a moment, what we generally are consuming more and more of is content, particularly video. And as we continue to consume data at rates never seen before, opportunities – and the environmental impact the consumption of data is having on the world around us – will increase as well.
It’s why 2024 promises to be transformative in more ways than one, with the network at the forefront of enabling those changes.
AI will begin to impact the network in transformative ways
The public ‘faces’ of AI – those that the everyday consumer is aware of, such as large language models such as ChatGPT and Bard – are simply the tip of the iceberg when it comes to AI and its disruptive power. For one, with AI expected to drive a multi-fold increase in demand for bandwidth and network performance, AI is redefining how data centers are networked.
A study by OpenAI found that the growth in AI training parameters will increase 35 times every 24 months, and the growth of AI training compute will increase 10 times every 18 months. All of this is done without much of that data seeing the “light of day” i.e. the end user – less than 10 percent of data produced by a data center leaves the premises.
As the data we consume gets smarter, it stands to reason that our networks must become smarter to cater to it. And for AI to realize its potential and to ensure the effectiveness of the myriad use cases in various states of invention or deployment, the network that provides it needs to be ‘always on’, with low latency, incredible capacity and compute power beyond current capabilities.
Thus, our networks not only need more bandwidth and lower latency to deliver these applications, but an upgrade to their fundamental architecture.
As new networks are built out and existing networks are upgraded in 2024, we’ll see the vision of the adaptive network – one that relies on programmability, analytics, software control and automation – become the standard. This network will be essential as it can self-heal, re-route traffic as required, learn from its traffic and data, and optimize its traffic and reduce congestion.
We’ll also see a massive demand for compute and connectivity requirements extending out to the network edge in 2024. This will require operators to build network fabrics that are as open to “machines” as they are to individuals, allowing consumption of network and compute resources through APIs. This will be an important monetization factor for service providers and cloud providers alike.
Our rapacious data appetite will lead to more edge and regional data centers
Our demand for data continues to grow, and our emerging interest in and demand for AI only exacerbates the amount of bandwidth and capacity requirements we need from our networks. According to some estimates, around 120 zettabytes of data will be generated this year, increasing to 181 zettabytes in 2025.
But let’s take AI out of it for a second and focus on what we’re consuming more than anything else: video. Recent research estimates that video – be it from YouTube, streaming services such as Netflix and video-native social media applications like TikTok – accounts for 65 percent of the world’s internet traffic. As content providers begin to deliver 4K as the standard, suddenly Full HD will seem archaic – and we’ll soon be curious about 8K. That’s already a quadrupling of bandwidth requirements just to deliver the same movie, just a little clearer.
That’s why, in 2024, we will see even more data center build-outs, new submarine cable networks, satellite broadband roll-outs and more to cater to this demand, and not necessarily in North America or the US, typically the world’s hubs for data centers. Already Microsoft has declared it will invest AUD$5 billion into building data centers in Australia to “help Australia seize the AI era,” for example, and we will see similar news emerge outside of North America and Europe as hyperscalers, network providers and bandwidth providers build out their ecosystems, delivering more capacity at the edge to meet the world’s insatiable demand for data.
A focus on sustainability: It’s now or never
Massive amounts of large-scale data center builds are still happening across the world, while edge data centers are rolling out in earnest across each continent. However, all of these rollouts require significant amounts of energy to power them, and the broader Information Communication Technology (ICT) industry is seen as one of the heavier-emitting sectors.
This last point needs to change if we are any hope of meeting net zero targets: specifically, what we’re powering and how it’s being powered.
In 2024, I predict we will see a more deliberate selection of data center locations (wherever possible) to avail of renewable energy – for instance, will eco-countries like Norway (with incredible renewable energy infrastructure already in place along with cool temperatures) see an influx of data center deployments?
The trade-off between building expensive power lines vs. cheap high-speed fiber connectivity will become an important factor in data center builds, while within the data center we’ll see a focus on more efficient hardware from pluggables to the servers themselves.
You’ll see technology like WaveLogic 6 deployed which enables transmission capacity to be doubled with the same footprint and result in a 50 percent power-per-bit reduction, for instance, all while enabling network operators to scale their networks to address future requirements. This will become the benchmark for data center organizations, submarine cable networks, network providers and others as sustainability becomes a paramount concern.
Meanwhile, networks will continue to see a greater focus on the use of software in lieu of hardware. Software-Defined-Networking (SDN) and Virtualized Network Functions (VNFs) can utilize bandwidth much more efficiently, which will ultimately reduce the amount of power-hungry network hardware needed.
What’s certain is that we simply can’t continue at this pace of data and energy consumption, and expect to meet net-zero targets.
