While our industry plays catchup, the AI era is threatening to obsolete traditional data centre and telecom infrastructure. Dramatic as this may sound, a new world is being described by visionaries like Marc Ganzi, CEO of Digital Bridge, and tech innovators like Jensen Huang, CEO of NVIDIA. Together they describe an inflection point announcing a new dawn for connectivity and computational firepower.
The Constraints of Legacy Systems: A Call for R(e)volution
Considering the rapid evolution of AI, how long can traditional data centers sustain the growing demands without fundamental changes? Ganzi’s foresight into a once-in-a-lifetime opportunity to build AI infrastructure is paralleled by the urgent need to overcome the limitations of legacy systems. “We are on a 10-year journey to truly enable the infrastructure to support Gen AI,” Ganzi asserted last week at Metro Connect USA, recognizing the current systems’ inability to scale bandwidth or reduce latency to meet the demands of an AI and IoT-powered future.
Such bottlenecks present both technological hurdles and barriers to potential market growth and innovation. Before we get into into the trillion-dollar implications, let’s consider the warning from NVIDIA’s latest earnings call, where the conversation about legacy systems took on an even more urgent tone.
The Trillion Dollar Write-down: NVIDIA’s Warning
Huang’s commentary during NVIDIA’s remarkable Q4 earnings call this week lends further credence to Ganzi’s transformative narrative. “The world has reached the tipping point of a new computing era,” he declared bullishly, as NVIDIA’s earnings soar and data centre revenues triple YOY, indicative of a market rapidly transitioning from general-purpose to accelerated computing. This shift is not merely a response to technology’s evolving demands; it’s an acknowledgment that the $1 trillion installed base of data centre infrastructure must adapt or risk obsolescence. As another well-known innovator recently said, “Let that sink in”.
Having established the financial stakes, we now turn to the operational transformation underway. These operational shifts are beyond iterative; they represent a fundamental reimagining of data centres and telecoms fabrics, now evolving into the so-called “AI factories”.
The AI Factories: A New Era for Data Centres and Telecoms
Imagine the data center of the future: will it be merely an upgraded version of todays, or will it be something entirely new, an ‘AI factory’ fine-tuned for the era of intelligent computing?
Juang’s prophecy of data centres transitioning from general-purpose to accelerated computing encapsulates a broader industry trend. These aren’t merely data centres anymore; they are becoming AI factories, purpose-built refineries dedicated to purifying raw data into actionable intelligence. This evolution parallels the connectivity infrastructure sector’s trajectory. Both these pivots are not just trends; they are necessities driven by the computational demands of modern AI, where Moore’s Law no longer suffices. The profound growth in AI workloads mentioned in NVIDIA’s earnings call is demanding a different form of network connectivity. The concept of AI factories brings us directly to the other leg of the issue: the legacy telecoms constraints that are increasingly becoming bottlenecks to progress.
Legacy Constraints: The Bottlenecks of Progress
As all industries inevitably move towards increasingly digital and interconnected frameworks, the limitations of legacy infrastructure become glaringly apparent. A report by Dell’Oro Group underscores an exponential trajectory in capacity demands, with Ethernet data centre port speeds expected to exceed 800 Gbps by 2027. Yet, current data centre architectures topping out at 100 Gbps are struggling to keep pace. The scalability of bandwidth remains throttled by outdated designs, unable to flexibly accommodate the terabit data torrents required by AI models and IoT networks.
With these bottlenecks identified, the industry’s path forward points to innovative solutions like Network as a Service (NaaS), which promises a future of flexibility and adaptability.
Could Network as a Service (NaaS) be the key to unlocking unprecedented flexibility in our networks, and how might this reshape the competitive landscape?
Network as a Service: A Flexible Future
Emerging NaaS architectures include built-in cloud-style principles that promise a departure from the rigid, hardware-centric models of the past. These new competitive alternatives offer a vision of global self-service automation, which cuts through the complexity of interoperability and the inertia of manual NetOps. NaaS offers flexibility, enabling on-demand orchestration of network resources, “limitless” resources and baked-in alignment with the self-service and usage-based billing ethos of the cloud. As we contemplate the promise of NaaS, it’s clear that this is just one facet of a larger trend: the rise of next-gen connectivity providers poised to bridge the gap between today’s needs and tomorrow’s possibilities.
Crossing the Chasm
Promising a bridge across these technological chasms, the inevitable rise of next-generation connectivity providers has become increasingly apparent. These bridge builders are redefining the norm with innovative solutions that offer scalability, reduced latency, and interoperable platforms. They stand out as the architects of a new infrastructure paradigm, one that seamlessly scales, adapts, and responds to the dynamic needs of AI-driven Enterprise.
The Narrative of Transformation: Insights from Industry Leaders
Ganzi’s observation of the industry’s cyclical opportunities, paired with Huang’s assertion of the necessity to accelerate every workload possible, underscores both the urgency and potential of this pivotal era. “Making sound investment decisions is the most important thing to do as a sector,” Ganzi emphasizes, urging a collective navigational effort through the industry’s challenges and opportunities.
Huang’s reflection on NVIDIA’s data centre growth trajectory, driven by the demand for AI and LLMs (large language models), supports the narrative of this transformation. “The versatility and leading performance of our data centre platform enables a high return on investment for many use cases,” he explains, highlighting the alignment of technical capabilities with broad multi-industry use cases.
Solutions and Business Model Alignments
In this transformative phase, the industry is witnessing the natural arrangement of elegant business models with technical capabilities. For example, AI workloads require a new breed of data centres capable of low-latency, high-bandwidth connectivity. The solution lies not in adding “bigger faster CPU” hardware but in hardware acceleration using GPU, FPGA and ASICS and rearchitecting networks to be scalable and routable, supporting xPU-to-xPU connectivity across multiple provider platforms and geographies.
The embrace of high-speed networking is a testament to the industry’s recognition of AI’s burgeoning demands. Forecasting by Dell’Oro Group anticipates nearly all back-end AI network port speeds to reach 800 Gbps and above by 2027, a clear indicator of the market’s trajectory towards higher performance standards. Front-end ports offering access to the Internet will surely keep up, evidenced by the new generation of NaaS WAN providers.
The variability in deployment approaches, as seen with major hyperscalers, underscores the need for customizable solutions. Who will bring hyperscale tech to Enterprise? There’s no one-size-fits-all; the nuanced requirements of AI applications demand a diverse, multivendor approach, blending standardized and proprietary technologies to create a robust, future-proof telecoms and data centre ecosystem. How will we, as an industry, reconcile the urgency to adapt with the complexity of executing such a transformation, and what role will you play in this new era?
Conclusion: Decoding the Puzzle
As we collectively gaze out over this newly discovered precipice, the shift towards next-generation data centre and connectivity providers is part industry trend, part industry imperative. The insights provided by Ganzi and Huang are more than crystal ball forecasts; they are blueprints for a future where connectivity and intelligence converge seamlessly. The AI brain needs a complimentary central nervous system.
The overarching narrative of “transform or die” offers a lens through which we can view the exponential needs of a data-driven, AI-empowered future — a future where the digital infrastructure is both a conduit for data and the foundation upon which the next era of innovation is built.
To be continued…
In subsequent articles in this series, we will deep dive into the practical applications of the transformative shift in cloud computing and AI integration within data centres. We’ll explore real-world case studies showcasing companies at the cutting edge of this evolution, examining the strategic overhauls they’ve implemented to thrive in the AI era. Our journey will uncover the step-by-step strategies that facilitate a smooth transition from legacy systems to advanced, AI-ready infrastructures.
Expect to glean industry best practices and actionable insights that stakeholders at all levels can use to align with the future of data management. We’ll introduce a comprehensive decision-making framework tailored to guide your company through assessing its readiness for adopting these technological advancements.
Our discussions will extend to selecting the right vendors, with a keen focus on the criteria that ensure their solutions not only integrate seamlessly with your current systems but also support your growth trajectory.
And because such a significant shift comes with its set of risks, we will provide robust risk management strategies to safeguard your operations throughout this transition.
Stay tuned as we embark on this series that promises to arm you with the necessary insights to not only participate in the conversation but to lead the charge in the next era of digital infrastructure.
The future is here, and it’s time to prepare.