YOU ARE AT:AI-Machine-LearningEnterprise exchanges: The future of IXPs (Reader Forum)

Enterprise exchanges: The future of IXPs (Reader Forum)

Internet exchange points, or IXPs, are evolving to meet enterprise demands for AI, edge compute and security

Internet exchange Points (IXPs) have adapted to meet today’s enterprise needs, evolving into some of the world’s most secure, sustainable and high-performing connectivity providers. Handling the regional exchange of internet traffic between different networks, they traditionally enabled members to interconnect directly rather than through third-party networks, enhancing the efficiency and speed of data transfer while reducing costs. But the advent of cloud services and, more recently, the move by enterprises to leverage edge compute and deploy AI-driven latency-sensitive and mission-critical use cases, has provided IXPs with new opportunities, and they are evolving to meet them with a more value-based approach.

Enterprise needs are changing

The growth in cloud applications and their adoption by enterprises has significantly increased the volume of traffic moving across hybrid private-public cloud networks. As enterprises expand beyond their own data centers, they’re developing complex inter-relationships with a much wider variety of cloud providers, internet content providers (ICPs), network service providers and partners. This distributed environment has created a diverse set of interconnection needs.

As a result, enterprises seeking performance and moving aggressively into edge compute and AI applications are driving the intensity and complexity of these interactions between different players. Edge computing refers to the processing and storing of data closer to the source to allow rapid and real-time analysis. The global edge computing market is projected to grow to $181.96 billion by 2032, exhibiting a compound annual growth rate (CAGR) of 38.2% during the forecast period.

Meanwhile, off-the-shelf large language model (LLM) productivity tools have traditionally not required fine-tuning with proprietary enterprise data or placed heavy demands on bandwidth and latency. But that is already rapidly changing. As enterprises scale AI applications, networks must provide the performance and scalability needed to support them. These AI applications will interact with end users at the network edge, making IXPs strong candidates for providing the necessary local connectivity. The edge AI market grew from $49.30 billion in 2024 to $53.54 billion in 2025 and is projected to continue growing at a CAGR of 8.84%, reaching $81.99 billion by 2030.

According to Equinix, using an Interconnection Oriented Architecture (IOA) is the most efficient way to build digital services, exchange with participants or access population centers. The economics of data, density, velocity and experience demand localized exchanges to move the highest volumes of data with the lowest latency to dense clusters of participants and population centers.

All of this signals a tremendous opportunity for IXPs, provided they customize their services even further to securely offer a range of speeds, performance and scale to meet the emerging needs of edge computing and enterprise AI applications. They must evolve from a volume-based approach to a value-based approach.

The evolved enterprise exchange

To that end, the most evolved IXPs are adapting to these new conditions by tailoring their local POPs (Points of Presence) more specifically to the enterprise customer’s needs. IXPs will not only thrive by connecting the biggest pipes to drive the largest volumes of traffic, but there is also growing demand to meet the specialized needs of the most innovative enterprise applications. Enterprises will require more than just volume; they will demand strict service level agreements (SLAs), enhanced security and localized data privacy.

Meeting these demands will require IXPs to deploy the latest networking technologies including, IP routing, optical transport, network processing units (NPUs) and network automation.

Crucially, they must deliver these capabilities in a sustainable way. As connection speeds increase to 400GE, 800GE and beyond, modern IXPs must sustainably improve power and space efficiency. For example, NL-ix reduced power consumption from 0.9115 watts to 0.1065 watts per gigabit delivered in the parts of its network where it deployed modern network processing technology.

By distributing these modern routing fabrics throughout its entire service infrastructure, IXPs can offer the performance expected from the most stringent enterprise applications to all members across all POPs.

Additionally, deploying Ethernet virtual private networks (VPNs), or EVPNs, can transform IXP networks by increasing security, scale, resiliency, availability and efficiency while retaining the benefits from existing VPN technology. It’s a game changer and will help meet the needs of the emerging enterprises.

Equally important, the latest in optical transport must be deployed to provide the foundation for all IP services. Optical Data Center Interconnect (DCI) will be a key use case, and looking forward, advances in optical networking will play an important part in meeting the enterprise challenge. New optical technology — such as higher speed pluggable digital coherent optics (DCO), the use of new spectrum bands and the introduction of new types of optical fiber such as hollow core — will all be required.

Multi-layer automation will also be required to drive both efficiency and accuracy within operations. Since IXP networking services will be customized to specific enterprise applications, automating tasks as simple as configuration is essential. However, perhaps of most importance is to be able to visualize both the optical transport and the IP routing network “as a whole”. This will allow the operator to proactively detect issues in advance, making their services more robust and available. As a foundation to automate both IP and optical elements, the network should be built with open interfaces and model-driven management architectures, while providing ease of integration and ease of use.​

Many of these same enterprise applications that involve interactions with client data from banking, insurance, healthcare and other industries with sensitive client information, will often require to abide by local data sovereignty rules. The localized nature of IXPs ensures that enterprise data does not have to leave its local boundaries. This positions IXPs to play an increasingly critical role as enterprises continue to distribute their compute to the edge and enterprise applications continue to proliferate and mature.

As a result, this most recent evolution of the IXP looks very different than its predecessor. While it shares some of the DNA, it is much more focused on adding value and sustainability than peering volumes of traffic cheaply. Still, security cannot be ignored and can be considered the next phase in its evolution.

IXPs as the first line of defense

Beyond issues with meeting SLAs and maintaining data sovereignty, it’s also critical that IXPs can do all this securely as these applications become business- and mission-critical for enterprise customers. DDoS (Distributed Denial of Service) attacks are increasing in frequency and sophistication. IXPs have been slow to respond to these threats, operating as neutral network intermediaries that move traffic from source to destination without interference.

However, as nerve centers for the AI-driven internet, IXPs are uniquely positioned to address DDoS attacks by leveraging their widely distributed IP routing fabric or data planes. Essentially, DDoS mitigation can be accomplished with sophisticated programmable filtering capabilities in the network processors. AI-driven analysis of the IXPs’ network and the internet can be used to understand from where attacks are originating. Since modern IXPs have distributed their advanced routing fabric with advanced traffic filters to all customer access points, they can conveniently use this fabric to do granular, network-based mitigation with little or no impact on services or customer traffic — without the need for expensive scrubbers.

Ultimately, these kinds of value-added services are changing the relationship IXPs have with their enterprise customers. As enterprises continue to evolve — particularly in adapting to the realities of AI in their market offerings — innovative and adaptive IXPs are in a unique position as the enterprise exchange of the future, not only to survive in this evolving environment, but to thrive.

ABOUT AUTHOR