How Big a Market for AI Data Center Interconnection?

Lumen Technologies has announced it will receive 10 percent of Corning’s total optical fiber production for a period of two years, to support additional capacity supporting “data center to data center” connections supporting artificial intelligence operations. 

The additional fiber is said to double Lumen route miles, but also support an order of magnitude expansion of “inside the data center” optical connections as well. Some might say the new move supports Lumen’s earlier move to emphasize its private connectivity fabric.

And while estimates vary, all observers would note that Lumen is a major provider of public optical networking services in the U.S. market. By some estimates, Lumen is the largest supplier. 

Optimists might say the new effort to create capacity connecting AI data centers is “new business.” Skeptics might say it is simply the evolution of the existing business. But Christopher Stansbury, CFO, says the new AI-related deals are “largely all incremental.” 

“This isn't cannibalization of legacy at all,” he notes. 

Research FirmAI "Data Center to Data Center" Interconnection Services (Capacity) ForecastOverall "Data Center to Data Center" Capacity Market ForecastGartner

The AI "data center to data center" interconnection services market is expected to grow from $4.2 billion in 2023 to $9.1 billion by 2027, a CAGR of 16.8 percent, according to analysts at Gartner.

Gartner also forecasts the overall "data center to data center" capacity market is forecast to grow from $48 billion in 2023 to $72 billion by 2027, a CAGR of 10.6 percent.

According to IDC, the AI-driven "data center to data center" interconnection services will account for 22 percent of the total "data center to data center" capacity market by 2025, up from 15 percent in 2023.

The global "data center to data center" capacity market is projected to reach $64 billion by 2025, growing at a CAGR of 12.7 percent from 2023, IDC also believes.

Forrester researchers believe the market for AI-enabled "data center to data center" interconnection services will grow at a CAGR of 19 percent from 2023 to 2027, reaching $8.5 billion by 2027.

Forrester also estimates the overall "data center to data center" capacity market will grow at a CAGR of 11 percent from 2023 to 2027, reaching $68 billion by 2027.

Markets and Markets predicts the AI "data center to data center" interconnection services market will grow from $4.6 billion in 2023 to $10.2 billion by 2027, at a CAGR of 17.3 percent .The global "data center to data center" capacity market is forecast to grow from $51 billion in 2023 to $78 billion by 2027, at a CAGR of 11.2 percent, the firm says. 

Either way, “data center to data center” connections are an important niche or segment of the capacity business, including sales to third party customers as well as captive capacity owned and operated by hyperscalers including Alphabet, Amazon and Meta, which consume much capacity to support their own operations. 

Though it is virtually impossible to know with great precision, most estimates of active traffic on global long-haul networks point to the importance of “data center to data center” connections as a percentage of total active connectivity. For example, broad consensus seems to exist that between 30 percent and perhaps as much as half of active capacity now is used to support connections between data centers and other data centers. 

By definition, that data center traffic is all enterprise or wholesale, while traffic between internet PoPs might be either retail or wholesale; consumer or enterprise. 

Category	Global Capacity
Data Center Interconnections	30-50%
Internet Points of Presence (PoPs)	30-40%
Enterprise Networks	10-15%
Other	5-10%

The contribution of AI data center interconnection to Lumen financial results remains to be seen. But it is an important segment of the connectivity business and among the largest segments.

Data Center Capacity to Double Over the Next Four Years

Amazon, Microsoft and Google now represent 60 percent of all hyperscale data center capacity, while doubling their capacity over the last four years, says Synergy Research. The firm expects total hyperscale data center capacity will double again in the next four years.

Synergy Research sees non-hyperscale colocation data centers accounting for 22 percent of worldwide capacity, with on-premise data centers rounding out 37 percent of the total. 

source: Synergy Research

In turn, that capacity drives demand for connectivity as well, as a huge portion of global long-haul capacity is used to connect data centers. Estimates peg the percentage of “data center to data center” capacity between 30 percent to 50 percent of total. Add in connections to internet points of presence and as much as 75 percent of total global backbone networking capacity connects data centers with other data centers and internet points of presence. 

Category	Global Capacity
Data Center Interconnections	30-50%
Internet Points of Presence (PoPs)	30-40%
Enterprise Networks	10-15%
Other	5-10%

The advent of artificial intelligence is likely to drive capacity trends. 

The AI "data center to data center" interconnection services market is expected to grow from $4.2 billion in 2023 to $9.1 billion by 2027, a CAGR of 16.8 percent, according to analysts at Gartner.

Gartner also forecasts the overall "data center to data center" capacity market is forecast to grow from $48 billion in 2023 to $72 billion by 2027, a CAGR of 10.6 percent.

According to IDC, the AI-driven "data center to data center" interconnection services will account for 22 percent of the total "data center to data center" capacity market by 2025, up from 15 percent in 2023.

The global "data center to data center" capacity market is projected to reach $64 billion by 2025, growing at a CAGR of 12.7 percent from 2023, IDC also believes.

Forrester researchers believe the market for AI-enabled "data center to data center" interconnection services will grow at a CAGR of 19 percent from 2023 to 2027, reaching $8.5 billion by 2027.

Forrester also estimates the overall "data center to data center" capacity market will grow at a CAGR of 11 percent from 2023 to 2027, reaching $68 billion by 2027.

Markets and Markets predicts the AI "data center to data center" interconnection services market will grow from $4.6 billion in 2023 to $10.2 billion by 2027, at a CAGR of 17.3 percent .The global "data center to data center" capacity market is forecast to grow from $51 billion in 2023 to $78 billion by 2027, at a CAGR of 11.2 percent, the firm says. 

The point is that if data center capacity keeps doubling every four years, then  “data center to data center” connections are going to grow as well. The issue is “how much” growth will be needed and how long the trend might last.

Why Cloud Data Centers Drive WAN Provider Revenue

source: Cisco

Traffic patterns are rapidly shifting from the traditional north-south client server model to an east-west server to server, said Jayshree Ullal, Arista president and CEO.

What that implies is a shift of data traffic. “We found that most of the server generated traffic in

the cloud data centers stays within a rack, while the opposite is truefor campus data centers,” say Theophilus Benson and Aditya Akella of the University of Wisconsin–Madison and David A. Maltz of Microsoft Research–Redmond.

And make no mistake, traffic generated by cloud computing increasingly dominates the pattern of global wide area network, metro and data center traffic. Data center traffic, as measured by Cisco,  includes data center-to-user traffic along with data center-to-data center traffic and traffic that remains within data centers.

Cisco’s latest Global Cloud Index estimates that global data center traffic will grow nearly 300 percent between 2013 and 2018.

By 2018, 76 percent of all data center traffic will come from the cloud, while 75 percent of data center workloads will be processed in the cloud.

But that might not even be the most significant prediction. Quantitatively, the impact of cloud computing on data center traffic is clear, Cisco argues.

Most Internet traffic has originated or terminated in a data center since 2008.

Where in the past most traffic (voice) functionally originated and terminate at a central office, though that traffic was passively transmitted to an end user telephone, now most global traffic originates and terminates at a data center.

And Cisco means that literally. Most traffic that ultimately traverses the WAN will follow communication within a data center between co-located servers.

The whole rest of the communications network essentially becomes a passive connection between an edge device requesting some operation, and the server that processes and serves up the result.

So in a meaningful sense, data centers are the central offices of the new network. As “switching” in the telco or mobile central offices might have been likened to the “brain” of voice operations, with the rest of the network the nervous system, so now data centers are the brains of the global network.

That will have any number of ramifications. The architecture of a voice network was build on the switching offices. Even the location of specific central offices was built on the amount of signal loss through a pair of copper wires connecting central offices with end user locations.

The architecture of new global networks is built on co-located servers inside data centers, and then connections between data centers over the wide area network, often with more traffic crossing the metro area than transiting the WAN, and then only traffic moving between a metro area data center and an end user location.

Although the amount of global traffic crossing the Internet and IP WAN networks is projected to reach 1.6 ZB per year by 2018,  the amount of annual global data center traffic in 2013 is already estimated to be 3.1 ZB.

The implications could not be clearer. By 2018, data center to end user traffic will constitute 17 percent of total. About nine percent of traffic will move from data center to data center. About 75 percent of global data center traffic will stay within the building, moving from server to server.

That explains the high interest by capacity providers in data centers. In the past, most of the revenue made by wide area network providers was supplying capacity across the wide area network.

In the future, it is likely much wide area network provider revenue will be made supporting data communications between servers and entities within data centers.

In other words, cloud data centers now are a major driver of WAN provider revenues. As potential owners of cloud data centers, WAN providers earn revenue by enabling data communications within the data centers, while also positioning to capture data center to data center revenues.

Data Center Traffic Drives WAN Capacity, But Most Traffic is Inside the Building

Global Internet traffic volume forecasts historically have been tricky, and on occasion have been wildly too optimistic.

Forecasts also are complicated by the new reality that so much Internet traffic is exchanged inside data centers, and does not cross the wide area network. That has implications for any company investing in, operating or selling wide area network infrastructure and services.

A further complication is that “data center traffic” often includes both traffic that crosses wide area networks and traffic that is exchanged solely within a data center.

The fact of note is that traffic carried inside a data center exceeds global wide area network traffic volume. The implication is that forecasters will have to be careful when dealing with concepts such as “data center IP traffic.”

Some of that traffic will drive demand for local access, metro and long-haul capacity. Some will not.

The analogy is to today’s data networks, which include WAN and local access connections and revenue, as well as local area network data transfers (Wi-Fi, for example) that do not represent revenue for service providers.

And inside-the-data-center traffic is a huge deal.

By 2018, as much as 75 percent of all IP traffic will occur within data centers, and will not cross the WAN, about even with the 2013 estimate that 77 percent of total IP traffic actually remains within a data center, Cisco predicts.

Although the amount of global traffic crossing the Internet and IP WAN networks is projected to reach 1.6 ZB per year, by 2018, the amount of annual global data center traffic in 2013 is already estimated to be 3.1 ZB, and by 2018, will triple to reach 8.6 ZB per year, according to Cisco.

Traffic between data centers is growing faster than either traffic to end-users or traffic within the data center, though, and by 2018, traffic between data centers will account for almost nine percent of total data center traffic, up from nearly seven percent at the end of 2013.

The high growth of this segment is due to the increasing prevalence of content distribution networks, the proliferation of cloud services and the need to shuttle data between clouds, and the growing volume of data that needs to be replicated across data centers.

Over the next five years, the study projects data center traffic to nearly triple, with cloud representing 76 percent of total data center traffic.

In 2013, cloud accounted for 54 percent of total data center traffic. By 2018, cloud will account for 76 percent of total data center traffic.

(Data center traffic includes data center-to-user traffic along with data center-to-data center traffic and traffic that remains within data centers.)

The study predicts that global data center traffic will nearly triple from 2013 to 2018 with a combined annual growth rate of 23 percent.

A few clear observations can be made about what drives global Internet traffic. It has been clear for some time that content bits--especially entertainment video--is the primary driver of volume across global and access networks.

It has been clear for some time that consumer-facing Internet apps likewise drive traffic, not enterprise traffic.

Most Internet traffic has originated or terminated in a data center since 2008, according to Cisco.

Now there is one more assumption we can make: data center traffic will continue to dominate Internet traffic for the foreseeable future, and will be driven by cloud applications, services, and infrastructure.

By 2018 76 percent of data center traffic, will be cloud traffic, Cisco now says.

source: Cisco Global Cloud Index

U.S. Telcos Exiting Data Center Business?

Several developments confirm a shift of strategy by large U.S. telcos and the apparently-secure role of specialists in the data center business.

Digital Realty Trust, the second-largest data center REIT after Equinix, is adding more capacity, bringing online a two-million-square-foot property in Ashburn, Va.

At the same time, the large U.S. telcos either have concluded that data center ownership no longer is a “core” asset, and have sold, are trying to sell, or might sell, their data center portfolios.

It is a truism that communications service providers are in a race to replace declining legacy revenues with earnings based on next generation services that largely center, one way or the other, on Internet apps, services and devices.

Along the way, some approaches that seemed appealing in the past are being revised. For the most part, few large telcos, cable TV or other service providers have had much luck creating big new businesses in consumer over the top services, even if that effort continues.

Most large U.S. service providers, for example, also seem to be concluding that data center operations are not “core assets.” They understand that data center services must be bundled with other enterprise and business customer offers.

But they also seem to be concluding that capital is better deployed elsewhere, and that data center services can be supplied on a “lease rather than own” basis.

Verizon might be the first large U.S. service provider to conclude that its extensive global network assets might also be “not core assets.”

Verizon Communications now is said to be exploring a sale of its “global enterprise assets” built on the former MCI-Worldcom global network plus the data center business (Terremark).

Verizon Chief Financial Officer Fran Shammo said, during the company's third-quarter earnings call on Oct. 20, 2015, that it continues "to work through secular and economic challenges" with its global enterprise division, which posted a 4.9 percent decline in revenue in the quarter ended Sept. 30, 2015.

In many ways, that trend speaks to changes in the ways enterprises globally have changed “what” they buy in the connectivity arena, as well as “from whom.”

There is declining demand for legacy connectivity of every sort, and growing demand for IP connections and bandwidth. Need for data center support also is changing as cloud computing becomes mainstream and public computing resources from Amazon Web Services and others become viable options.

Verizon spent $8.4 billion to acquire MCI-Worldcom in 2006. Verizon bought Terremark Worldwide in 2011 for $1.4 billion. Altogether, that represented $9.8 billion in acquisition costs.

Some think the assets could be sold for $10 billion. Some estimate the assets produce about $2 billion in annual revenue, implying a sale price about five times annual revenue.

Of course, Verizon has a revenue profile distinct from other large U.S. service providers, as 85 percent of its total revenue is earned from its U.S. mobile segment. All consumer operations and fixed network enterprise together represent just 15 percent of revenue.

Still, the trend to divest data center assets, mostly acquired over the past decade, suggests a belief on the part of large service providers that sustainable advantage cannot be gained by owning such assets.

If so, we might also conclude that large U.S. telcos no longer see the data center business as a path to future growth, in the same way that most large telcos have had scant success creating large new businesses in the consumer over-the-top domain.

In addition to the possible sale of Verizon data center assets, AT&T and CenturyLink likewise are trying to sell their data center businesses.

Windstream already has divested its data center business.  

So what is going on? There appears to be nothing fundamentally “broken” with the data center businesses. But most larger U.S. telcos seem to believe they can obtain the benefits (services for their enterprise customers) without owning the facilities.

CenturyLink appeared to believe colocation center hosting would provide a boost several years ago when it bought Savvis.

While it plans to continue offering colocation services, CenturyLink says it is looking for alternatives to owning nearly 60 data centers around the world that support colocation, managed hosting, and cloud services.

Cincinnati Bell also is monetizing its data center assets, selling ownership shares of its CyrusOne data center business and raising cash to reduce debt.

CenturyLink business segment revenues might be “driven principally by increased market penetration of our network, hosting, cloud, and IT solution service offerings,” as CEO Glen Post said.

But capital might be better deployed elsewhere, CenturyLink suggests. “We expect colocation services will continue to be a service our customers will look for us for, but we do not necessarily believe we have to own the data center assets to be effective in delivery of those services,” said Post.

CenturyLink’s revenue for the cloud and hosting  business is about $600 million annually, and the company says it seeks to sell or otherwise restructure the data center operations, not the cloud and hosting businesses that use data center real estate.

Experimentation is to be expected. Large service providers must expect to replace half their existing revenue sources about every decade, and the diminution of all the legacy core services implies a future revenue base build nearly completely on “new” services.

Finding big new revenue sources will be complicated by changing value and roles within what we used to call the “communications” ecosystem. Simply, value and revenue growth are migrating to cloud apps and software products, and away from managed communication services.

One is reminded of moves many large telcos made, in past decades, in the “computing” industry, again without long-term success.

That is just part of the continuing evolution of the business. Many efforts will prove disappointing. And it appears we now can add “owning data centers” to the avenues that have proven disappointing. No matter. The search will continue.

Data Center Architecture Now Hinges on Connectivity Foundation

Cloud computing requries data centers. Data centers require communications. Communications requires computing and data centers. That means each industry potentially can assume additional roles inm in the other parts of the ecosystem. That, in turn, perennially spurs hopes that connectivity providers can operate in the cloud computing or data center roles.

So Orange now talks about a strategy that builds on connectivity but adds other digital roles beyond connectivity. Other "telcos" have similar hopes to shift from being "telcos" to becoming "techcos." It never is completely clear what that means, but the implication always is that new lines of business and revenue are the result.

Hopes aside, connectivity services now have become an essential foundation for cloud computing and data center value.

Some adaptations of data center architecture in the cloud computing era have to do with connectivity architecture more than the way servers are cooled, the types of processors and specialized chips used in servers or the way servers are clustered. 

A cluster is a large unit of deployment, involving hundreds of server cabinets with top of rack (TOR) switches aggregated on a set of cluster switches. Meta has rearchitected its switch and server network into “pods,” in an effort to recast the entire data center as one fabric. 

source: Meta 

The point is to create a modularized compute fabric able to function with lower-cost switches. That, in turn, means the fabric is not limited by the port density of the switches. It also enables lower-cost deployment as efficient standard units are the building blocks. 

That also leads to simpler management tasks and operational supervision as well. 

And though  much of that architecture refers to “north-south” communications within any single data center, the value of the north-south architecture also hinges on the robustness of the east-west connections between data centers and the rest of the internet ecosystem. 

source: Meta 

In fact, it is virtually impossible to describe a data center architecture without reference to wide area and local area connectivity, using either the older three-tier or newer spine-leaf models. The point is that data centers require connectivity as a fundamental part of the computing architecture. 

source: Ultimate Kronos Group 

That will likely be amplified as data centers move to support machine learning operations as well as higher-order artificial intelligence operations. 

Still, in the cloud computing era, no data center has much value unless it is connected by high-bandwidth optical fiber links to other data centers and internet access and transport networks. 

“From a connectivity perspective, these networks are heavily meshed fiber infrastructures to ensure that no one server is more than two network hops from each other,” says Corning.

The other change is a shift in importance of north-south (inside the data center) and east-west (across the cloud) connections and data movement. As important as north-south intra-data-center traffic remains, communications across wide area networks assumes new importance in the AI era. 

The traditional core-spine-leaf connectivity architecture between switches and centers increasingly looks to be replaced by a two-layer spine-leaf design that reduces latency. In other words, in addition to what happens inside the data center, there will be changes outside the data center, in the connectivity network design. 

source: Commscope

In the above illustration, the Entrance Room (ER) is the entrance facility to the data center.

The Main Distribution Area (MDA) holds equipment such as routers, LAN and SAN switches. The Zone Distribution Area (ZDA) is a consolidation point for all the data center network cabling and switches.

The Equipment Distribution Area (EDA) is the main server area where the racks and cabinets are located.

 

So even if connectivity remains a separate business from cloud computing and the data center business. But all are part of a single ecosystem these days. Cloud computing requires data centers and data centers require good connectivity.