Data Center Interconnection Network Effects Emerge, EQIX Moat Deepens

Overview

This report focuses on the data center interconnection market within US communications infrastructure. By building a proprietary database covering 645 key enterprise colocation facilities globally, it quantitatively analyzes the competitive landscape and economic value of the interconnection business. The core conclusion is that interconnection has evolved from a mere procurement criterion into a core growth engine in the AI era, characterized by significant network effects and high gross margins. Equinix (EQIX) possesses the widest moat due to its facility density, cloud on-ramp coverage, and layout in high-priced US markets; Digital Realty (DLR) and American Tower (AMT/CoreSite) also benefit from industry trends. The firm maintains 'Outperform' ratings for all three companies and believes the sector has further upside as AI releases demand for high-density interconnection.

Core views

The interconnection business is one of the segments with the best business models in the data center industry, typically boasting gross margins of 70%-90% and extremely high customer stickiness. Its core value lies in 'network effects': as the number of tenants in a campus increases, each new tenant becomes a potential connection target for existing tenants, causing interconnection revenue per cabinet to grow non-linearly while reducing churn through ecosystem lock-in. In the AI era, this effect is further amplified because large model training and inference require massive amounts of data to be transmitted with low latency between GPU clusters, storage, and cloud entry points, making private interconnection the only viable solution to avoid public internet bottlenecks. In terms of competitive landscape, Equinix (EQIX) demonstrates overwhelming advantages. Data shows that EQIX has 222 interconnection facilities, with an average of 58 partners connected per site, supporting 513,000 revenue-generating interconnection ports (as of Q1 2026). In contrast, second-ranked CoreSite (under AMT) has only 29 facilities with an average of 46 partners per site; although Digital Realty (DLR) ranks second in total networks, its average per site is only 34, reflecting its later transition to the enterprise segment. Regarding the key metric of cloud on-ramps, 60% of observed points are located in EQIX facilities, and among multi-hub facilities with more than 3 cloud on-ramps, EQIX accounts for 57%, while DLR holds only 23%. This ecosystem barrier is extremely difficult to replicate. Geographic structure and pricing power are also differentiating factors. EQIX's business portfolio is more skewed towards the US market, with an average cross-connect price of approximately $342/month, significantly higher than the ~$197 in other global regions. Although DLR has some extremely high-density nodes in Europe (e.g., the Frankfurt hub has 611 networks), its overall pricing is lower. Financially, EQIX's interconnection revenue reached $1.6 billion in 2025, with a CAGR of about 9%; DLR was $479 million, growing at about 8%; CoreSite has a smaller base but healthy growth of 13%. The report argues that AI not only increases rack demand but also boosts the number of ports and value per rack, benefiting top operators with neutral ecosystems and automated configuration capabilities.

Analysis framework

The report adopts an analytical framework of 'proprietary data mining + horizontal benchmarking + unit economics model'. First, instead of relying on vague industry estimates, the firm built its own proprietary database covering 645 key facilities globally, tracking over 21,000 interconnection relationships. This covers seven categories of partners including hyperscale cloud providers, tier-1 carriers, CDNs, ISPs, and emerging AI clouds (Neoclouds), thereby achieving a leap from qualitative description to quantitative empirical evidence. Secondly, the analysis logic follows the 'Structure-Conduct-Performance' paradigm: first defining the physical and virtual forms of interconnection and their evolution in the AI era (Structure), then quantifying the ecological niche differences of various operators through indicators such as facility density, cloud on-ramp share, and networks per site (Conduct), and finally deducing earnings quality and investment value by combining gross margins, revenue growth, and regional pricing (Performance). On the valuation front, consistent with REIT characteristics, the P/AFFO (Price to Adjusted Funds From Operations) multiple method is uniformly used, setting target prices anchored to expected 2027 AFFO to ensure comparability.

Methodology notes

• Competition and Strategy FrameworkMoat / competitive advantage

  Network Effects as Core Moat

  The report defines the interconnection business as a typical network effect model: the value of connections within a data center grows quadratically with the number of participants. This effect allows first-movers (like EQIX) to form ecosystem lock-in. Latecomers, even if providing equivalent power and space, cannot replicate the accumulated interconnection density, explaining why the interconnection business can maintain high gross margins of 70-90% in the long term.

• Industry/Sector Analysis FrameworkSupply and Demand Framework

  AI Changes Physical Characteristics of Data Center Traffic

  Traditional data center traffic was primarily north-south (in/out of the internet), whereas AI training has led to an explosion in east-west traffic (between rooms/clusters). Based on this, the report derives the 'data gravity' logic: to avoid public internet transmission bottlenecks, computing power must be deployed next to densely interconnected neutral hubs, which reconstructs the logic of data center location selection and premiums from the demand side.

• Valuation Method

  P/AFFO Valuation Method

  For data center REITs, the report does not use general PE or EBITDA multiples, but instead adopts Price to Adjusted Funds From Operations (AFFO). AFFO excludes non-cash items like depreciation and amortization and deducts maintenance capital expenditures. It better reflects the distributable cash flow creation capability of REITs than net income and is the standard anchor for pricing in this industry.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• Equinix (EQIX.US)

  Core Beneficiary: Possesses the densest interconnection network globally and the highest cloud on-ramp coverage, with its moat further deepening under AI drive

  Strengths

  Facility density and number of interconnection partners are far ahead; large exposure to high-priced US markets; monopolistic advantage in cloud on-ramps; steady growth in interconnection revenue

  Weaknesses

  Relatively high valuation (2027E P/AFFO 25x); significant year-to-date gains already realized

  Comparison

  Compared to DLR, interconnection density per site is 70% higher, and cloud on-ramp share is more than double; compared to AMT, interconnection revenue scale is an order of magnitude larger

  Risks

  Slowing demand for enterprise data centers; intensified competition in interconnection; pricing pressure due to data center supply surplus

• Digital Realty Trust (DLR.US)

  Key Participant: Transitioning from wholesale to enterprise-grade; high density in European nodes but overall interconnection ecosystem weaker than EQIX

  Strengths

  Extremely high density in core European nodes like Frankfurt; large total asset scale; valuation discount relative to EQIX

  Weaknesses

  Lower average number of interconnection partners per site (34); cloud on-ramp coverage far below EQIX; transition to enterprise-grade still takes time

  Comparison

  Ranks second in total networks, but per-site density is only about 60% of EQIX; interconnection revenue growth slightly lower than EQIX

  Risks

  <1MW segment growth missing expectations; slowing enterprise demand; price wars triggered by market share decline

• American Tower (AMT.US)

  Potential Candidate: High-quality CoreSite assets, fast interconnection growth but small volume

  Strengths

  CoreSite interconnection density per site is second only to EQIX (46); interconnection revenue growth leads the industry at 13%; tower business provides stable cash flow

  Weaknesses

  CoreSite overall scale is much smaller than EQIX and DLR; tower business faces interest rate sensitivity issues

  Comparison

  CoreSite density is close to EQIX but facility count is only 1/8th; overall valuation multiples are lower than pure data center REITs

  Risks

  Deterioration of major customer credit status; rising treasury yields suppressing valuations; D2D technology replacing terrestrial complementary facilities

Key data

• Number of EQIX Interconnection Facilities222Most globally, averaging 58 partners per site, supporting 513,000 revenue-generating interconnection ports
• Cloud On-Ramp Market Share60%60% of observed cloud on-ramps are located in EQIX facilities; 57% share in multi-cloud hubs
• Interconnection Business Gross Margin70-90%Significantly higher than traditional cabinet leasing business, with high stickiness
• US vs Global Average Cross-Connect Price$342 vs $197EQIX has greater exposure to the US market, enjoying higher pricing power
• EQIX 2027E AFFO/Share$48.63Target price of $1,222 based on 25x 2027 expected AFFO

Impact & implications

For the data center industry, interconnection capability has replaced pure power supply as the key variable distinguishing asset tiers. Operators with high-density interconnection ecosystems will gain excess returns in the AI wave, as they can not only charge higher connection fees but also improve overall occupancy and renewal rates through ecosystem stickiness. For investors, this means prioritizing allocation to targets with structural advantages in interconnection network effects (such as EQIX), rather than just looking at power reserve scale. Meanwhile, as Neoclouds (such as CoreWeave, Nebius) accelerate their entry into traditional interconnection hubs, the value of this ecosystem is being revalued by a new round of AI infrastructure cycles.

Risks

• Slowing growth in enterprise data center demand may lead to interconnection port growth falling short of expectations
• Intensified competition in interconnection services, especially challenges from emerging neutral interconnection platforms or cloud providers' self-built networks
• Continuous expansion of data center supply may weaken operator pricing power and compress interconnection business margins
• Macroeconomic fluctuations or changes in the interest rate environment affecting REIT valuations and customers' willingness to spend on IT

What to watch

• Quarterly interconnection revenue growth rates for each operator and changes in their proportion of total revenue
• Addition of new cloud on-ramps and Neocloud entry status
• Trends in average interconnection partners per site and cross-connect prices
• Actual deployment progress and bandwidth consumption of AI-related workloads in colocation facilities