Equinix, one of the world’s biggest data-center companies, said mounting barriers to bringing new projects online will make existing capacity more valuable and potentially strengthen its grip on the market — and pricing

Equinix customers anchoring data in vendor-neutral hubs cut data movement fees by up to 75%. The lever is geography: where data lives, how far it travels, and whether you're free to move it.

At Equinix Horizon 2026, 1️⃣ idea held the room: distributed AI only delivers when it connects well. Equinix Fabric One™ handles the routing and Equinix® Inference Exchange runs models close to users. Get the recap:

More snapshots from Equinix Horizon! 📸 We had so much fun showcasing demos, connecting with partners and customers and having conversations with industry leaders. ICYMI:

Proud moment for Equinix as we received the 2026 Mplify Alliance NaaS Service Innovation award! Recognition for driving innovation and delivering exceptional value to customers.

$EQIX KEY READ-THROUGHS FROM EQUINIX’S Q3 2026 EARNINGS CALL Equinix’s 10/28/26 call strengthens the case that the economic benefits of AI infrastructure spending are extending beyond hyperscale campuses into enterprise colocation, electrical equipment, cooling, and interconnected computing platforms. The most actionable implication is that constrained supply is supporting pricing on existing infrastructure while enterprise deployments broaden the demand base for new equipment. The principal negative implication is the growing distinction between infrastructure demand and investment returns: strong bookings do not eliminate construction, financing, utilization, or counterparty risk. The highest-conviction exposures favor established interconnection platforms and infrastructure suppliers over heavily financed computing operators. Several suppliers and peers had already reported strong results, so Equinix provides corroboration and greater visibility rather than an entirely new earnings signal. Country labels below refer to primary equity listing markets; impact assessments concern business fundamentals, not predicted share-price movements. DATA-CENTER OPERATORS: RETAIL PRICING POWER IS BROADENING BEYOND HYPERSCALE LEASING (READ-THROUGH 1) Affected companies: Digital Realty Trust (DLR: US), positive with moderate consolidated earnings relevance; American Tower (AMT: US), strongly positive for CoreSite but smaller at the consolidated level. The critical evidence is Equinix’s 12% increase in new-deal pricing, 35% growth in annualized gross bookings, 1.7% quarterly churn, and 7% colocation revenue growth within stabilized assets. Management described the market as being at “the very early stage of durable pricing strength going forward” and attributed part of the opportunity to industry capital expenditure being concentrated outside retail colocation. This combination supports a more durable pricing opportunity than a temporary shortage affecting a small number of hyperscale transactions. The transmission mechanism is renewal repricing and higher revenue from existing facilities. Established, network-dense campuses can increase realized pricing as contracts expire without incurring the full construction cost of a new campus. Interconnection raises switching costs because customers must preserve access to networks, cloud platforms, counterparties, and operational workflows. Equinix’s stabilized interconnection revenue growth of 10%, exceeding stabilized colocation growth of 7%, reinforces the value of these ecosystems. Digital Realty’s interconnection-oriented and smaller-deployment businesses and American Tower’s CoreSite platform have relevant exposure to this mechanism. The most actionable near-term catalyst is Digital Realty’s scheduled 10/29/26 earnings release. Equinix’s results strengthen the case for continued favorable renewal spreads, healthy smaller-deployment leasing, and conversion of signed capacity into revenue. However, Digital Realty already reported 25.4% cash renewal spreads in Q2 and guided to 9% to 11% for 2026. Equinix’s 12% new-deal pricing increase is not directly comparable with Digital Realty’s renewal spreads, and it does not justify extrapolating Digital Realty’s exceptional Q2 result. The incremental signal is broader pricing support across retail infrastructure, not a guaranteed repetition of unusually strong individual renewals. American Tower’s 10/27/26 results already showed CoreSite revenue growth of 17.4% and data-center cash revenue growth of 16.3%. Equinix therefore corroborates an already-reported trend rather than predicting an unreported American Tower earnings beat. CoreSite generated approximately $313 million of American Tower’s $2.785 billion quarterly revenue, limiting the consolidated impact relative to a more concentrated data-center investment. The longer-duration implication is favorable for established metropolitan interconnection campuses. Equinix expects approximately 80% of its 3-gigawatt development pipeline to be in its top 25 metros and emphasized that “Where a megawatt sits matters.” This supports a valuation distinction between connected, commercially established capacity and undifferentiated power availability. It does not establish that every data-center developer can reproduce Equinix’s pricing, churn, or returns. ELECTRICAL EQUIPMENT: THE SIGNAL IS BACKLOG EXECUTION AND REPEATABLE DEMAND, NOT NECESSARILY ANOTHER ORDER SURGE (READ-THROUGH 2) Affected companies: Eaton (ETN: US), positive with moderate consolidated significance and greater relevance to its electrical businesses; Schneider Electric (SU: France), positive with moderate consolidated significance and substantial relevance to data-center power infrastructure. Equinix expects 2026 retail cabinet deliveries to exceed its original plan by nearly 40%, including more than 13,000 cabinets in Q4. It has 48 projects underway across 32 markets. Management also stated that it had “pre-booked and pre-ordered large parts of our MCE footprint” and described standardized designs that allow equipment to be transferred between projects when construction schedules change. These disclosures support demand for switchgear, power distribution, uninterruptible power supplies, protection systems, and related electrical infrastructure. The important distinction is between ordering and commissioning. Equipment supporting imminent cabinet deliveries may already have been ordered, manufactured, or shipped. Consequently, the immediate read-through is stronger confidence in project completion and supplier backlog conversion, not an assumption that the 40% increase in Equinix’s delivery plan creates a matching new order increase for suppliers. Eaton’s next identifiable trading catalyst is its 11/03/26 earnings release. Equinix’s execution supports a constructive view of electrical backlog delivery and customer readiness, although it does not establish an Eaton-specific revenue beat. Eaton had already disclosed 43% year-over-year electrical-sector backlog growth in Q2. The incremental question is how effectively that backlog converts into revenue, operating profit, and cash as manufacturers expand capacity and customers complete projects. Schneider Electric had already reported Q3 results on 10/16/26, including approximately 19% organic growth in Energy Management and 22% organic growth in Systems, with data centers supporting electrical distribution, UPS, prefabricated solutions, and cooling. Equinix’s call therefore reinforces Schneider’s 2027 demand visibility rather than constituting a fresh Q3 surprise. The 20-year Georgia utility agreement adds a second transmission channel. Equinix committed to funding the substation and transmission infrastructure required by its campus and covering associated utility costs. This demonstrates that development can create funded demand beyond equipment inside the data center. However, the call did not identify the electrical equipment vendors or construction contractors for that agreement, so it does not support assigning a specific contract value to Eaton or Schneider. A relationship distinction is essential: Eaton was named on the call as a customer purchasing Equinix managed solutions and connectivity, not as an announced supplier for the expansion program. The positive supplier read-through rests on Eaton’s verified product exposure to the end market, not on that customer reference. Over the next 12 to 36 months, standardized designs, advance procurement, and expansion across established campuses should support repeat business and manufacturing utilization. The principal limitation is that stronger demand does not automatically imply higher supplier margins. Capacity additions, project mix, manufacturing costs, and competition determine how much of the infrastructure investment becomes incremental profit. THERMAL INFRASTRUCTURE: ENTERPRISE LIQUID COOLING IS BECOMING A SECOND DEMAND CHANNEL (READ-THROUGH 3) Affected companies: Vertiv Holdings (VRT: US), positive with potentially meaningful company-level relevance; Schneider Electric (SU: France), positive for cooling and integrated data-center systems, with a more diluted consolidated impact. The strongest evidence is Equinix’s disclosure that liquid-cooling bookings increased more than 10 times year over year. Incoming customers averaged 7 kilowatts per cabinet versus 5 kilowatts for departing customers. The company also described a 1.2-megawatt liquid-cooled deployment for Texas A&M Engineering Experiment Station and the launch of global innovation centers where enterprises can test liquid-cooled infrastructure before deploying it at scale. The transmission mechanism is higher thermal-management content and greater system complexity, rather than cabinet growth alone. Liquid-cooled deployments require combinations of coolant distribution, heat exchange, pumping, controls, commissioning, and ongoing service. Enterprise adoption broadens the addressable customer base beyond the largest AI training campuses and creates opportunities to sell integrated systems and lifecycle support. This is particularly relevant to Vertiv because of its concentrated exposure to critical digital infrastructure. Schneider benefits through its cooling portfolio and the ability to combine thermal equipment with electrical and prefabricated systems. The investment implication is that meaningful cooling demand can develop within existing enterprise colocation ecosystems, not only through a small number of exceptionally large hyperscale projects. The magnitude requires discipline. Equinix did not disclose the absolute value of liquid-cooling bookings, so growth exceeding 10 times could reflect a small starting base. The increase from 5 to 7 kilowatts per cabinet also does not mean that every incoming cabinet requires liquid cooling. The strongest conclusion is accelerating adoption and a richer infrastructure mix, not a 10-fold increase in the overall cooling market. Near-term novelty is limited because Vertiv reported earlier on 10/28/26, disclosing 61% sales growth, 55% organic growth, and organic order growth exceeding 100%. Schneider had also already reported strong data-center systems demand. Equinix adds customer-side confirmation that the equipment cycle is connected to actual deployments and enterprise adoption. It should not be represented as a leading indicator of supplier results already released. The longer-duration catalyst is conversion of enterprise testing and initial installations into repeat deployments during 2027 and beyond. Neither Vertiv nor Schneider was identified as the awarded liquid-cooling supplier for the specific Texas A&M installation, so no project-level revenue allocation is justified. Holding Vertiv, Eaton, and Schneider also retains substantial common exposure to data-center investment, despite their different product mixes. AI CLOUD OPERATORS: STRONG DEMAND DOES NOT REMOVE FINANCING AND RETURN-ON-CAPITAL RISK (READ-THROUGH 4) Affected companies: CoreWeave (CRWV: US), negative for relative financing resilience and equity-risk assessment, with potentially high sensitivity; Equinix (EQIX: US), comparatively positive as a more diversified infrastructure exposure. Management stated that the neocloud market contains more than 100 providers and that “there’s going to be winners and losers as this plays out.” Equinix emphasized strict credit protections and disclosed that the entire neocloud cohort represented less than 2% of monthly recurring revenue. Separately, management highlighted its 3.2% blended borrowing rate, predominantly fixed-rate debt, and average maturity exceeding 6 years as competitive advantages. The important read-through is a divergence in business-model risk, not evidence of weakening AI demand. A neutral interconnection platform can serve multiple competing compute providers while limiting dependence on any individual provider. A GPU-cloud operator must fund computing equipment and associated infrastructure, achieve utilization, meet service commitments, and absorb financing and depreciation costs. Strong customer demand improves the opportunity but does not eliminate the cash requirements preceding revenue collection. CoreWeave’s Q2 disclosure illustrates this distinction. It reported $2.575 billion of revenue and $1.510 billion of adjusted EBITDA, but also $640 million of net interest expense and only $128 million of adjusted operating income. Net interest expense was approximately 25% of revenue. These are dated Q2 figures rather than a current balance-sheet estimate, but they demonstrate why EBITDA growth alone is an incomplete measure of equity value creation. The counterevidence is equally important. CoreWeave reported approximately $104 billion of backlog, excluding more than $25 billion of additional commitments secured in early Q3, and disclosed substantial capital raising. Equinix’s call does not establish a CoreWeave customer default, financing failure, cancellation, or earnings miss. CoreWeave is the relevant listed exposure because of its financing-intensive operating model, not because Equinix identified a company-specific adverse event. The near-term catalyst is CoreWeave’s 11/10/26 earnings release. The economically important variables are deployment timing, funding cost, capital spending, utilization, and the conversion of contracted demand into cash earnings. Another increase in backlog would not, by itself, resolve those questions. The longer-duration negative is that competition among compute providers can remain intense while power, connectivity, and capital retain scarcity value. This can leave infrastructure owners with stronger bargaining positions than operators purchasing or leasing the infrastructure. That mechanism does not imply immediate repricing of existing contracts or justify applying Equinix’s 12% retail new-deal increase to CoreWeave’s costs. The positive offset is access to enterprise customers. Equinix added 31 neocloud network nodes, bringing the total to 150, and described providers increasingly targeting enterprise inference. That expands distribution opportunities. It does not establish that the resulting workloads will earn attractive returns after equipment, financing, and service costs. The high-conviction conclusion is a preference for financing resilience and demonstrated cash conversion, not a blanket negative view of AI infrastructure demand. AI SEMICONDUCTORS AND ENTERPRISE HARDWARE: MODEL CHOICE CAN EXPAND INFRASTRUCTURE DEMAND WITHOUT SELECTING A SINGLE HARDWARE WINNER (READ-THROUGH 5) Affected companies: NVIDIA (NVDA: US), positive for enterprise inference distribution; Dell Technologies (DELL: US), positive for integrated AI systems and services; Advanced Micro Devices (AMD: US), positive for accelerator deployment validation. The disclosed projects have limited immediate group-level financial significance; the stronger implication concerns the longer-duration demand structure. The call identified NVIDIA as the compute partner for Equinix Inference Exchange, with Together AI providing the initial inference platform and access to more than 200 open models. Separately, Equinix named Dell and AMD as partners in the 1.2-megawatt Texas A&M deployment. These are distinct commercial arrangements and should not be combined into a single technology stack or supplier allocation. For NVIDIA, the transmission mechanism is an additional route to enterprise adoption outside a single hyperscaler or proprietary model ecosystem. Enterprises can select among models while continuing to consume accelerated computing. Model flexibility therefore does not automatically undermine NVIDIA’s hardware opportunity. Equinix contributes connectivity, deployment locations, power, and access to enterprise customers, reducing infrastructure friction between experimentation and production. The limit is that the call did not disclose GPU volumes, hardware ownership, purchase commitments, revenue-sharing terms, or customer utilization. It supports a positive distribution and adoption thesis, not a quantified NVIDIA revenue increase. Lower token costs can also reflect efficiency improvements, so model usage and GPU demand should not be assumed to grow proportionately. For Dell and AMD, the evidence is more deployment-specific. Dell’s 07/24/26 project announcement identified a Texas A&M platform using AMD Instinct MI355X accelerators, with Dell supplying integrated systems and services and deployment at an Equinix facility in Dallas. The revenue mechanism includes server integration and associated services for Dell and accelerator content for AMD. The project demonstrates that meaningful AI infrastructure can be deployed through enterprise and research channels outside NVIDIA-only configurations. That is positive validation for AMD, but it is not evidence of broad NVIDIA share loss. The 1.2-megawatt project is too small to justify a material consolidated earnings revision for any of the hardware vendors. It was also announced before the Q3 call and should not be counted again as a newly awarded Q3 contract. Near-term, AMD’s 11/03/26 results provide a broader checkpoint on deployment conversion, while the Equinix announcements alone do not justify changing quarterly estimates for NVIDIA, Dell, or AMD. The more relevant commercial milestone is Inference Exchange’s planned Q1 2027 availability, following beta activity in Q4 2026. The longer-duration implication is that enterprise AI infrastructure spending can broaden across model providers, accelerator architectures, server integrators, and service channels. The strongest investment signal is diversification of deployment pathways, not a conclusion that all AI vendors benefit equally or that public-cloud spending must decline as private and colocated infrastructure grows.

Check out this demo from Ed Baichtal (@edbaichtal) at Equinix Horizon. Walk through an Equinix Distributed AI™ Hub solution with @PaloAltoNtwks. 👇 Explore more from Equinix Horizon: