Data Centers

The filing introduces dividend restrictions tied to preferred stock status: no junior stock dividends or repurchases may occur unless accumulated preferred dividends are paid in full or reserved. This is a standard preferred-stock-related risk, not specific to data center operations.

“so long as any share of Preferred Stock of such series remains outstanding, no dividend or distribution may be declared or paid on the Class A Common Stock, Class C Capital Stock or any other shares of junior stock”

No new risk factors are explicitly introduced in this filing relative to prior periods. Currency risk in display (yen exposure) is discussed but is being managed through hedges through 2030. Hyperscaler concentration is implicit in the Meta and similar customer agreements but is not flagged as a new risk.

One material new risk flagged: 800-volt DC architecture adoption risk, though CEO frames it as net positive. Language shifted from 'monitoring' to 'we have made a number of analyses'—indicating active risk assessment. CEO estimates 20% of Legrand sales (primarily rack PDUs, UPS) will face neutral-to-negative impact; 80% neutral-to-positive. Grid access delays re-characterized as material (2-3 year lead times for connection) but positioned as industry-wide, not Legrand-specific. Tariff risk remains active but already absorbed once; management signals willingness to re-price if raw mats exceed central case. No new geopolitical, supply chain, or customer concentration risks introduced.

“It will have a neutral to negative impact on about 20% of Legrand sales and a neutral to positive impact on 80% of Legrand sales.”

The Item 1A disclosure states 'The risk factors discussed under the heading Risk Factors and elsewhere in the Company's and the Operating Partnership's Annual Report on Form 10-K for the year ended December 31, 2025 continue to apply to our business.' No new risk factors are introduced in this Q1 2026 10-Q excerpt. The forward-looking statement boilerplate reiterates prior risks (tariffs, supply chain, inflation, geopolitical turmoil) without adding novel risk language.

“The risk factors discussed under the heading "Risk Factors" and elsewhere in the Company's and the Operating Partnership's Annual Report on Form 10-K for the year ended December 31, 2025 continue to apply to our business.”

Memory price volatility is introduced as a new risk factor affecting Windows OEM and on-premises server demand, with potential spillover into capital expenditures and cloud gross margins. Amy states: 'We mentioned the potential impact on Windows OEM and on-premises server markets, from increased memory pricing earlier. In addition, rising memory prices would impact capital expenditures, though the impact on Microsoft cloud gross margins will build more gradually. As these assets depreciate over six years.' Additionally, the durability and concentration risk of the OpenAI contract (45% of commercial RPO) is raised by an analyst and addressed by management, though management reframes it as a strength (diversified $350B non-OpenAI RPO growing 28% YoY).

“Rising memory prices would impact capital expenditures, though the impact on Microsoft cloud gross margins will build more gradually as these assets depreciate over six years.”

No explicit forward guidance on operating metrics, revenue, capex, or financial targets is provided in this 8-K. The document is transactional, not guidance-bearing.

Guidance is highly confident and committed. Q1 2026 core sales expected to grow 15% year-over-year to $4.2B–$4.3B range, with EPS growing 26% to $0.66–$0.70 range. Full-year 2026 CapEx planned at $1.7B. Management upgraded both internal SpringBoard plan ($6.5B incremental sales by 2026, up from $6B) and high-confidence plan ($5.75B, up from $4B), closing the gap between plans due to increased visibility and customer commitments. Operating margin target remains at 20% or above; management expects margins could exceed 20% but prioritizes growth and return on invested capital over margin expansion.

“For the first quarter, we expect year-over-year growth to accelerate, with core sales up approximately 15% to a range of $4.2 billion to $4.3 billion.”

Guidance hedge level: confident for 2026 data center (10-20% organic growth), cautious for buildings (flat-to-slightly-positive Europe, slightly negative U.S.). For 2030 ambitions, language upgraded from 'around 20%' margin to 'above 20%' and 'confident in its ability to reach the upper end' of EUR 15B sales target. Exact phrasing: 'we are very confident' (data center); 'we start to be somehow a bit more positive' (France/Europe); 'I don't want to sound too optimistic because we've been waiting for the rebound for 2 years' (European buildings).

“Legrand is targeting the following in 2026, sales growth, excluding currency effects of between plus 10% and plus 15%...we are very confident on our ability to do something between plus 10% and plus 20% in data center.”

Capital expenditure guidance is cautious with optionality: 'For the remainder of 2026, we currently expect to incur approximately $2.8 billion to $3.3 billion of capital expenditures... This amount could go up or down, potentially materially, based on numerous factors.' Development stabilized yield is confident at 11.4% on 1,169 MW but qualified: 'No assurance can be given that we will complete any of these projects on the terms currently contemplated.' Rental rate expectations are confident on 2026 renewals but qualified by market supply: 'we expect average aggregate rental rates... to be positive... subject to the supply of available data center capacity.' Overall hedge level: cautious.

“For the remainder of 2026, we currently expect to incur approximately $2.8 billion to $3.3 billion of capital expenditures... This amount could go up or down, potentially materially.”

Guidance is classified as confident with specific quantified ranges. Q3 guidance: revenue $80.65-$81.75B (15-17% growth), COGS $26.65-$26.85B (22-23% growth), OpEx $17.8-$17.9B (10-11% growth). Azure growth guidance 37-38% YoY in Q3 constant currency. Windows OEM decline 'roughly 10%' in Q3, with 'wider than normal' range of outcomes due to memory pricing. FX effects quantified at +3 points total revenue growth. Management uses phrases like 'we expect,' 'should,' and 'will continue to' indicating committed guidance, though contingencies around capacity delivery timing and revenue recognition are acknowledged.

“We expect revenue of $80.65 to $81.75 billion or growth of 15 to 17%. Demand continues to exceed available supply. Operating margins should be down slightly year over year.”

Alphabet committed $13.3 billion to mandatory convertible preferred stock with automatic conversion into common equity in May 2029, locking in 6.25% annual dividend costs and signaling capital allocation toward sustained infrastructure investment over the next three years without immediate equity dilution.

“167,500,000 series A depositary shares and 167,500,000 series B depositary shares, each representing a 1/20th interest in a share of 6.25% Mandatory Convertible Preferred Stock, liquidation preference $1,000.00 per share”

Reid Hoffman, a director since 2017, will not stand for re-election at the 2026 annual shareholder meeting. His departure is explicitly unrelated to any operational or strategic disagreement with management.

“His decision not to stand for re-election is not as a result of any disagreement with management on any matter relating to the Company's operations, policies, or practices.”

Corning announced a $6B multiyear agreement with Meta to supply optical fiber, cable, and connectivity solutions for GenAI data centers, with manufacturing and capacity expansion in North Carolina. This is the anchor contract for an upgraded SpringBoard plan that now targets $11B in incremental annualized sales by 2028 (versus original $8B), driven by similar agreements with other major hyperscaler customers. This signals a fundamental shift in optical communications from cyclical to secular growth, formalized by customer prepayments and take-or-pay agreements. It also represents the first major customer agreement of this scale and structure in optical, validating the strategy proven in display (Gen 10.5) and automotive (Apple cover glass) segments.

“Corning Incorporated and Meta announced a multiyear, up to $6 billion agreement to support Meta's apps, technologies, and AI ambitions using our newest innovations in optical fiber, cable, and connectivity solutions.”

Most important new statement: Legrand has elevated its 2030 margin ambition from 'around 20%' (prior CMD) to 'above 20%' while maintaining confidence in reaching upper end of EUR 15B sales target—representing a shift from defensive to offensive margin posture. This is grounded in data center's contribution rising from 14% of sales (EUR 0.7B in 2020) to 26% (EUR 2.4B in 2025), a 3.4x expansion in 5 years, with margin accretion embedded in 2026-2030 plan. Secondary: Accelsius investment signals active portfolio gap-filling in 2-phase direct-to-chip cooling, marking transition from 'evaluating' to 'investing' in next-generation cooling architectures.

“Legrand is confident in its ability to reach the upper end of its 2030 sales target range around EUR 15 billion...average adjusted operating margin above 20% of sales compared with the previously targeted level of around 20%.”

Shareholder voting reveals institutional resistance to climate and labor governance pressure on data center operations. A proposal requesting additional reporting on data centers' impact on climate commitments received only 18.3% support, while a worker-oriented AI advisory council proposal received 0.0006% support, indicating shareholders view management's current governance as adequate.

“A shareholder proposal requesting additional reporting on impact of data centers on climate commitments was not approved”

Demand for Corning's GenAI optical products is unprecedented. Enterprise business grew 61% in 2025, with hyperscale data center segment growing significantly faster. Management states if they could make more of their new high-density products, they could sell more. Meta agreement ($6B multiyear) and similar agreements with other major customers signal formalized long-term commitments to capacity, with manufacturing expansion in North Carolina dedicating capacity specifically to these customers. Booking timelines are tightening: Corning is dedicating capacity through long-term prepayment and take-or-pay agreements.

“If we could make more of these new products, we could sell more. And it is for those types of products that we are dedicating this capacity through these agreements.”

Management cites unprecedented AI-driven data center demand with language shifted from exploratory to contracting and committed. Data center sales grew 40% organically in 2025 versus 7% in buildings; 2026 guidance assumes 10-20% data center growth. Hyperscaler CapEx plans are explicitly referenced as validation—CEO states CapEx growth of 50-100% between 2025-2026 from hyperscalers. Backlog language is strong ('very strong,' 'very good') but CEO cautions against extrapolating orders to sales given 8-12 week lead times. No specific capacity additions in MW cited; instead, management discusses doubling capacity investment YoY while maintaining 3-3.5% CapEx-to-sales ratio. Pre-leased/sold-out language not used; instead 'above 1' book-to-bill mentioned.

“The lesson of last year is that it's difficult to anticipate...we did plus 40% or close to plus 40%...based on those information, we are very confident on our ability to do something between plus 10% and plus 20%.”

Management language on AI-driven demand has shifted from 'strong demand' (2025 baseline) to explicit capacity acceleration metrics: 1,169 MW under development as of March 31, 2026 represents a 52% sequential increase vs. Dec 31, 2025. Pre-leasing rate at 61% on in-flight projects is a hard signal of customer commitments. Management expects 'average aggregate rental rates on renewed data center leases for 2026 expirations to be positive.' No new hyperscaler customers named, but booking timelines implied to be tight (high pre-lease percentage signals advanced planning).

“As of March 31, 2026, we had 1,169 megawatts of projects underway across multiple metropolitan areas around the world, representing an approximately 52% increase of capacity under development as compared to December 31, 2025. As of March 31, 2026, 61% of the 1,169 megawatts of projects underway was pre-leased.”

Management cites unprecedented, accelerating AI demand with customer demand exceeding available supply. Nearly one gigawatt of capacity was added in Q2 alone. Forward guidance for Azure is 37-38% revenue growth in Q3 YoY, with demand continuing to exceed supply. Management emphasizes that capacity delivery is multiyear and geographically distributed; Atlanta and Wisconsin are referenced as multiyear delivery projects, not immediate capacity sources. Over 250 customers are on track to process over 1 trillion tokens on Foundry this year. No 'sold out' or 'pre-leased' language is used; instead, demand-supply imbalance is the controlling constraint.

“All up, we added nearly one gigawatt of total capacity this quarter alone. Customer demand continues to exceed our supply.”

AI-driven demand is characterized as accelerating and unprecedented. Trailing 12-month organic orders grew 81%, with Q4 up 252% YoY. Book-to-bill stands at 2.9x. Orders are becoming larger and arriving with 12-18 month delivery windows. Management cites broad-based strength across all regions and all product technologies, with pipeline growing despite Q4's massive order intake. No language of capacity being sold out; emphasis is on 'win and scale.'

“Trailing 12-month organic orders grew 81%. Fourth quarter orders were up more than 250%. Book-to-bill almost 3x. Very strong performance.”

Liquid cooling is not mentioned. High-density optical architecture is central to the Meta agreement and optical communications growth. Corning's new GenAI products deliver superior optical performance in half the space with significantly reduced installation cost. Management references 'new high-density products in fiber, in cable, and in connectivity' as the focus of capacity expansion. CPO (co-packaged optics) is mentioned as a future opportunity not yet in SpringBoard targets. Scale-up (photonics closer to GPUs and switches) is cited as a major long-term opportunity not modeled in the 2028 plan due to timing uncertainty.

“These innovations enable our customers to have better and more reliable optical performance in about half the space with significantly reduced installation cost.”

Liquid cooling explicitly highlighted as core growth vector. Language has shifted from pilot to core portfolio: 2-phase direct-to-chip cooling now represents 5% of data center sales and is being actively built through Accelsius investment (equity + commercial/tech partnership). Rack density thresholds cited: current architecture at $2-3M per megawatt (closer to $3M); 800-volt DC architecture expected to reach $3-4M per megawatt, implying higher kW/rack. No specific kW/rack figures stated. GPU/AI accelerator infrastructure not explicitly mentioned; instead 'compute management' and 'compute infrastructure' used to describe monitoring PDUs, KVMs, transceivers representing 50% of DC sales. Named partnerships: Google, Meta, Oracle, Microsoft, QTS, Equinix; Accelsius investment.

“We recently invested in Accelsius in the U.S., a pioneer in 2-phase direct-to-chip liquid cooling. This investment will further strengthen the group's portfolio of solutions for AI and HPC data centers.”

Liquid cooling is not mentioned in this Q1 2026 filing. Rack density thresholds and GPU/AI accelerator partnerships are not discussed. Technology architecture specifics are absent, suggesting either stability in prior architecture decisions or no material changes to report.

Liquid cooling is explicitly mentioned as enabling Fairwater's two-story design for higher GPU density. No language shift from 'pilot' to 'default' is stated; liquid cooling appears as an implemented feature at select sites (Fairwater). Rack density improvements are implicit in the claim of '50% increase in throughput' for OpenAI inferencing and the optimization focus on 'tokens per watt per dollar.' GPU accelerators (NVIDIA, AMD, and Microsoft's Maya) and custom chips (Maya 200, Cobalt 200) are detailed. Maya 200 delivers '10 plus flops at FP4 precision with over 30% improved TCO compared to the latest generation hardware.' Cobalt 200 delivers 'over 50% higher performance compared to our first custom-built processor for cloud-native workloads.' No named partnerships with chip vendors beyond NVIDIA and AMD are stated.

“Fairwater's two-story design and liquid cooling allow us to run higher GPU density and thereby improve both performance and latencies for high-scale training.”

Liquid cooling has moved decisively from pilot to deployed-at-scale status. SmartRun (prefab white-space solution) is 'being deployed across several large customers at scale' and now specifically referenced with Compass Data Centers. PurgeRite acquisition signals Vertiv's commitment to fluid management as core competitive moat. Rack density thresholds and GPU architecture specifics are not disclosed; emphasis is on 'hundreds of kilowatts per rack' and system-level integration. Liquid cooling language has shifted from exploratory to operational/commercial, indicating threshold crossing.

“SmartRun is being deployed across several large customers at scale. It is flexible and scalable across multiple generations of silicon.”

No transformer, switchgear, or lead-time constraints are mentioned. CapEx is planned at $1.7B for 2026 (up from under $1B in 2025) to support capacity expansion. Labor and construction costs are not flagged as constraints. The supply chain strategy focuses on customer-funded capacity through prepayments and long-term commitments, mirroring the Gen 10.5 display glass model. Optical products are the primary focus of CapEx investment.

“We plan to spend about $1.7 billion in CapEx. For reference, we spent a little under $1 billion this year.”

Supply chain constraints mentioned only indirectly through tariff absorption. U.S. tariff impact was $140M actual (vs. $30M initially embedded), with $100M compensated through pricing and $40M through supply chain optimization (USMCA eligibility improvements). No transformer or switchgear lead times mentioned; CEO states lead times for most products are 8-12 weeks. Tariff exposure on electrical equipment flagged for 2026 with embedded assumption of $1-2% raw material/component inflation. If raw mats exceed central scenario, pricing will adjust upward. No construction cost or labor availability constraints flagged.

“We had $140 million. We compensated $40 million by optimizing our supply chain...the remaining $100 million of tariff were compensated through pricing.”

Digital Realty flags tariffs and supply chain disruptions as forward-looking risks ('increased tariffs, global supply chain or procurement disruptions, or increased supply chain costs') but does not cite specific transformer or switchgear lead times, product categories, or geographies. Construction cost and labor availability are mentioned generically in the risk section; the MD&A notes 'increases in building operations expense, common area maintenance expense and data center labor' driving stabilized opex +$21.9M YoY, but no constraint language appears.

“increased tariffs, global supply chain or procurement disruptions, or increased supply chain costs”

Transformer or switchgear lead times are not mentioned. Tariff exposure on electrical equipment is not flagged. Construction costs and labor availability are not cited as constraints. Amy Hood mentions memory price increases as a source of volatility in Windows OEM and on-premises server purchasing patterns, and notes that 'rising memory prices would impact capital expenditures, though the impact on Microsoft cloud gross margins will build more gradually as these assets depreciate over six years.' This is a downstream cost impact, not a supply chain lead-time constraint.

“We mentioned the potential impact on Windows OEM and on-premises server markets from increased memory pricing. Rising memory prices would impact capital expenditures.”

Lead time expansion is acknowledged on 'some product lines' but positioned as aligned with customer demand (12-18 month delivery cycles), not supply constraint. No specific transformer or switchgear lead times cited. Tariff exposure is flagged: management expects 'materially offset unfavorable margin impact from tariffs' by Q1 2026 on an exit rate basis. CapEx expansion is being deployed to unlock capacity; no labor availability constraints mentioned. Supplier relationships are described as 'extensions of our operations,' with coordinated scaling underway.

“We expect on an exit rate basis, to have materially offset unfavorable margin impact from tariffs as of the first quarter of this year.”

Corning is consolidating U.S. optical manufacturing in North Carolina, anchored by the Meta agreement and similar hyperscaler commitments. Management states about 60% of sales are outside the U.S. and 40% inside, with the expectation that this ratio will continue. However, factory location will follow customer location: if more AI buildout occurs in the West, more U.S. manufacturing; if automotive and glass growth occurs in Asia, manufacturing will shift eastward. The company is developing new markets for existing capabilities (e.g., Gorilla Glass from display into automotive). Geographic diversification is stable, not intensifying.

“We today are about 60% outside the U.S. and about 40% in, and we would expect something in that zone to continue. But what will really drive the location of our factories will tend to be where our customers are.”

Market saturation and density are the primary geography drivers, not expansion into new markets. In 2025: Europe +1.9% (Germany, Italy, Netherlands, UK healthy); North America +16% (driven by data centers); Rest of World +2.7% (Asia Pacific, Africa, Middle East good; South America retreating). For 2026, management expects flat-to-slightly-positive building in Europe (residential completions moving -9% to +2%; permits +4%; renovation +1%) and slightly negative in U.S. buildings. China remains depressed (residential -50% over 3-4 years) but represents only 2% of sales. Data center growth is U.S.-dominant at ~50% of group data center growth; Europe and Rest of World at ~20% each. Geographic diversification not intensifying; instead, reliance on hyperscaler footprint.

“In Europe, in a market that remains mixed overall, sales were up plus 1.9%...In North and Central America, sales were up a strong plus 16%.”

Northern Virginia dominates annualized rent at 20.0% of portfolio; Frankfurt and Chicago tie at 6.4% each. Management notes 'strong demand in most of our key metropolitan areas' but emphasizes constraints: 'subject to the supply of available data center capacity in these metropolitan areas.' New development is concentrated in existing markets (Northern Virginia, Paris, Frankfurt cited as biggest contributors to non-stabilized revenue growth). Acquisition activity emphasizes existing footprint strengthening (Sofia, Milan, Portland, Atlanta in Q1 2026) rather than greenfield expansion. Geographic diversification language is stable, not intensifying.

“We continue to see strong demand in most of our key metropolitan areas for data center capacity and, subject to the supply of available data center capacity in these metropolitan areas, we expect average aggregate rental rates on renewed data center leases for 2026 expirations to be positive.”

Northern Virginia, Dublin, and Singapore are not explicitly named as constrained markets. Seven countries are cited for new DC investments to support sovereignty and local data residency; these are selected based on customer control and regulatory requirements, not power or land availability. Geographic diversification language is intensifying: Satya states 'We are expanding our solutions and global footprint to match' sovereignty requirements and notes 'We offer the most comprehensive set of sovereignty solutions across public, private, and national partner clouds.' Amy notes that capacity additions will occur 'globally' and 'A lot of that will be added in The United States. To locations you've mentioned, but it also needs to be added across the globe to meet the customer demand that we're seeing and the increased usage.'

“We announced DC investments in seven countries this quarter alone supporting local data residency needs. We're working incredibly hard adding capacity globally.”

Americas is the primary growth engine, guided to high 30s organic sales growth in 2026. EMEA is described as a 'coiled spring uncoiling' with strong Q4 orders and expected return to growth in H2 2026 (flat to down mid-single digits in H1). APAC is accelerating, guided to mid-20s growth, despite China remaining 'muted' due to macroeconomic conditions. Nordics specifically called out as moving well. Geographic diversification language has intensified: decision-makers' realization that 'things need to happen beyond North America' is now palpable. Site selection drivers are not explicitly detailed.

“The coiled spring is uncoiling. The market sentiment has significantly improved. Pipeline growth has accelerated. We saw strong orders in Q4, and we expect that to continue in '26.”

Corning's solar wafer facility initially operated on temporary power and water systems due to inability to secure utility-provided permanent infrastructure on schedule. Management is conducting an extended maintenance shutdown in Q2 2026 to transition to permanent power systems and upgrade production equipment. The company cites government support and customer funding as enabling the buildout but does not flag broader grid interconnection delays or SMR interest.

“Part of that meant bringing up our facility on temporary power and water systems because we couldn't get the utilities to build the permanent systems on our schedule.”

Management does not cite grid interconnection delays as a constraint on Legrand's operations. However, CEO acknowledges that grid access is a structural industry challenge: some countries require 2-3 years for data center grid connection, which may shift project timelines from 2026 to 2028. This is positioned as an industry-wide PUE efficiency driver rather than a Legrand-specific bottleneck. No counterparties or PPAs named. Bring-your-own-power or on-site generation not mentioned. SMRs not mentioned in context of Legrand's offerings.

“It is a fact that you have geographies where access to grid is a constraint...it can take up to 2, 3, 4 years for a data center to get connected to the grid.”

Digital Realty does not cite specific grid interconnection delays by geography or timeline in this filing. The MD&A acknowledges potential energy cost inflation and regulatory risk but frames these as future policy risks rather than current operational constraints. Language remains speculative ('could significantly increase,' 'may not be able to pass on') rather than contractual on power sourcing. No SMR mentions, on-site generation specifics, or new PPAs are disclosed in Q1 2026.

“The cost of electric power comprises a significant component of our operating expenses. Any additional taxation or regulation of energy use... could significantly increase our costs.”

Management did not cite grid interconnection delays by name or timeline. However, sovereignty and geographic diversification are primary drivers: seven new data center investments announced in Q2 specifically to support local data residency and customer control requirements. Fairwater's Atlanta and Wisconsin sites are connected via AI WAN to form a 'first-of-kind AI super factory' with two-story design and liquid cooling for higher GPU density. No specific power purchase agreements (PPAs) are named, nor are small modular reactors (SMRs) mentioned. Language is contractual and operational rather than exploratory.

“We announced DC investments in seven countries this quarter alone supporting local data residency needs. Sovereignty is increasingly top of mind for customers.”

Management does not cite grid interconnection delays as a material constraint. The focus is on customer-driven lead times (12-18 months for large orders) rather than grid-driven delays. No mention of bring-your-own-power, SMRs, or new PPA execution. Power strategy centers on system-level design with customers, not grid-constrained sourcing.