◈ A SKALD CORPORATION INITIATIVE // CONCEPT DOCUMENT 2026
VARDA GIVES BACK MORE THAN IT TAKES.
Net-Positive Energy Infrastructure
Every data center ever built has made the same deal with the world: give us power and water, and we'll give you computation. The heat, the waste, the burden on the grid — that's just the cost of doing business. Everyone agreed this was unavoidable.
It isn't.
Varda is a data center that produces more clean energy than it draws from the grid.
It returns more clean water than it consumes.
It heats homes, grows food, and creates jobs — as a byproduct of running servers.
It is net positive, by design, at every scale from a neighborhood building to a hyperscale campus.
This is not a proposal to be more efficient.
This is a different idea of what infrastructure is for.
The technology to do this exists today. The AWG systems, thermoelectric generators, geothermal loops, district heating infrastructure — none of it is experimental. What's missing is the decision to combine them deliberately around the question: what if a data center were a gift to the community it sits in?
Varda is the answer to that question, worked out in detail, at five scales of construction — from a 2,500 sq ft neighborhood facility to a 23-acre hyperscale campus. Every tier is net positive. Every tier creates jobs, produces food, and returns clean water. The scale changes. The principle doesn't.
WHAT VARDA DOES DIFFERENTLY
01
⚡
NET ENERGY PRODUCER
Rooftop solar, wind turbines, and thermoelectric generators recapture waste heat as electricity. At hyperscale, the surplus powers 45,000 homes. Even the smallest Tier 1 facility exports power back to the grid.
02
💧
WATER RETURNED CLEAN
Servers exhale humid hot air. Varda captures it, condenses it, filters it to potable standard, and returns it to the local water table or community. In water-stressed regions, this becomes critical civic infrastructure.
03
🌡
WASTE HEAT AS A SERVICE
Every watt of computation produces heat. That heat is routed to greenhouse grow pods for year-round food production, district heating pipelines for nearby homes, and absorption chillers that convert it back into cooling.
04
🌿
FOOD FROM COMPUTE
Greenhouse grow pods fed by waste heat and recaptured water produce food year-round. At Tier 5, the agri-food campus yields 150,000–300,000 kg of produce per month — enough to supply hospitals, schools, and grocery chains.
05
👷
JOBS BY DESIGN
Every Varda facility has a structured community hire track built into the operating model — not as a PR gesture but as a functional role. At hyperscale, 30–50 community positions per year flow into a permanent career pipeline.
06
⚖
SCALES WITHOUT COMPROMISE
The same net-positive principle applies from a 100 kW neighborhood node to a 500 MW hyperscale campus. The systems scale together. The community benefits scale with them. There is no version of Varda that extracts.
Varda doesn't invent new technology. It assembles existing breakthroughs — many of them created by people who were solving entirely different problems — into a single coherent system. Two of those breakthroughs deserve to be named directly.
◈ WATER COLLECTION
WARKA WATER TOWER
Arturo Vittori & Andreas Vogler, Architecture and Vision — 2012
Italian architect Arturo Vittori visited the Ethiopian highlands in 2012 and watched women and children walk for hours to collect contaminated water. He came home and designed a 9.5-meter tower made of bamboo and polyester mesh that needs no electricity, no pumps, and no infrastructure. It collects fog, dew, and rain through condensation — the same way a spider's web collects morning moisture — and gravity-feeds clean water into a reservoir at its base.
A single Warka tower weighs 80 kg, costs around $550 to build, and produces up to 100 liters of clean water per day. It won the World Design Impact Prize in 2016 and has been deployed in Ethiopia, Togo, Cameroon, and Haiti.
Varda uses the Warka Tower as a passive water collection layer — deployed around the facility perimeter and roofline, working at night and in fog conditions when the active AWG system is less productive. Together they form a redundant, layered water system. The Warka towers also give Varda its exterior identity: standing structures that are visibly, unmistakably not a conventional data center.
DESIGNED AS OPEN-SOURCE · warkawater.org · World Design Impact Prize 2016 · Aga Khan Award Shortlist 2019
◈ WIND ENERGY
GEOWIND GW1200
Young June Jeon, GeoWind — South Korea, 2023
South Korean physicist Young June Jeon looked at the geometry of a regular icosahedron — twenty triangular faces, twelve vertices, one of the most structurally stable forms in nature — and realized it was also a perfect wind turbine frame. His GeoWind turbine builds a vertical-axis wind turbine inside that geometry, distributing aerodynamic loads across the entire structure rather than concentrating stress at any single point.
The result operates in turbulent, low-speed urban wind — the kind that stalls conventional turbines — starting at just 4 m/s and spinning omnidirectionally across 360 degrees with no yaw mechanism needed. It's quiet, modular, portable, and doubles as a climate data node, measuring wind speed, temperature, humidity, and pressure in real time. It won the CES 2026 Innovation Award in the Urban Wind Power category.
A data center rooftop is one of the worst environments for conventional turbines — chaotic airflow, HVAC exhaust turbulence, edge vortices. It's almost exactly the environment GeoWind was designed for. Varda deploys GeoWind arrays on every rooftop and perimeter structure, treating the chaotic airflow as a feature rather than a problem. In developing regions, Jeon releases the design open-source.
PATENT PCT/KR2023/019844 · CES 2026 INNOVATION AWARD · iF Design Award · geowind.kr · Pre-POC validation ongoing
VARDA-1 // NET-POSITIVE ENERGY INFRASTRUCTURE
Net-Positive Energy Infrastructure — Blueprint Rev. 1.0
DRAWING NO. VARDA-BP-001
CLASSIFICATION: SYSTEMS / ENERGY LOOP
SCALE: SCHEMATIC
STATUS: ● CONCEPTUAL ACTIVE
ENERGY FLOWS:
ELECTRICITY (RENEWABLE)
THERMAL / HEAT ENERGY
WATER / CONDENSATE RETURN
VAPOR / HUMID AIR
RECLAIMED / SECONDARY ENERGY
GRID BACKUP (MINIMAL)
RADIATIVE COOLING → SPACE
⚡ Energy Input Systems
Rooftop photovoltaic array (400–800 kW peak)
GeoWind GW1200 icosahedron VAWT arrays — see below
LiFePO₄ battery storage bank for load-shifting
UPS + inverter for clean DC/AC distribution
Grid tie-in as emergency-only backup (red loop)
TEG recirculation supplements internal bus
Primary power coverage target 85–95%
🌀 GeoWind Turbine Arrays
Designed by physicist Young June Jeon, GeoWind — South Korea, 2023
Commercial vertical farm, aquaponics, distribution
~25,000–50,000 households
Ops / NOC / Campus
40,000
4%
Full campus ops, R&D labs, employee facilities
500–1,000+ FTEs
SUBSYSTEM ENERGY I/O
⚡ Solar + GeoWind Megafarm
Solar peak
150–220 MW
GeoWind units
2,000–5,000 GW1200 campus-wide
GeoWind output
~20–50 MW combined
Battery storage
500 MWh
Self-sufficiency
~30–45% of total draw
💧 Water Recovery (AWG + Warka)
AWG megaplant output
500,000–800,000 L/day (active)
Warka towers (500–800)
+20,000–80,000 L/day (passive)
Population served (combined)
~12,000–25,000 people/day
🌡 Thermal Plant
TEG DC output
~20–35 MW
District heat city loop
~100–150 MW thermal
🌿 Agri-Food Campus
Crop yield
~150,000–300,000 kg/month
Beneficiaries
~25,000–50,000 households
⚙ City Heat Infrastructure
Buildings heated
~20,000–40,000 units
Revenue potential
~$30M–$80M/yr
♻ Net Balance
Net surplus exported
~60 MW avg
Water returned/day
500,000–800,000 L clean
At hyperscale, 60 MW net surplus powers approximately 45,000 average American homes continuously — the data center becomes a net electricity producer for its surrounding grid.
500,000–800,000 L/day of clean water exceeds the daily needs of 10,000–20,000 people. In water-stressed regions, this becomes critical civic infrastructure.
TEG output at 20–35 MW is enough to power a small city district entirely on recaptured waste heat — electricity produced as a byproduct of compute.
District heat at 100–150 MW thermal eliminates gas heating for 20,000–40,000 residential units — equivalent to ~50,000–100,000 tons of CO₂/yr avoided.
◈ FULL PROJECT PLAN FOR THIS SCALE
THE CASE // CONVENTIONAL VS. VARDA
What the industry accepts as inevitable — and what doesn't have to be
A conventional data center is an extraction machine. It takes power from the grid, turns most of it into heat, and expels that heat into the atmosphere. It draws enormous amounts of water for cooling and evaporates it. It returns nothing to the community it sits in except jobs — and usually, not many of those.
Nobody designed it to be this way deliberately. It's the result of a thousand engineering decisions made in isolation, each locally optimal, collectively wasteful. Varda is what happens when you ask a different question at the start: what should this building give back?
The answer draws on work already done by others. Arturo Vittori's Warka Water Tower showed that a passive bamboo structure could pull drinking water from fog and dew with zero electricity. Young June Jeon's GeoWind GW1200 showed that a geodesic icosahedron frame could generate power from the turbulent, chaotic wind on a city rooftop where conventional turbines fail. Neither was designed for a data center. Varda combines them — and a dozen other existing technologies — into a single system that does something none of them were designed to do alone.
SIDE BY SIDE — 5 MW MID-SCALE FACILITY
Category
CONVENTIONAL DATA CENTER
VARDA
Grid power draw
6 MW continuous, 100% from grid
6 MW total; 40–55% from on-site renewables
Energy returned to grid
None — net consumer
+600 kW average surplus exported
Waste heat
~4–5 MW expelled to atmosphere
Captured: district heat + TEG + greenhouse + chiller
Water consumption
~100,000–500,000 L/day evaporated via cooling towers (WUE avg 1.8 L/kWh; varies significantly by climate and cooling type)
Net water producer: +5,000–8,000 L/day returned clean — no evaporative cooling towers
Food production
None
1,500–3,000 kg/month from greenhouse complex
Homes heated
None
200–400 residential units via district heating
PUE (efficiency)
1.56 average (Uptime Institute, 2024); range 1.3–2.0+ across industry
Target <1.2 — 25–35% below industry average
CO₂ footprint
~15,000–25,000 tons CO₂/yr (grid-dependent)
Net negative when district heat displaces gas heating
Community jobs
5–15 FTE, typically specialist roles
21 FTE with structured community hire track
Community programs
None built-in
CSA food program, water access, STEM education, district heat
Revenue streams
Colocation only
Colocation + energy credits + heat contracts + food sales + water
Conventional: evaporates ~500M–2B L/year at hyperscale (varies by WUE and climate)
Varda: returns 180M–290M liters/year clean
NET CIVIC WATER INFRASTRUCTURE
HYPERSCALE ENERGY SURPLUS
Conventional: largest grid consumer in region
Varda: exports +60 MW to grid
POWERS ~45,000 HOMES
WHY HASN'T THIS BEEN BUILT?
The honest answer is that conventional data centers weren't designed to give back because nobody required them to. The systems Varda uses — AWG, TEG, district heating, geothermal loops — all exist. None of them are experimental. They've been deployed in isolation around the world for decades.
What Varda does is combine them intentionally around a single design principle: every output of this building is an input for something else. The heat doesn't leave. The water doesn't leave. The energy surplus doesn't disappear. Everything transmutes.
The cost premium is real. A Varda-1 Mid-Scale costs roughly 15–25% more to build than a conventional facility of the same IT capacity. That premium is recovered through energy credits, district heating revenue, water value, and food production — and through the political capital that comes from being the kind of infrastructure a city actually wants in it.
The question was never whether it was possible. It was whether anyone would decide it was worth it.
WHERE IT WORKS // SITE SELECTION CRITERIA
What makes a location right for Varda — and where to start
Varda isn't location-agnostic. The net-positive systems perform differently depending on climate, grid conditions, water stress, and community density. Understanding the site criteria helps identify where a Varda facility delivers the most value — and where it's hardest to build.
SITE SELECTION CRITERIA
⚡
GRID CONDITIONS
Renewable energy is the backbone of Varda's net-positive energy claim. Sites with access to deregulated electricity markets allow Varda to both purchase renewable PPAs and sell surplus back to the grid. Net metering policies, grid interconnect capacity, and the cleanliness of the local grid all affect the math.
Ideal: Deregulated market, high renewable penetration, net metering, available interconnect capacity within 5–10 miles. Texas ERCOT, MISO, PJM are strong candidates.
💧
WATER STRESS
The AWG system's community value multiplies in water-stressed regions. Returning 50,000–800,000 liters of clean water per day to a community that struggles with water security transforms the facility from infrastructure into essential civic service — and creates powerful political alignment for approval and operation.
Ideal: High water stress index — Texas, Arizona, Nevada, California inland, Middle East, North Africa, parts of India and Australia. The higher the stress, the greater the civic value.
🌡
CLIMATE FOR AWG
The AWG system extracts moisture from server exhaust air. Higher ambient temperatures mean servers run hotter and produce more humid exhaust — which means more water yield. Hot climates produce better AWG output and also create greater community demand for the resulting clean water.
Ideal: Average temperatures above 20°C (68°F), moderate to high humidity. Southern US, Gulf Coast, Southeast Asia, Middle East. Cold climates reduce AWG yield significantly.
🌿
COMMUNITY DENSITY
District heating, food distribution, water access, and STEM programs all require a proximate community to deliver value to. Varda performs best when it can serve a neighborhood, district, or small city within a 1–5 mile radius. Remote or industrial sites reduce community benefit impact even if the energy math works.
Ideal: 5,000–50,000 people within 3 miles. Established neighborhoods, growing suburbs, secondary cities — not greenfield industrial parks.
🏗
GEOTHERMAL CONDITIONS
The geothermal ground loop is Varda's thermal sink for excess heat and its pre-heat source in winter. The depth and cost of boring varies significantly by geology. Sedimentary basins with stable temperature gradients are ideal. Rocky or highly variable substrates increase drilling cost substantially.
Ideal: Sedimentary geology, stable 10–15°C ground temperature at 100–400ft, no major aquifer conflicts. Most of the US South and Midwest qualifies well.
🏛
REGULATORY ENVIRONMENT
Varda introduces novel regulatory questions: AWG water classification (is it drinking water? stormwater? reclaimed?), TEG as a distributed generation asset, district heating as a utility service, and food production adjacent to a data center. Jurisdictions with flexible permitting and innovation-friendly regulators move faster.
Ideal: State or municipal governments actively seeking sustainable infrastructure. Jurisdictions with economic development incentives, green bond programs, or sustainability mandates. Texas, Colorado, North Carolina, Arizona have been receptive to data center innovation.
STRONGEST CANDIDATE MARKETS — INITIAL BUILD
SAN ANTONIO, TX
Growing tech sector, hot climate (strong AWG yield), ERCOT deregulated grid, significant water stress in long-term forecasts, politically receptive to infrastructure investment, proximity to Austin tech corridor. CPS Energy has active renewable procurement. Strong case for Tier 1–2 demonstration build.
WATER STRESSDEREGULATED GRIDGROWTH MARKET
PHOENIX / SCOTTSDALE, AZ
Extreme water stress makes AWG community value extremely high. Major data center market already established. Strong solar irradiance (best in US). State-level interest in water security solutions. High ambient temperature maximizes AWG yield. Political alignment with water infrastructure investment.
EXTREME WATER STRESSPEAK SOLAR
RALEIGH-DURHAM, NC
Established data center corridor (Research Triangle), favorable utility environment, strong university partnerships for STEM programs, growing population creating demand for community services, mild enough climate for reasonable AWG performance, state economic incentives for tech infrastructure.
TECH CORRIDORUNIVERSITY PARTNERSHIPS
DENVER / FRONT RANGE, CO
Strong renewable energy grid, state-level sustainability commitments, growing data center market, good geothermal conditions on the eastern plains, active policy environment for distributed energy resources. District heating has existing infrastructure precedent in Denver.
RENEWABLE GRIDDISTRICT HEAT PRECEDENT
INTERNATIONAL: UAE / QATAR
Extreme water stress, massive sovereign wealth appetite for infrastructure investment, strong political will for desalination and water security, high ambient temperatures, existing data center demand, and an explicit national interest in being seen as leaders in sustainable infrastructure. AWG at scale is a near-strategic asset here.
CRITICAL WATER STRESSSOVEREIGN CAPITAL
SINGAPORE / SOUTHEAST ASIA
Singapore has a moratorium on new data center builds due to energy and water constraints — Varda's net-positive profile is exactly what that moratorium was designed to incentivize. High humidity maximizes AWG yield. Government investment in sustainable infrastructure is significant and consistent.
POLICY ALIGNMENTMAX AWG YIELD
◈ Site selection analysis is conceptual. Actual site viability requires utility interconnect studies, geotechnical investigation, environmental review, and municipal engagement. These profiles represent qualitative alignment with Varda's design principles, not a formal feasibility determination.
THE NUMBERS // ECONOMICS AT EACH SCALE
Revenue streams, operating costs, and the financial case for net-positive infrastructure
Varda is not a charity project. It is designed to be financially self-sustaining — and at scale, profitable — through a diversified revenue model that no conventional data center can replicate. Because Varda produces multiple valuable outputs from the same inputs, its revenue base is fundamentally broader than colocation alone.
These figures are conceptual estimates based on current market rates for each revenue category. They are not financial projections and should not be treated as such. They are intended to demonstrate that the economic model is coherent — that the additional construction cost of a Varda facility can be recovered, and that the community benefits are not subsidized charity but byproducts of a sound business.
REVENUE STREAMS
⚡ COLOCATION / COMPUTE
Primary revenue. Customers pay for rack space, power, and connectivity. Typical rate: $150–$400/kW/month for wholesale; $400–$1,200/kW for retail colo. Anchor: 80% occupancy at market rates pays the base operating cost.
💧 ENERGY SURPLUS EXPORT
Surplus renewable generation sold back to the grid or through PPAs. Typical rate: $0.03–$0.08/kWh depending on market and contract type. At Tier 3 Mid: +600 kW avg × 8,760 hrs = ~5.3 million kWh/yr = $160K–$420K/yr
🌡 DISTRICT HEATING CONTRACTS
Residential and commercial heating supply agreements — selling waste heat that would otherwise be expelled. Typical rate: $0.04–$0.08/kWh thermal equivalent. At Tier 3 Mid: 1.2–1.8 MW × 8,760 hrs = $420K–$1.3M/yr
🌿 FOOD PRODUCTION
Wholesale produce from greenhouse complex to restaurants, food banks, CSA subscribers, and grocery distributors. Typical rate: $2–$8/kg wholesale depending on crop type. At Tier 3 Mid: 1,500–3,000 kg/month = $36K–$288K/yr
💰 CARBON OFFSETS & ESG CREDITS
Verified carbon reduction from grid displacement, heat offset, and water return can generate tradeable carbon credits and ESG compliance value. Typical rate: $15–$80/ton CO₂ equivalent. Value: Varies significantly by certification and market.
🏛 MUNICIPAL PARTNERSHIPS
Water supply agreements, emergency water reserve contracts, district heating utility designation — governments pay for reliable infrastructure. At scale, these become material revenue lines with long-term contractual stability. Value: Highly location-dependent; potentially $500K–$10M+/yr at regional/hyperscale.
ECONOMICS BY TIER — ILLUSTRATIVE ESTIMATES
Scale
Build Cost
Annual Ops Cost
Colo Revenue (80% occ.)
Ancillary Revenue Est.
Notes
TIER 1 — MICRO
$1.8M–$3.2M
$175K–$240K
$180K–$480K/yr
$30K–$80K/yr
Energy + minimal food. Payback 6–12 yrs depending on colo rates.
TIER 2 — EDGE
$8M–$14M
$480K–$620K
$900K–$2.4M/yr
$120K–$350K/yr
Energy + district heat pilot + food. Strong unit economics at moderate colo rates.
TIER 3 — MID
$55M–$95M
$2.2M–$3.0M
$9M–$24M/yr
$600K–$2.0M/yr
All revenue streams active. Ancillary revenue offsets premium build cost within 3–5 yrs.
TIER 4 — REGIONAL
$450M–$750M
$8M–$12M
$90M–$240M/yr
$5M–$12M/yr
Municipal water and heat contracts become material. ESG premium attracts lower-cost capital.
TIER 5 — HYPERSCALE
$4.5B–$8.0B
$60M–$90M
$900M–$2.4B/yr
$50M–$120M+/yr
At hyperscale, ancillary revenue is a material business line in its own right. City-scale water and heat contracts + large carbon offset portfolio + food wholesale operation.
THE COST PREMIUM — AND WHY IT'S WORTH IT
A Varda facility costs roughly 15–25% more to build than a conventional data center of equivalent IT capacity. That premium is driven by the AWG system, TEG panels, geothermal loop, greenhouse complex, and the additional engineering required to integrate them.
That premium is recovered through three mechanisms: ancillary revenue (energy, heat, food, water contracts), lower operating costs (on-site renewables reduce grid dependency), and cost of capital (green bond financing, municipal partnerships, and ESG-driven institutional investment carry lower interest rates than conventional data center debt).
At Tier 3 and above, the ancillary revenue lines are large enough that the facility's economics are materially better than a conventional facility — not despite the premium build cost, but because of what that premium unlocks.
At Tier 1 and 2, the economics are tighter. The Micro and Edge builds are best understood as demonstration projects — proof of concept at a scale where the risk is manageable and the community impact is visible and immediate. They prove the model so that Tier 3 and beyond can be built with confidence.
◈ All figures are illustrative estimates based on publicly available market rate data for colocation, energy, heating, and agricultural products as of 2024–2025. They do not constitute financial projections, investment advice, or a business plan. Actual economics depend heavily on location, energy market conditions, occupancy rates, and contract terms. Consult qualified professionals before making any investment or development decisions.
HORIZON TECHNOLOGIES // WHAT COMES NEXT
Emerging and adjacent systems that could extend Varda's capabilities — not part of the current spec, but worth watching
The Varda design spec is built on technology that exists and is deployable today. But the systems below represent the next layer — some proven at small scale, some still in lab or early commercial phase, all directly relevant to what Varda is trying to do. None of these are baked into the current design. They're presented here as a living inventory of possibilities: things to evaluate as they mature, things to pilot at specific sites, things that may make the next generation of Varda significantly more capable than this one.
The honest framing: some of these are ready. Some are five years out. Some may not work at Varda's scale at all. The point isn't to promise them — it's to show that the frontier keeps moving in Varda's direction.
WATER COLLECTION & EXTRACTION
🌫
CLOUDFISHER FOG NETS
Peter Trautwein / Aqualonis & WasserStiftung — Germany, 2013
Industrial-scale fog collection nets developed by Peter Trautwein for the German Water Foundation. The CloudFisher Pro uses a 3D polyester mesh (54 m² per unit) that traps fog droplets and gravity-feeds them to a collection trough — no electricity, no moving parts, withstands winds up to 120 km/h. In Morocco, 31 units across 1,686 m² produce up to 37,000 liters per foggy day, supplying 1,300 residents.
How it fits Varda: Where the Warka Tower is community-scaled and beautiful, the CloudFisher is industrial-scaled and high-throughput. At Tier 3 and above, CloudFisher panels on the building perimeter and roof parapet could add significant passive water yield alongside the AWG and Warka systems. Best in coastal or elevated sites with persistent fog. Yields 10–22 L/m²/day depending on conditions.
STATUS: COMMERCIALLY DEPLOYED · aqualonis.com · UNFCCC Momentum for Change Award
🔬
MOF-BASED DESERT WATER HARVESTING
Prof. Omar Yaghi, UC Berkeley / MIT — 2017–present · 2025 Nobel Prize in Chemistry
Metal-Organic Frameworks (MOFs) are ultraporous crystalline materials that adsorb water vapor from air at relative humidity levels as low as 10% — far below what conventional AWG or fog nets can capture. At night, the MOF absorbs moisture from cool desert air. During the day, ambient sunlight heats the material, releasing the water as vapor which condenses into a collection chamber. No electricity required. Professor Omar Yaghi, who won the 2025 Nobel Prize in Chemistry for his work on reticular chemistry, has demonstrated devices producing 2.8 L of water per kg of MOF per day at 20% humidity in Arizona desert testing.
How it fits Varda: This closes the critical gap in the current water system: the AWG and Warka towers both underperform in very dry air. MOF harvesting works exactly where they struggle — Phoenix at 15% humidity, UAE coastal desert, Chihuahuan desert. A MOF array integrated into the AWG system or deployed on the building envelope could make Varda's water claim valid in the world's driest climates. The waste heat from TEG systems could also drive the MOF desorption cycle, replacing solar heating entirely.
Established materials science · Commercial deployments in EU building sector
Phase Change Materials absorb large amounts of heat when they transition from solid to liquid (at a specific temperature), storing thermal energy without any mechanical system. Salt hydrates, paraffin wax, and bio-based PCMs can be embedded in walls, ceilings, or thermal battery units. They absorb waste heat during peak compute hours and release it slowly overnight — smoothing the thermal recovery curve and extending the useful window for district heating and greenhouse warming.
How it fits Varda: The current thermal recovery system is real-time — heat produced now is used now. PCMs allow heat produced at 2pm peak load to be delivered to homes at 8pm when heating demand is higher. They decouple generation from delivery, making district heating far more practical for residential customers. Particularly valuable in Tier 2–3 where district heating viability is marginal without storage.
STATUS: COMMERCIALLY AVAILABLE · Widely deployed in European building sector · Multiple vendors
ENERGY RECOVERY & GENERATION
🌌
RADIATIVE SKY COOLING
SkyCool Systems — Eli Goldstein & Aaswath Raman, Stanford University · 2017
A multilayer optical film that reflects 97% of incoming sunlight while simultaneously emitting heat as infrared radiation at wavelengths that pass through the Earth's atmosphere into the cold of space. The result: the panel stays cooler than the surrounding air, 24 hours a day, even under direct sun. SkyCool embeds fluid pipes beneath the panels — the cooling effect pre-chills water that flows through the facility's cooling loop, reducing chiller energy use by 15–25% in deployed commercial systems (up to 40% projected in optimal arid conditions). No electricity in. No water evaporated. The sky is the heat sink.
How it fits Varda: Already included in the Varda core spec as a rooftop add-on. Listed here because it deserves fuller context — this is one of the most elegant energy reduction technologies available. In hot dry climates (San Antonio, Phoenix, UAE), where both sky clarity and cooling demand are highest, SkyCool panels can offset a significant fraction of the chiller load that the current system still draws grid power for. ARPA-E funded. Data center application specifically cited by Stanford team.
STATUS: COMMERCIALLY DEPLOYED · skycoolsystems.com · ARPA-E funded · Stanford / UCLA origin · CES featured
👣
KINETIC FLOOR ENERGY HARVESTING
Energy Floors, Pavegen, and academic piezoelectric research — UK/Netherlands · 2009–present
Tiles that convert mechanical pressure from footsteps, equipment vibration, and cooling system vibration into electricity. Kinetic floor tiles using electromechanical systems generate up to 35W per module under heavy foot traffic. Piezoelectric alternatives produce microwatts to milliwatts per step — less output, but solid-state and more durable. A data center floor experiences constant low-level vibration from cooling fans, server chassis resonance, and HVAC systems — a different source than human footsteps, but persistent and predictable.
How it fits Varda: The energy output is small relative to Varda's scale — this is not a meaningful power source. The reason to consider it is philosophical more than practical: it completes the "every output is an input" principle down to the most granular level. It also has sensor value — kinetic tiles are excellent occupancy and movement monitors, which has real operational utility in a data center. Worth piloting in high-traffic areas (NOC, loading docks, greenhouse pathways) at Tier 2 and above.
STATUS: COMMERCIALLY AVAILABLE · Energy Floors BV (Netherlands) · Pavegen (UK) · Deployed in airports, stadiums, transit hubs
🌊
MICRO-HYDRO ON COOLING LOOPS
Established micro-hydro engineering · Adapted from industrial process recovery
The chilled water and condenser water loops in a data center move large volumes of fluid at significant velocity and head pressure. Small in-pipe turbines (pressure recovery turbines or inline micro-hydro) can capture the kinetic energy of flowing water on the return lines — the same principle as regenerative braking in electric vehicles. Output per unit is small (typically 1–10 kW per recovery point) but multiple recovery points across a large facility aggregate meaningfully. Zero additional water consumed, zero moving parts exposed to the environment.
How it fits Varda: At Tier 3 and above, the cooling water volume justifies the installation cost. A mid-scale facility circulating tens of thousands of liters per hour could recover 20–80 kW from strategic turbine placement on return manifolds. Not transformative but consistent with Varda's design philosophy — capture everything. Works alongside existing pumps, requires no infrastructure change beyond inline turbine installation.
STATUS: ENGINEERING STAGE · Components exist commercially · Data center application not yet standardized · Requires flow analysis per site
🦠
MICROBIAL FUEL CELLS (MFC)
Active research — multiple universities · Early commercial stage
Bacteria naturally generate small electrical currents as they break down organic matter in wastewater and biological waste streams. Microbial fuel cells harness this by placing electrodes in a chamber where bacteria oxidize organic compounds — the electrons released flow through an external circuit as electricity. The AWG filtration system produces a small but consistent biological waste stream from condensate collection. Greenhouse aquaponics systems produce organic waste continuously.
How it fits Varda: The output is small — milliwatts to watts per liter of waste treated. But the framing matters: the waste stream is already being produced. Running it through an MFC converts a disposal problem into a small generation asset while simultaneously treating the waste. Most interesting at Tier 3 and above where waste volumes are sufficient to make the system non-trivial. Also aligns with Varda's ethos — the last waste stream produces the last electrons.
STATUS: EARLY COMMERCIAL · Multiple university spin-outs · Not yet standardized for data center application · 5–10 year horizon for meaningful scale
The greenhouse grow pods in the current Varda spec are good. CEA takes them further: LED lighting calibrated to photosynthetically active radiation (PAR) spectra specific to each crop, AI-driven nutrient delivery, fully controlled atmosphere (CO₂ enrichment, temperature, humidity). Commercial CEA operators report 10–350x higher yields per square foot than conventional farming. Varda's waste heat and AWG water are exactly the inputs CEA needs — stable warmth and clean water on demand.
How it fits Varda: The current grow pod spec is conservative. Upgrading to full CEA at Tier 3 and above transforms the greenhouse from a community benefit add-on into a genuine commercial food operation. At Tier 4 Regional, a fully optimized CEA facility could produce 50,000–100,000 kg/month — enough to supply hospital and school systems. The compute waste literally grows the food.
STATUS: COMMERCIALLY PROVEN · Multiple large-scale operators · High capex but strong unit economics at scale
🐝
POLLINATOR HABITAT INFRASTRUCTURE
Established ecological practice · Solar farm co-deployment model
Solar farms co-deployed with native pollinator habitat (wildflower meadows, native grasses, bee hotels) have demonstrated measurable increases in both solar panel efficiency (from natural cooling) and surrounding agricultural productivity. Several major solar operators now deploy "agrivoltaic" designs as standard. The land area around a Varda facility — particularly at Tier 3 and above — could support significant pollinator habitat while also providing the solar farm its efficiency benefit.
How it fits Varda: The facility's AWG water return and greenhouse drainage could naturally irrigate pollinator habitat on the perimeter. This creates a visible, tangible ecology around the facility — bees, butterflies, native plants — that makes the net-positive nature of the site legible to the surrounding community in the most immediate possible way. No complex engineering. Primarily a design decision.
STATUS: ESTABLISHED PRACTICE · No new technology required · Primarily a site design choice · Widely documented in agrivoltaic literature
🐟
ADVANCED AQUAPONICS / RECIRCULATING AQUACULTURE
Recirculating Aquaculture Systems (RAS) · Commercially proven at scale
Recirculating Aquaculture Systems (RAS) use highly controlled water environments to raise fish at densities 10–100x higher than traditional aquaculture, with 90–95% water recirculation (almost no water discharged). The waste from fish directly fertilizes the plant systems in aquaponics loops. Commercial RAS operators raise salmon, tilapia, trout, and shrimp in landlocked facilities. Varda's stable waste heat and AWG water supply are ideal inputs — cold-blooded fish are dramatically easier to raise at consistent temperatures.
How it fits Varda: The current spec includes basic aquaponics. Upgrading to full commercial RAS at Tier 4 and above adds a high-value protein production facility that requires no land beyond the existing greenhouse footprint. At Tier 5 hyperscale, a serious RAS operation could produce 20,000+ kg/month of fish protein — a material food security contribution. Waste heat from TEG panels is an ideal supplemental heating source.
Electric vehicle batteries that are no longer suitable for automotive use (typically below 80% capacity) retain 70–80% of their original energy storage capacity — enough for stationary grid storage. Repurposing them into data center battery arrays instead of virgin LiFePO₄ dramatically reduces the embodied carbon and cost of the storage system. Several major automakers now have formal second-life programs. The cost per kWh of second-life battery storage is approximately 40–60% lower than new battery systems.
How it fits Varda: The current spec calls for new LiFePO₄ battery storage at every tier. Substituting second-life EV packs — where available and where the capacity degradation is acceptable — reduces both cost and embodied carbon of the storage system. Particularly relevant at Tier 1 and 2 where cost sensitivity is highest and where the modular nature of second-life packs matches the smaller scale. Requires battery management system (BMS) integration work.
STATUS: COMMERCIALLY AVAILABLE · Nissan xStorage, BMW i Reuse, multiple integrators · Growing supply as EV fleet ages · 5–10 yr proven lifespan in stationary storage
🧠
AI-OPTIMIZED ENERGY DISPATCH
DeepMind (Google) demonstrated 40% cooling energy reduction · Multiple vendors
Machine learning systems trained on a facility's operational data can predict thermal load, optimize cooling dispatch, manage battery charge/discharge cycles, and coordinate renewable generation against grid pricing in real time — far more effectively than static control logic. Google's DeepMind AI achieved a 40% reduction in data center cooling energy use after being trained on historical operational data. The multi-system complexity of Varda (AWG, TEG, geothermal, greenhouse, district heat, multiple renewables) makes it an unusually rich environment for AI optimization.
How it fits Varda: At Tier 3 and above, the interdependency of Varda's systems is complex enough that human-managed control logic will always leave optimization on the table. An AI layer sitting above the BMS and DCIM systems — trained on the specific thermal, water, and energy patterns of the facility — could materially improve the net-positive performance over time. This is less a technology addition than a software layer, and one that gets better the longer the facility runs.
STATUS: PROVEN · DeepMind/Google published 2016 · Multiple commercial DCIM vendors now offer ML optimization · Increasingly standard at hyperscale
Servers are submerged directly in thermally conductive, electrically inert dielectric fluid. No air cooling at all — the fluid absorbs heat directly from chips with near-zero thermal resistance. PUE drops to 1.02–1.05, significantly below the <1.2 target in the current Varda spec. The fluid exits the tank at high temperature (50–70°C), which is much higher grade than air-cooled exhaust heat — making TEG recovery more efficient and district heating supply temperatures more useful for residential heating.
How it fits Varda: The current spec notes immersion as "optional" in Tier 3 and above. Upgrading to full immersion cooling significantly improves every downstream system in Varda — higher grade heat means better TEG output, better AWG yield from high-temperature exhaust, and better district heating temperatures. The trade-off is higher initial cost and more complex operations. At Tier 4+ where IT density is highest, the case is compelling.
STATUS: COMMERCIALLY PROVEN · GRC (Green Revolution Cooling) largest operator · Growing adoption at hyperscale · PUE 1.02–1.05 demonstrated
A NOTE ON MATURITY
These technologies sit at different points on the readiness curve. CloudFisher fog nets and CEA are commercially proven today. SkyCool radiative cooling is deployed but still scaling. MOF water harvesting is lab-to-field validated but not yet commercially available at scale. Microbial fuel cells and kinetic floor tiles are real but their contribution at Varda's scale is currently more symbolic than material.
The honest framing for all of these is: Varda is designed to be built with what exists now. What's on this page is what the second and third generation of Varda facilities will be built with — as each technology matures, it slots into a design that was always intended to receive it.
The goal was never to wait for the perfect technology. It was to build the best possible version of this idea today, with a design that gets better over time.
◈ Technology status descriptions reflect publicly available information as of 2025. Readiness levels and commercial availability vary by region, application scale, and site conditions. This section is informational — not a procurement recommendation or performance guarantee.
SOURCES & CITATIONS
All named technologies, performance figures, and factual claims — with primary sources
Varda is built on the work of many researchers, designers, engineers, and scientists. This page documents the primary sources for every named technology and key factual claim in this document. Where a figure appears as a design estimate rather than a cited fact, it is noted as such. All external links open in a new tab.
Note: Varda's own design estimates — energy balances, water yields, staffing, costs, build timelines — are engineering projections based on the cited technologies and standard industry methodology. They are not independently verified. All performance claims for third-party technologies are sourced from the references below.
WARKA WATER TOWER
Arturo Vittori & Andreas Vogler, Architecture and Vision — 2012
These are the calculations behind each major design estimate in the Varda document. All figures are engineering projections, not guaranteed outputs. Inputs are stated explicitly so any estimate can be recalculated for a different site, climate, or technology assumption. Variables that change significantly by location are flagged.
⚡ SOLAR PV OUTPUT ESTIMATES
INPUTS
Panel efficiency20% (monocrystalline silicon, standard commercial)
Solar irradiance (peak)1,000 W/m² (standard test condition)
System losses~20% (inverter, wiring, soiling, temperature)
Peak sun hours (San Antonio)~5.5 hrs/day average annual
Document states 60–90 kW peak for Tier 1 — this reflects a larger usable roof area than 600 sq ft alone (accounting for full building footprint + additional panels). The 600 sq ft figure is the primary array; total roof is ~800–1,000 sq ft usable.
For Tier 5 (200,000 sq ft = 18,580 m²): 18,580 × 200 W/m² effective = ~3,716 kW = 3.7 MW peak (rooftop only). Ground mount adds 150–220 MW total as stated — requires adjacent land area.
🌀 GEOWIND OUTPUT ESTIMATES
INPUTS
GW1200 rated output2–5 kW per unit (estimated; pre-POC, not commercially validated)
Cut-in wind speed4 m/s (stated by GeoWind)
Avg rooftop wind speed~5–7 m/s (typical urban rooftop)
Capacity factor (urban VAWT)~15–25% (conservative for turbulent rooftop conditions)
⚑ Pre-POC caveatGeoWind has not published validated field power curves. These estimates are based on comparable small VAWT literature, not GeoWind-specific data.
At 20% capacity factor: 42 kW × 0.20 = ~8.4 kW average continuous output
Annual: 8.4 kW × 8,760 hrs = ~73,600 kWh/year
Document states "~40–80 kW" wind output for Tier 2 — this reflects peak output, not continuous average. Peak and average are both honest but mean different things.
The climate data collection function is independent of power output — each unit acts as a sensor node regardless of generation capacity.
⚡ NET ENERGY BALANCE — TIER 1 MICRO (+18 kW)
INPUTS
IT load100 kW
PUE1.3 (maturing system)
Total facility power draw100 kW × 1.3 = 130 kW
Solar output (avg continuous)~60–90 kW peak → ~55–70 kW average (5.5 peak sun hrs ÷ 24)
Wind output (avg continuous)~3–6 kW average (4 units × 3.5 kW × 20% CF)
TEG output~2–4 kW DC (see TEG calculation below)
CALCULATION
Total renewable generation (avg) = 55–70 kW solar + 3–6 kW wind + 2–4 kW TEG = 60–80 kW average
Battery smoothing allows renewable generation to cover load during peak sun hours and draw from battery at night.
Net balance = Generation – Load = 60–80 kW – 130 kW = −50 to −70 kW (grid still needed for baseload gap)
The "+18 kW" surplus figure represents the daily average export during peak generation hours, not a 24-hour continuous surplus. With battery storage, the facility exports during the day and draws from grid at night. Net daily kWh balance can be positive.
Net daily energy: Solar 300–450 kWh + Wind 72–144 kWh + TEG 48–96 kWh = 420–690 kWh generated/day
⚑ Tier 1 Micro is NOT fully renewable-powered. At 65–75% self-sufficiency, it still draws ~30% from grid. "Net positive" at this tier means net kWh surplus on the best solar days, and net grid export over the billing cycle via net metering — not 24/7 off-grid operation. This is clearly the hardest tier to make net-positive and represents a design aspiration more than a proven outcome.
💧 AWG WATER YIELD ESTIMATES
INPUTS
Server exhaust temp~35–45°C (hot aisle exit temperature)
Exhaust relative humidity~60–80% RH (servers humidify air through cooling)
Condenser efficiencyCommercial AWG: ~0.3–0.5 L of water per kWh of electricity input
Airflow (Tier 1)~5,000 CFM = ~141 m³/min from 10–20 racks
⚑ Climate-dependentHot humid climates (TX, FL, SE Asia) yield 2–4× more than cool dry climates
CALCULATION — TIER 1 MICRO
Air volume at 5,000 CFM = 141 m³/min = 8,500 m³/hr
At 40°C, 70% RH: absolute humidity ≈ 34 g/m³
At condenser exit (15°C): absolute humidity ≈ 13 g/m³
Document states 50–150 L/day — this is conservative, accounting for lower real-world condenser efficiency, variable rack loading, and the fact that not all exhaust air routes through the AWG system. The conservative figure is appropriate for a design spec.
⚑ The higher theoretical figure (540–890 L/day) represents ideal conditions in a hot humid climate with optimized airflow routing. The 50–150 L/day in the spec is the design floor for planning purposes — actual yield will likely exceed this in warm climates.
⚛ TEG POWER OUTPUT ESTIMATES
INPUTS
Hot side temperature (server exhaust)~40–60°C at heat exchanger boundary
Cold side temperature (ambient / geothermal)~15–25°C
Temperature differential (ΔT)~20–40°C
TEG module efficiency5–8% at this ΔT (standard bismuth telluride modules)
Heat flux available (Tier 1, 100 kW IT)~80–95 kW thermal (from PUE calculation above)
CALCULATION — TIER 1 MICRO
Available thermal power at HX boundary = ~80 kW (conservative)
TEG array captures ~30–40% of HX boundary heat (not all heat routed through TEG panels)
At 7% module efficiency: 28 kW × 0.07 = ~2 kW DC electrical output
Document states "~2–4 kW DC" — the upper end (4 kW) assumes better ΔT (higher server load, cooler ambient) and 40% heat routing efficiency. Both ends are defensible.
Tier 5 (500 MW IT): Heat available ~400 MW thermal → TEG input ~140 MW → at 7% = ~9.8 MW DC. Document states 20–35 MW — this assumes higher ΔT from immersion cooling (60–80°C hot side) and higher routing efficiency at scale. The upper end (35 MW) is optimistic and requires immersion cooling + optimized heat routing.
🌡 DISTRICT HEATING — HOMES SERVED ESTIMATES
INPUTS
Average US home heating load~7,000–12,000 kWh/year (~800–1,400 W avg over heating season)
District heating season~5–6 months (Oct–Mar in most US climates)
Distribution losses~15–25% in typical district heating pipe networks
⚑ Climate-dependentSan Antonio heating load ~30% of Boston. District heating more valuable in colder climates.
Available thermal export = 1,500 kW (midpoint) × 0.80 after distribution losses = 1,200 kW delivered
Average home heating demand during season = ~3,000 W peak, ~1,200 W average continuous during heating months
Homes served = 1,200,000 W ÷ 1,200 W/home = ~1,000 homes theoretical
Document states 200–400 homes for Tier 3 — this conservative figure accounts for: peak vs. average demand, partial heat coverage (supplement not replace), not all homes being connected, and system availability constraints. The conservative figure is appropriate.
⚑ District heating is most valuable in cold climates. In San Antonio (mild winters), the same 1.2 MW thermal would serve more homes for shorter periods. In Chicago or Denver, fewer homes but for longer. The figures in the document assume a mid-latitude US climate with a moderate heating season.
Document states 1,500–3,000 kg/month for Tier 3 — this assumes a fully optimized CEA operation with vertical stacking (2–4 growing layers), not a single flat greenhouse floor. The higher figure requires vertical farming infrastructure, which the spec notes as an option.
⚑ The higher production figures (Tier 3–5) assume vertical farming technology (multiple growing layers per square foot of floor area), not single-layer greenhouse. Single-layer yield would be 3–5× lower. The Horizon Technologies page flags CEA as the upgrade path to achieve these yields.
🌿 WARKA TOWER SCALING ESTIMATES
INPUTS
Output per tower (Vittori)Up to 100 L/day under ideal fog/dew conditions
Conservative planning figure40–60 L/day average across a year in a warm humid climate
Tower footprint~4.2 m diameter at base cables; ~18 m² ground footprint
⚑ Highly site-dependentOutput drops to near-zero in dry climates with no fog. Best in coastal, humid, or elevated sites.
CALCULATION — TIER 1 MICRO (6–8 towers)
Conservative: 7 towers × 40 L/day average = 280 L/day
Document states "+240–800 L/day (passive)" — consistent with this range
Combined with AWG (50–150 L/day active) = 290–950 L/day total water system output at Tier 1
Tier 5 (500–800 towers): 650 towers × 50 L/day avg = ~32,500 L/day. Document states "+20,000–80,000 L/day" — consistent at mid-range. Upper end requires high fog frequency.
⚑ The Warka tower contribution is the most climate-variable element in the water system. In San Antonio (hot, semi-arid), fog frequency is low and dew collection dominates — yields will be at the lower end. In coastal or high-humidity sites, yields approach the upper end. The AWG (active) system is the reliable baseline; Warka is the passive supplement.
🏗 CONSTRUCTION COST ESTIMATES — METHODOLOGY
BENCHMARKS USED
US data center construction cost$8–$15M/MW IT capacity (JLL, CBRE 2024 benchmarks)
Solar PV installed cost (utility)$0.90–$1.30/W installed (NREL, 2024)
Total estimate: $1.58M–$2.64M — consistent with stated $1.8M–$3.2M range (which includes higher-end site and market conditions)
⚑ The cost ranges in the project plans are deliberately wide to account for regional variation, market conditions, and site complexity. The low end assumes favorable conditions (competitive market, standard geology, existing utility infrastructure). The high end reflects difficult sites, premium markets, or complex permitting.
◈ These calculations are provided for transparency and to enable independent verification or adaptation. All inputs should be validated against site-specific data before any real-world design or investment decision. Varda is a concept document — these calculations are the thinking behind the estimates, not a certified engineering analysis.
◈ All links were verified as of April 2026. External links open in a new tab. Varda is a concept document produced by Skald Corporation — not a commercial product, not investment advice, not an engineering specification. The technologies cited are real; the Varda design that combines them is conceptual.