Data Centers vs. Moore’s Law: How Should Chicago Area Investors Value Data Center and Adjacent Properties

Until roughly 2022, a thoughtful commercial real estate investor could underwrite the digital infrastructure asset class with a fairly comforting assumption: every two years, chips would get smaller, hotter loads would shrink, and the same building could process more workload on the same megawatts. That was the Moore’s Law subsidy. It has, in any practical sense, ended. AI training compute is now growing roughly 50 to 100 times faster than transistor density, and rack densities that sat at 5 to 8 kilowatts for a generation are routinely passing 100 to 120 kilowatts. The result for Chicagoland investors is that the underlying physics of a data center, the land around it, and even industrial parcels three exits down the tollway have all repriced. This piece walks through how the math now works, why Chicago has been pulled off the bench as a primary market, and how owners and investors should think about valuation for both data center assets and the adjacent properties caught in their wake.

The Compute Math Has Inverted

1. Moore’s Law Is Functionally Over

The traditional Moore’s Law cadence, doubling transistor density every two years, began breaking down in the early 2000s as thermal limits stalled clock speeds. Intel’s own leadership has conceded the doubling now runs closer to three years, and the International Technology Roadmap for Semiconductors retired its Moore’s Law-based plan back in 2016. The industry’s “More than Moore” strategy now relies on chiplets, 3D packaging, and specialized accelerators rather than pure scaling.

2. AI Demand Is Growing Off the Chart

AI training compute has grown roughly 300,000x since 2012, with a doubling time of about 3.4 months. Bain & Company estimates global AI compute demand will require roughly $500 billion in annual data center capital spending by 2030, against power constraints that will likely take a decade to resolve. Industry forecasts call for roughly 97 GW of new data center capacity coming online between 2025 and 2030, nearly doubling global supply in five years.

3. The Real Estate Implication

For decades, software and chip efficiency gains largely cancelled out demand growth, and developers built about the same number of data centers per year. That subsidy is gone. Every gigawatt of new model training requires roughly a gigawatt of new physical infrastructure, because there is no algorithmic discount coming on a timeline that matters to a 10-year hold. For investors, this means physical real estate, the land, the substation rights, and the cooling water access, is no longer a commodity. It is a scarce, durable, monopolistic input.

Why Chicago Got Pulled Off the Bench

For a long time, Chicagoland was a respectable secondary data center market. The Northern Virginia capacity wall changed that. With Dominion Energy’s interconnection queue stretching past three years in Ashburn and no meaningful contiguous space left in the corridor, hyperscalers went looking for the next viable Tier 1 hub.

1. The 2026 Snapshot

Industry research puts Chicago at 1.9 GW of installed capacity in 2025, climbing to roughly 2.03 GW in 2026 and a forecast 2.81 GW by 2031. Vacancy fell to 1.9% on the most recent reading, and asking rents climbed roughly 33% year over year for the metro. There is no contiguous block of 5 MW or larger available, and major operators (Prime, Aligned, CyrusOne, Equinix, CoreWeave, Compass) have all signed or announced significant expansions in the past 18 months.

2. The Power Picture

ComEd’s situation is the real story. The utility currently sits with 75 large-load applications totaling roughly 14 GW in its queue, a combined load greater than the system’s all-time peak. Industry analysts report that some new ComEd-served projects face power delivery delays out to 2031 or later. To protect existing ratepayers, ComEd now requires 10-year letters of credit on projects of 50 MW or more, and the Illinois Commerce Commission opened a formal investigation into large-demand tariffs in March 2026. At the same time, ComEd is executing a $1 billion substation expansion program centered on Elk Grove, with target completion in late 2026.

3. The Incentive Stack

Illinois’s 2019 data center sales-and-use tax exemption (10.25% on qualifying equipment for projects investing at least $250 million and employing 20 qualified staff at 120% of county average wages) has helped attract more than $11 billion in committed investment since enactment. Combined with industrial power tariffs near 6.9 cents per kilowatt-hour and Lake Michigan water access for cooling, the policy environment has been a meaningful pull factor that the surrounding Midwest states cannot fully match.

What Actually Drives Value Now

Underwriting a data center using industrial real estate comps is now actively misleading. Power, not space, is the value driver.

1. The New Income Metric

The institutional standard is now dollars per kilowatt per month rather than dollars per square foot. Wholesale asking rents in Chicago for a 250 to 500 kW requirement run roughly $155 to $165 per kW per month, with premium markets like Northern Virginia and Singapore stretching meaningfully higher. As the valuation team at CBIZ puts it bluntly, data centers are no longer space-constrained assets, they are power-constrained assets. A facility with available square footage but capped power cannot generate incremental revenue, and the appraisal needs to reflect that.

2. Where Cap Rates Sit

Stabilized hyperscale-leased assets are trading at cap rates around 6%, comparable to premium industrial and multifamily, with implied REIT cap rates down roughly 66 basis points from the 2022 peak even amid 10-year Treasury volatility in the 3.97% to 4.40% range. Pre-leasing on new construction exceeds 90% nationally, meaning open-market vacancy is functionally zero in any primary market.

3. The Hidden Risk: Functional Obsolescence

Data center buildings carry a 30 to 50 year economic life on paper, but the critical power and cooling infrastructure inside them needs replacement or upgrading every 8 to 15 years. A facility designed for 10 kW racks cannot easily retrofit for 100 kW liquid-cooled AI training racks. That is a real underwriting risk that should haunt anyone valuing an older Chicagoland colocation facility against a hyperscale comp.

4. The Sponsor Premium

Capital is flowing earlier in the development cycle and behaving more like infrastructure finance than traditional real estate equity. Tenant credit, balance sheet depth, and the ability to deliver commissioned megawatts (not just announced megawatts) have become the primary differentiators. Data Center Watch reports that more than $64 billion in U.S. data center projects have been blocked or delayed since 2023 due to community opposition, zoning, and grid issues. Sponsor execution risk is now priced into every transaction.

Adjacent Property Values, How the Halo Effect Works

The more interesting opportunity for many Chicagoland investors is not the data center itself, but the parcels around it. The halo effect is real, measurable, and underwritten.

1. The Elk Grove Comp Set

When Stream Data Centers needed 34 acres in the Roppolo subdivision in unincorporated Elk Grove Village, it paid roughly $950,000 per home for 55 single-family houses. Industry observers cited by Network World have noted that residential-zoned land in Elk Grove worth $20 per square foot at its prior use trades for $45 to $50 per square foot once entitled for data center use. T5 Data Centers’ purchase at 950-1050 Morse Avenue closed at $21.3 million, more than three times the $6.7 million the seller had paid. Those are not modest premiums. They are step-function repricings of land that had been on the books at industrial or even residential basis.

2. What Makes a Parcel “Halo-Eligible”

Not every neighboring acre catches the lift. The parcels that get repriced typically share a tight set of traits:

• Proximity to a major fiber spine (the Cermak loop, O’Hare, the I-88 and I-90 corridors)
• Within a few miles of an existing or planned ComEd substation with available capacity
• Industrial or convertible-to-industrial zoning, or a cooperative municipality willing to annex and rezone
• Sufficient acreage and configuration for hyperscale footprints, ideally 20 contiguous acres or more
• Access to substantial water for cooling, a meaningful constraint outside the Lake Michigan service area

3. The Submarkets to Watch

Elk Grove Village remains the bellwether, but the constraint set is pushing developers further out. Aurora along the I-88 corridor has absorbed CyrusOne’s 40 MW first phase and an Edged campus, with more acreage in active discussion. Wood Dale, Itasca, Carol Stream, Hoffman Estates (where Compass is redeveloping the former Sears HQ into a 200-acre, $10 billion hyperscale campus), and Minooka are all now in the live pipeline. NTT Data’s acquisition of office buildings on Arlington Heights Road in Itasca is a useful early signal that office-to-data-center conversion is becoming a real underwriting case for the right physical structure (large-floorplate, slab-on-grade, near substations). Equinix’s announced first Illinois hyperscale campus in Minooka stretches the geography meaningfully southwest of the traditional O’Hare cluster.

4. Industrial Tenants Are Being Priced Out

The flip side of the halo is that traditional industrial users (light manufacturing, third-party logistics, last-mile fulfillment) are being squeezed out of the same submarkets. For a landlord with a Class B industrial building on a power-rich parcel, the highest-and-best-use calculation has changed: the data center conversion exit may now beat the multi-tenant industrial lease-up scenario. That is a meaningful pivot for owners who have been operating these assets on autopilot for a decade.

A Five-Step Underwriting Framework for Chicago Investors

For owners and investors evaluating data center or data center-adjacent assets in Chicagoland, the following sequence consistently separates real opportunities from headline-driven mispricing:

1. Verify deliverable power, not nameplate power. Pull the ComEd queue position, study substation timing, and confirm whether the project sits behind a 10-year letter of credit requirement. Announced megawatts and commissioned megawatts are not the same thing.
2. Underwrite on dollars per kilowatt per month, then sanity-check against dollars per square foot. Anchor against the Chicago wholesale band of roughly $155 to $165 per kW per month and apply a project-specific adjustment based on tenant credit, contract length, and power availability.
3. Model functional obsolescence at the equipment layer. Assume a refresh cycle of 10 to 12 years on critical mechanical and electrical systems, and underwrite reserves accordingly. A flat 25-year DCF on aging gear is fiction.
4. Map the halo radius for adjacent assets. Identify substation-served parcels within roughly two miles of an active campus, screen for zoning flexibility, and look for owners with basis low enough to make a conversion exit attractive.
5. Stress-test the regulatory case. The Illinois Commerce Commission investigation, potential BIPA revisions, and any rollback of the 2019 tax exemption are all live political risks. Discount for policy change, especially on assets dependent on the exemption for stabilized yield.

Risks That Should Sharpen the Pencil

Even with the tailwinds, a few risks deserve more attention than the headlines suggest. Global construction costs are running roughly $10.7 million per megawatt in 2025, forecast to rise to $11.3 million per megawatt in 2026, with AI-ready facilities running double those figures because of liquid cooling and density requirements. Capital intensity has reached a point where, as DataBank’s leadership has observed, OpenAI alone reportedly needs to raise $207 billion of equity before reaching profitability. No company in history has ever raised that much capital, and any concentration risk in tenant rosters needs to be priced.

The shift from AI training to inference workloads, expected to dominate by 2027, will also redistribute some capacity toward smaller, more regional facilities. That could either help or hurt specific Chicago suburbs depending on their fiber and power profiles. Markets with strong network density (the Cermak loop, Elk Grove’s O’Hare-adjacent fiber) should benefit; pure greenfield campuses optimized only for training may not.

Most importantly, the AI demand curve is not guaranteed. If model architectures become dramatically more efficient (the algorithmic equivalent of a Moore’s Law replacement), the absolute scale of compute demand could moderate. Underwriting should reflect that scenario as a credible downside case rather than ignore it.

What Comes Next

Chicagoland is, for now, firmly on the right side of the digital infrastructure trade. The combination of available water, comparatively cheap industrial land, a meaningful tax exemption, and Northern Virginia’s grid wall has made the region a primary destination for AI-era hyperscale capital. The end of the Moore’s Law subsidy is the structural reason this is happening, and that subsidy is not coming back on a timeline that any current underwriting horizon should assume. For investors, the right posture is neither uncritical enthusiasm nor reflexive skepticism. It is disciplined underwriting that treats power as the primary asset, recognizes the halo effect on adjacent parcels, and prices in real risks around equipment obsolescence, regulatory change, and the capital intensity of the build cycle. Owners with the right parcels in the right submarkets should be calling their broker, and owners trying to value digital-adjacent properties without these tools are leaving real money on the table.

To position your Chicagoland data center, industrial, or adjacent land holding for this cycle, contact our team of Chicago commercial real estate agents!