Two diverging streams of golden light on a deep-navy field — the two-speed market

ai-smart-buildings.com

The Intelligent Building Brief

Issue №1 · June 1, 2026 · a 12-minute read

Intelligence for the people who own, operate, finance, and build commercial real estate. Capital markets, sector flows, the PropTech tape, and the energy/AI shift underneath all of it — what's signal, what's noise, and what we think you should do about it.

IN THIS ISSUE  ·  From the Editor  ·  The Lead  ·  The Numbers  ·  📊 The Maturity Wall  ·  Capital Markets  ·  📊 Cap-Rate Ladder  ·  Sector Watch  ·  📊 Sector Scorecard  ·  📊 Where the AI Money Goes  ·  PropTech Lens  ·  📊 PropTech Concentration  ·  📊 Office, City by City  ·  Regional Radar  ·  The Playbook  ·  Contrarian Corner  ·  Tools & Signal  ·  📅 The Lookahead  ·  Our Conviction Calls

FROM THE EDITOR

Welcome to issue one. Here's the deal we're making with you: every week we read the capital-markets prints, the sector data, the PropTech tape, and the energy/AI shift underneath all of it — and instead of just handing you the numbers, we tell you what we actually think they mean.

This week the conviction is simple. The biggest capital wave in a generation — roughly $700B of AI infrastructure spend — is being reported as a technology story. It isn't. More than 60% of it pours into power, cooling, and construction: it's a real-estate-and-energy story wearing a tech headline, and most of our industry hasn't repriced for that yet. The owners who internalize it first — who underwrite power access and carbon exposure the way they underwrite location and credit — will spend the next five years buying from the ones who didn't.

We'll be wrong about some things, and we'll tell you when we are. But you'll always get our real read, not a hedge — that's the whole point of a brief written by people who've sat on the operator's side of the table.

— The Editor · AISB Intelligence Desk

01 · The Lead

The two-speed market is now the whole story

If you take one frame into your next investment-committee meeting, make it this: 2026 is not a recovery, it's a divergence. The headline numbers say the freeze is over — and they're real. U.S. investment-sales volume hit $112.6B in Q1, up 18% YoY, back in line with 2017–2018 levels; CRE lending is projected to jump 38% to ~$805B. Cap rates are stabilizing, and for the first time since 2022 the consensus is that yields are past their peak.

But underneath that thaw, capital is sorting into winners and losers with unusual violence — a $1.26T debt wall, 2015-vintage loans refinancing from ~4.5% into ~6.5%, and distress concentrated almost entirely in commodity office. The recovery and the crisis are happening in the same market, at the same time, to different assets.

A data-center building wired into a glowing power grid at dusk

The binding constraint has moved from chips to electricity — the grid, not the building, is the scarce asset.

⏩ FAST LANE — capital chasing

Data centers · Industrial · Multifamily · Medical office · Trophy / Class A+ office · Powered land

  

⏸ SLOW LANE — capital fleeing

Commodity / Class B-C office · Over-levered 2015-vintage refis · Anything priced on flat utilities & zero carbon cost

Why this matters: the winning move in 2026 isn't "is the market up or down" — it's "which side of the divergence is this asset on, and can I underwrite the gap." The winners price obsolescence risk, refi risk, and energy/carbon risk into one number before they bid.

02 · The Numbers

Six figures that frame the year

$1.26T

debt-maturity wall, peaking 2027 (~$875B in 2026). S&P

  

$112.6B / +18%

Q1 sales volume YoY; apartments led at $32.1B. Altus
 

4.4%

data-center cap rate — lowest of any class. CRE Daily

  

19% / 83.7%

office vacancy / delinquency on matured office loans. S&P
 

$16.7B → $20B+

PropTech VC 2025→2026; 31 firms took 72%. CRETI

  

~$700B / 60%

AI capex; >60% to power/cooling/build, not chips. DCK

03 · 📊 The Maturity Wall

$1.26 trillion — and 2027 is the peak

CRE & multifamily mortgage debt maturing, by year ($B)

2024
  
 $950B
2026
  
 $875B
2027 ▲
  
 $1.26T

It's a rising slope, not a single bad year — smoothed by extend-and-blend. The danger isn't the total; it's the concentration in assets that can't refinance. Multifamily and industrial roll over with a haircut. Commodity office increasingly doesn't roll over at all.

EDITOR'S TAKE

"Ramp, not cliff" is the new consensus — and consensus is usually half-right. It's a ramp for multifamily and industrial. For commodity office it's still a cliff; the market is just being polite about who walks off it. Don't let the soothing aggregate lull you on a specific Class B tower.

04 · Capital Markets

The wall is a ramp — only if you start climbing

Lending stabilized at ~$583B in 2025 and is projected at ~$805B in 2026 (+38%) — a market re-opening, not bracing for collapse. But the ramp is only gentle for assets with a credible income story.

▸ Hold 2026–2027 maturities? Open the refi conversation 12–18 months early. Lender cooperation is a depleting resource.

▸ Have dry powder? The cleanest entries are recapitalizations of fundamentally-fine assets owned by over-levered sponsors — not heroic office turnarounds.

▸ Underwriting anything? Add two lines to every model: refi-gap risk and energy/carbon risk (§12).

05 · 📊 The Cap-Rate Ladder

Where capital is hottest

Longer bar = hotter sector = more yield compression. Cap rate shown at right.

Data centers
  
 4.4%
Data center (CMBS)
  
 6.5%
Medical office
  
 6.9%

Data centers sit alone at the top. The rest of the market is repricing toward income durability — which is exactly what the next section maps out.

06 · Sector Watch

Where the capital is actually going

Data centers are the asset class of the cycle — and the window is closing. Priced richer than anything else on ~$700B hyperscaler capex (+60%) and record leasing. The live edge isn't owning a stabilized DC — it's controlling the scarce inputs: power, interconnection, cooling.

The binding constraint moved from chips to electricity. Nadella: "GPUs sit idle because we can't find electricity." PJM capacity cleared at $329/MW — 11.4× two years prior. Gas-turbine lead times: 2–3 → 7–8 years.

Office is two assets sharing a name — trophy tightening (A+ occupancy 78.6%), commodity in runoff (19% vacancy). Quiet winners: multifamily (led Q1, 0.5% delinquency), industrial (96.8% occupied), medical office (+78% investment).

📊 Sector Scorecard — capital flow · distress risk · 2026 verdict

Sector Capital Distress 2026 verdict
Data centers HOT LOW Crowded — chase inputs, not assets
Industrial STRONG LOW Core hold
Multifamily STRONG LOW Refi-able, durable
Medical office RISING LOW Demographic tailwind
Retail (necessity) STEADY MOD Selective — anchor matters
Office — Trophy/A+ RETURNING LOW-MOD Tightening — reset basis
Office — Commodity FROZEN SEVERE Convert / recap / hand back

EDITOR'S TAKE

If your portfolio is heavy on the bottom row and you're waiting for "office to come back," stop waiting. Trophy is coming back; commodity is being repriced as land. The fastest way to be wrong in 2026 is to treat "office" as one bet — the green and the red rows above share a name and nothing else.

07 · 📊 Where the AI Money Goes

The 60/40 that reshapes CRE

~$700B of 2026 hyperscaler AI capex, by destination

60%+  Power · Cooling · Construction <40% Compute

The most under-appreciated fact in CRE this year: the AI boom is mechanically a buildings-and-energy boom. The majority of the biggest capex wave in a generation never touches a chip — it pours into the physical stack our industry owns and operates.

EDITOR'S TAKE

This is the chart we'd staple to every IC memo in 2026. The market keeps pricing AI as NVIDIA's story; the cash flow says it's our story. When more than 60% of the largest capex wave in a generation lands in the physical stack, "building intelligence" stops being a soft cost and becomes the asset. This is the single belief that organizes the rest of this brief.

08 · The PropTech Lens

VC is back — but only for AI

After three brutal years, VC returned to PropTech — $16.7B in 2025 (+67.9%), record 2026 pace. The filter: AI or nothing. Money funds products that do the work (autonomous underwriting, lease abstraction, construction robotics — Bedrock Robotics hit a $1.75B unicorn), not "AI-enhanced dashboards."

The structural shift worth your attention: the biggest owners are building their own AI instead of buying PropTech. Blackstone and Brookfield are striking multibillion-dollar deals with Anthropic and OpenAI for custom underwriting/portfolio systems. Counter-signal: CBRE runs Nexus AI across 1 billion sq ft; 68% of institutional investors call AI platforms a primary 2026 acquisition focus.

Honest take: unless you're Blackstone, don't build from scratch or buy ten point tools. Pick a few survivors with real autonomous capability, keep your data portable, and treat your portfolio data as a strategic asset — the giants already do.

09 · 📊 PropTech Concentration

31 firms took 72% of the money

Share of 2025 PropTech VC dollars

72% → top 31 firms 28% → everyone else

The capital came back, but it isn't spread out. Adopt accordingly: the safest tools are the funded survivors — not the clever long-tail startup that may not have a vendor in three years.

10 · 📊 Office, City by City

The geography of return

Peak-day office occupancy, 10-city Barometer (Kastle, mid-May 2026)

Austin
  
89.5%
Dallas
  
72.9%
New York
  
72.0%
Chicago
  
69.7%
Houston
  
69.4%
Washington DC
  
65.9%
San Francisco
  
55.1%
Los Angeles
  
54.9%
San Jose
  
53.3%
Philadelphia
  
51.2%

The return-to-office story is really ten different stories. Sun Belt and Texas lead; coastal gateways lag. Underwrite office occupancy to the metro, never the national average.

EDITOR'S TAKE

The national occupancy average is a number we'd ban from underwriting. Austin and San Jose are 36 points apart — they aren't in the same market, the same cycle, or arguably the same decade. Price the metro, the submarket, and the building. The average is for headlines; you're not buying the average.

11 · Regional Radar

Three markets, three forcing functions

🇺🇸 United States — the divergence capital

Recovery is real but bifurcated; the debt wall and office distress are concentrated here, and so is the data-center gold rush. PJM's 11.4× capacity repricing is the canary for power-cost inflation nationwide.

🇪🇺 Europe — regulation is the forcing function

The recast EPBD transposition deadline passed May 29; nine member states (incl. Germany) are already in infringement proceedings. Building automation is mandatory above 290 kW. Compliance is now a transaction gate, not a virtue.

🌏 APAC — the build-out frontier

Taiwan's inaugural carbon fee collected ~NT$4.5B from ~512–550 emitters; from 2026 every ≥5 MW project must file an energy-use plan pre-construction and hit PUE ≤1.3. Energy efficiency has become a literal permit prerequisite.

12 · The Playbook

Underwrite energy & carbon risk into a 2026 acquisition

Energy and carbon used to be an ESG footnote; in 2026 they're a cash-flow and CapEx line sophisticated buyers already price into bids — and the seller's broker hopes you won't. Five moves before you sign:

1. Model carbon penalty on the future caps. NYC LL97 = $268/ton CO₂e over the limit, annually (>25,000 sq ft). The 2030–2034 caps bite harder. Gap × $268 = a recurring NOI haircut to capitalize.

2. Treat rising power rates as your base case. Stress the energy line +20–40% and see if the deal still works.

3. Price the retrofit runway into the bid. A serious HVAC/controls retrofit takes 18–36 months — if the asset will be over its 2030 cap, that's a 2026–2027 CapEx decision.

4. Distrust every "green" claim in the OM. Demand IPMVP M&V with a weather-normalized 12-month baseline meeting ASHRAE Guideline 14 fit before any savings number. No M&V = marketing, not underwriting.

5. Credit the upside: grid flexibility is a revenue line. A building that shifts ~20% of load off-peak can earn demand-response revenue; a BMS that speaks OpenADR is optionality to price in your favor.

Bottom line: the energy-and-carbon line is the most mispriced variable in CRE underwriting right now. Getting it right is a genuine edge over the buyer who still treats it as a footnote.

13 · Contrarian Corner

The hottest asset in CRE is a substation

Electrical substation transformers at golden hour

Everyone is fighting over the debt wall and the office discount. Meanwhile the real repricing is happening one layer down, in the thing nobody used to underwrite: access to power.

When a data center prices at a 4.4% cap, you're not paying for concrete — you're paying for a megawatt already connected to a grid that can't add new ones fast enough (gas-turbine lead times: 7–8 years; PJM capacity: +1,040% in two years). The scarce, defensible, appreciating asset isn't the building. It's the interconnection. Powered land, on-site generation, and certified demand-flexibility are becoming the highest-margin position in the stack.

The contrarian bet: in 2026 the smartest CRE capital stops asking "what can I build here?" and starts asking "what can I power here?" — and the buildings follow the electrons, not the other way around.

14 · Tools & The Signal

Three tools, one signal

IPMVP (EVO) — the M&V framework behind the Playbook, free. Honest take: dense, but the language your engineer, vendor, and lender should all speak when money rides on a savings number.

Brokerage AI platforms (CBRE Nexus AI, JLL, Cushman) — portfolio-scale analytics. Honest take: useful for benchmarking if you're a client; remember you're feeding their training set.

Fault Detection & Diagnostics (FDD) (category). Honest take: the most reliable, lowest-drama win in building intelligence — it surfaces failures that quietly bleed NOI. If you're early, start here, not with a moonshot.

The Signal — track, not yet urgent

The energy-intelligence layer is going from optional to mandatory. The analysis that used to be a nice-to-have is becoming the artifact you must produce to transact, permit, or refinance. The owners who instrument early won't just run cheaper buildings — they'll be the ones who can prove it when a buyer, lender, or regulator asks.

15 · 📅 The Lookahead

Dates that move the market

When What — and why it matters
Jul 1, 2026Germany EnEfG data-center heat-reuse 10% mandate enters force.
Aug 29, 2026NYC LL97 final filing deadline (with extension) — first real penalty exposure crystallizes.
Q3 2026Debt-wall pressure builds toward the 2027 peak — refi window narrows.
2027EU SRI delegated act for HVAC >290 kW; debt-maturity wall peaks at $1.26T.
2030LL97 step-down caps bite; EU requires all new buildings zero-emission. The retrofit clock is already running.

16 · OUR CONVICTION CALLS

Where the digest ends and our opinion begins

Four calls we'd put our name on this week — with how strongly we hold them. These are opinions, not facts. Hold us to them.

HIGH In data-center-adjacent CRE, power access has overtaken location as the primary value driver. Underwrite the interconnection, not the dirt.
 
HIGH The energy-and-carbon line is the most mispriced variable in CRE underwriting today. The spread between buyers who price it and buyers who don't is where 2026's best risk-adjusted returns live.
 
MED-HIGH Plug-and-play PropTech gets squeezed from both ends — giants building in-house, capital concentrating in 31 winners. Bet on survivors; keep your data portable; the schema you lock into today is the lock-in you regret in 2028.
 
MEDIUM Trophy office is a genuine buy at a reset basis; commodity office is a value trap dressed as a discount. The gap between them widens from here, not narrows.

WHERE WE MIGHT BE WRONG

If rates fall faster than expected, the refi ramp gets gentle enough that even commodity office muddles through — and our office-bifurcation call softens. We're watching the 10-year Treasury, not just the headlines. If it breaks decisively below where it sits today, ping us — we'll change our minds in print.

The Intelligent Building Brief is free, and that's the point. If it earned twelve minutes of your week, forward it to one colleague who'd say the same. That's the entire growth plan.

— ai-smart-buildings.com