Six issues tracking the capital stack, supply waves, regulatory moves, and the real tenant ledger across Sunbelt multifamily.
State capitalism, AI compute economics, policy platforms, and why the swing vote in every market is the tenant's paycheck.
Both sides of the rent roll: buying right is what lets you charge less, and operating discipline beats headline sentiment.
Where in the capital stack buildings actually break: floating-rate maturities, debt yield covenants, and the 65% signal.
Reading beneath national rent averages: why aggregate indexes mask submarket dispersion, concessions, and micro-location strength.
Bank fragility, extend-and-pretend exhaustion, supply cliff mechanics, and 16 Sunbelt markets where the acquisition window opened.
12 metro areas, 49 submarkets, 18,752 properties. The delivery peak schedule and fundamental underwriting reality across the Sunbelt.
Two countries, two kinds of AI, and the tenants caught in between.
An apartment owner watches the AI trade for one reason: it decides what our tenants can pay. If the trade corrects, companies cut costs, labor is the biggest line, and the cheapest model that does the job wins. The Capitalists are building the expensive machine. The Communists may be selling the cheap one.
American private capital put $285.9 billion into AI in 2025, more than 23 times China's $12.4 billion. As of March 2026, Stanford's AI Index put the best American model 2.7% ahead of the best Chinese one and called the gap effectively closed. Stanford HAI, AI Index 2026 Yet a leading Chinese model sells its output for $3.96 per million tokens; comparable American models list at $10 to $20. DeepSeek; Anthropic
Read the jerseys closely. The 'Capitalists' now own 9.9% of Intel Bloomberg Law, and the 'Communists' say private firms produce most of their economy. China Daily / Caixin (official line) Neither team plays the position on its jersey. The 2026 game is two kinds of state capitalism: in one, private capital leads and the state backs it; in the other, the state leads and private capital follows. We'll skip the verdict and show you the ledger.
Following our look at tenant balance sheets in Issue 05, this issue is about who pays: who borrowed, who got rescued, who got the raise, and who is left holding the rent. Our answer, as operators, is simple. When the tenant wins, the operator wins. We'd rather build more homes and see paychecks rise than cap rents and starve the buildings.
We gave five models the same 73 tasks from our own pipeline. The cheapest scored as well as Claude, at a fraction of the cost.
On general tests the Capitalists still lead. Artificial Analysis scores Anthropic's Claude Opus 5.5 at 58 on its intelligence index; the best Chinese model, Xiaomi's MiMo-V2.6-Pro, scores 46. Artificial Analysis On LMArena, where people vote blind between two answers, US labs hold all 15 top spots, and the best Chinese model, Moonshot's Kimi K3, is 16th. LMArena
Switch to models anyone can download and run, and the order flips. The nine best open-weights models on Artificial Analysis are all Chinese. The best American one ranks 20th and scores 26, against MiMo's 46. Artificial Analysis
Closer to our work, Vals AI tests models on finance paperwork. On scanned mortgage tax certificates, Claude Opus 5 scores 72.1% at $25 per million output tokens. MiniMax-M3 scores 68.4% at $2.40, about a tenth of the price, and Alibaba's Qwen 3.5 Flash scores 67.4% at $0.40, a sixtieth of it. Vals AI On long credit agreements, Claude Opus 5 scores 73.2%. Kimi K3 scores 71.6% at three-fifths of Claude's output price, DeepSeek V4 Pro 65.4% at about a sixth, and the cheap American option, Thinking Machines' Inkling Small, 69.6% at under a twentieth. Vals AI Cheap is not only Chinese.
Developers vote with tokens and with dollars, and the two votes disagree. In the last week of September, DeepSeek drew 21.8% of all requests on OpenRouter, more than any other lab; Anthropic drew 2.5%. OpenRouter Follow the money and Claude leads. In data extraction, the task closest to pulling numbers from an offering memo, Claude models took 14.9% of spend among the ten biggest earners and Chinese models 7.4%. OpenRouter
Our own test said the same thing. On October 5 we ran the same 73 tasks, modeled on our own pipeline and built on made-up sample deals, through five models: screening deals, calling tools, checking data, looking up procedures, reading documents, writing code. Claude Opus 5.5, the most expensive of the five, got 65 right. So did Claude Sonnet 5.5 and DeepSeek V4.1 Flash. Xiaomi's MiMo-V2.6-Flash got 66, at 1/44th of Opus's cost per task. The whole test cost 46 cents, and Opus was 29 of them. Coral Reef Capital internal test Read that as a tie on routine work. Most of our pipeline is routine, and most models aced most of it. The gap shows up on hard reasoning, like the 58-to-46 spread above, and that is where we pay up. The cheap models were also slower, 6 to 12 seconds a task against Claude's 3 to 5, which matters when a person is waiting and not at all for an overnight batch. Coral Reef Capital internal test
Every model, Claude included, slipped in the same place: answering procedure questions without the documents in hand, 4 or 5 right out of 10. Coral Reef Capital internal test That is a lesson about memory, not money. So the rule is simple: use the cheap model for the volume, give every model the documents, and let plain code do the math. Cheap does not have to mean sending deal data overseas: several of these models can be downloaded and run in-house, or served by US hosts. Artificial Analysis Running one in-house is not free, though: it takes servers and engineering time, so the per-task prices above are cloud prices.
| Test | Capitalists | Communists |
|---|---|---|
| Mortgage tax certificates Vals AI, Sep 2026 | Claude Opus 5: 72.1% $5 / $25 | MiniMax-M3: 68.4%, $0.60 / $2.40 Qwen 3.5 Flash: 67.4%, $0.10 / $0.40 10.4x and 62.5x cheaper Vals AI |
| Long credit agreements Vals AI, Aug 2026 | Claude Opus 5: 73.2% $5 / $25 | Kimi K3: 71.6%, $3 / $15 DeepSeek V4 Pro: 65.4%, $1.32 / $3.96 1.7x and 6.3x cheaper Vals AI |
| Our test, 73 pipeline tasks Coral Reef Capital, Oct 5, 2026 | Claude Opus 5.5: 65 of 73 $0.0039 per task | MiMo-V2.6-Flash: 66 of 73, a tie $0.00009 per task 44x cheaper Coral Reef Capital internal test |
| General intelligence index Artificial Analysis, Oct 2026 | Claude Opus 5.5: 58 | MiMo-V2.6-Pro: 46 Artificial Analysis |
| Share of requests OpenRouter, week of Sep 28 | Anthropic: 2.5% | DeepSeek: 21.8% OpenRouter |
| Share of data-extraction spend OpenRouter top ten, Oct 2026 | Claude models: 14.9% | Chinese models: 7.4% OpenRouter |
Prices are list prices per million tokens (input / output); "cheaper" compares output prices, as shown by each scoreboard. Vals tested Claude Opus 5; Artificial Analysis lists Opus 5.5. Scoreboards are private companies; each one measures something different. Our own test is internal and not independently verified: one run per model, default settings, 6 to 11 tasks per group.
In practice the cheaper model wins most jobs. We route each one by what it needs: routine work to cheap models, judgment to the strongest, and every dollar figure to plain code, never a model. And nothing sensitive leaves the building: no tenant data, investor data or private deal documents go to any cloud model, cheap or expensive.
Six numbers to know this month. The other 69 are in the Full Tape at the back.
We have no rent control. We had a building boom, and it did what rent control promises.
Texas, North Carolina, Georgia and Tennessee all bar local rent control by state law; Texas allows it only in a governor-declared disaster. N.C. General Statutes; State statutes The lever our metros have is supply, and they pulled it: a record wave of apartment building, then rents flat or falling.
The wave is over. The delivery peak we mapped in Issue 01 is giving way to the permit cliff we tracked in Issue 02: units permitted in 5+ unit buildings are down from peak in every one of our core metros: Nashville −60%, Austin −54%, Atlanta −48%, Raleigh −44%, Charlotte −43%, Houston −40%, DFW −24%. U.S. Census Bureau Add 7%+ mortgages and +10% materials Treasury / Freddie Mac / Fed; BLS, and the next squeeze may be building, if demand holds. Fewer deliveries do not raise rents on their own; absorption and jobs decide that.
Below, the housing plans on the ballot this fall, sorted by what they do rather than who proposes them. We don't endorse candidates. We read plans.
Austin, the hardest-hit metro, is turning: effective rent rose 1.8% in the second quarter to $1,316 after twelve straight quarterly declines, and vacancy fell to 12.3% from a 15.8% peak. Northmarq via CRE Daily Nationally, wage growth has outpaced apartment rent growth for 38 straight months, by JPI and RealPage data shown at a Dallas Fed conference. Dallas Fed conference (JPI / RealPage data) The supply wave gave renters a raise.
| Lever | Who proposes it this fall | What it does to rent |
|---|---|---|
| Build more, faster | James Talarico (D, Texas Senate nominee): "cut regulation so homes build faster"; House author of the small-footprint apartment bill (SB 2835) Texas Tribune; Talarico campaign Abdul El-Sayed (D, Michigan Senate nominee): by-right multifamily, lower parking minimums, firm permit timelines, more LIHTC El-Sayed campaign Mike Rogers (R, Michigan Senate nominee): tax-free residential building zones Newsweek Zohran Mamdani (NYC mayor): 200,000 affordable homes over 10 years; $5.6B for public housing NYC Mayor's Office; NYC Mayor's Office | Lowers rent for the next renter. |
| Raise paychecks, cut household costs | Talarico: higher minimum wage and overtime; a "cost-of-living tax cut"; universal childcare Texas Tribune Mamdani: universal childcare and free buses (campaign cost estimates ~$6B and ~$800M a year) CNN El-Sayed: tax billionaire wealth El-Sayed campaign | Raises what tenants can pay without touching the building. |
| Help buyers | Ken Paxton (R, Texas Senate nominee): a $50,000 first-time-buyer deduction plus an extra $50,000 down-payment break Texas Tribune Rogers: 529 savings for down payments Newsweek Talarico: down-payment aid Texas Tribune | Adds buying power without adding a home. |
| Cap rents | Mamdani: 0% on 1- and 2-year stabilized leases from Oct 1, 2026 (~1M units), passed 7 to 1, landlord suit pending CRE Daily / PoliticsNY; the board's own index shows operating costs up 5.3% this year and a projected 4.1% next NYC Rent Guidelines Board; NYC Rent Guidelines Board; NYC Rent Guidelines Board | Helps today's tenant; squeezes building budgets. Barred in our four states, except in a Texas disaster declaration. |
| Restrict owners and pricing tools | El-Sayed: ban large corporations from owning homes and ban algorithmic rent-pricing software Newsweek | The software ban matches our view (see Special Teams). |
Supply lowers rent for the next renter. A raise lowers the burden without touching the building. A freeze helps today's tenant and starves the boiler: New York's own cost index rose 5.3% the year stabilized rent increases went to zero. In our states a freeze isn't on the table anyway. Homebuyer deductions and 529s add buying power without adding a home.
How far permits fell in each of our seven metros, and what share of local jobs a cheap model could reach.
| Market | 5+ Unit Permits, Peak | 2025 | Change | White-Collar Share |
|---|---|---|---|---|
| Austin | 25,642 (2021) | 11,749 | −54% | 38.3% |
| DFW | 32,242 (2022) | 24,607 | −24% | 34.4% |
| Houston | 27,124 (2022) | 16,385 | −40% | 29.5% |
| Charlotte | 10,127 (2023) | 5,820 | −43% | 31.2% |
| Raleigh | 9,846 (2022) | 5,542 | −44% | 33.4% |
| Atlanta | 21,288 (2022) | 11,052 | −48% | 33.2% |
| Nashville | 14,478 (2021) | 5,827 | −60% | 31.1% |
Permits: Census Building Permits Survey, units in 5+ unit buildings U.S. Census Bureau; 2024 metro redefinitions re-bounded Austin, Houston and Atlanta. White-collar share: BLS OEWS, May 2025 (management, business and financial, computer and math, legal, and office support) BLS OEWS.
$285.9 billion against $12.4 billion bought the US a 2.7% lead.
The private spending gap is real: $285.9 billion of US private AI investment in 2025 against $12.4 billion in China. Stanford itself adds that private figures likely understate China's total, because Beijing also spends through government guidance funds. Stanford HAI, AI Index 2026 So is the performance gap, what's left of it. US and Chinese models have traded the lead since early 2025; as of March 2026 the top US model leads by 2.7%. Stanford HAI, AI Index 2026
The price gap runs the other way. At peak rates, a leading Chinese model charges $1.32 per million input tokens and $3.96 per million output tokens. Comparable US models list at $2 to $4 input and $10 to $20 output: 2.5 to 5 times more for the output. DeepSeek; Anthropic The Capitalists lead the race to build the machine. The Communists may win the race to sell it cheap.
Since the cover invites it: free-market capitalism, state capitalism, social democracy, democratic socialism and communism each answer two questions differently. Who owns the capital, and who sets the price? Two more systems never appear on a ballot. Corporate socialism keeps the gains private and makes the losses public. Crony capitalism buys the rules instead of winning them.
Washington owns 9.9% of Intel. Beijing says private firms produce most of its economy.
The state is buying in. In 2025 Washington took 9.9% of Intel Bloomberg Law, became the largest shareholder of a rare-earth miner Yahoo Finance, and used a "golden share" to stop U.S. Steel from closing an Illinois plant. WSJ via Adnkronos
It finances our industry. Government-sponsored enterprises and agency securities hold $1.2 trillion of US multifamily mortgage debt, or 50% of it. Mortgage Bankers Assoc. via Scotsman Guide
It pays part of the rent and builds part of the supply. 2.3 million households use Housing Choice Vouchers NLIHC; the tax-credit program had financed 3.3 million units through 2021. HUD
It insures the coast. Texas's state-created insurers of last resort cover 288,676 windstorm policies and 118,486 more properties. TWIA / Texas FAIR Plan
The fair label: state capitalism, with private capital in front.
China's official statistics are hard to verify independently; read these as Beijing's own numbers.
Private enterprise runs most of the economy. Beijing's own line: private firms supply more than 50% of taxes, 60% of GDP, 70% of technical innovation and 80% of urban jobs. China Daily / Caixin (official line)
It mints billionaires. How many depends on who counts. Hurun puts the US at 1,000 and China at 1,110 Hurun Report; Forbes puts the US at 989 and China at 539. Forbes China has 4.1 times the people, so per person the US leads either way, by 3.7 to 7.6 times. World Bank
The state keeps the leash. Regulators halted Ant Group's record IPO in 2020 after summoning its founders. China Daily
And it ran the housing experiment. In a 2019 central bank survey, before the bust, urban homeownership was near 96%, with about 70% of household assets in property. PBOC via Yicai Then the bust: the largest developer ordered liquidated with more than $300 billion of liabilities BusinessDay (wire); real-estate investment down 17.2% in 2025 alone, new home sales area down 8.7%. China National Bureau of Statistics
The fair label: state capitalism, with the state in front.
Before anyone in our industry says 'socialism,' look at the cap table on our own loans. Half the multifamily debt in America sits with Fannie, Freddie and other government-backed agencies. We're not purists either.
Both parties borrow. They fight over who gets the money.
Ten-year borrowing approved under President Trump's first term: $8.4 trillion, or $4.8 trillion excluding COVID relief. Under President Biden through mid-2024: $4.3 trillion, or $2.2 trillion excluding the American Rescue Plan. CRFB (Committee for a Responsible Federal Budget) Most of it passed with both parties' votes. The Bipartisan Budget Acts of 2018 and 2019 added $2.1 trillion and the CARES Act $1.9 trillion. CRFB The 2025 tax-and-spending law adds $3.4 trillion over 2025 to 2034, by CBO's estimate of the law as enacted. CBO / CRFB
Where they differ is the instrument. The 2017 tax law, passed on party lines, added $1.9 trillion. CRFB Executive actions added $1.2 trillion under Biden and less than $20 billion under Trump. CRFB Both borrowed heavily. The winners change.
And when something breaks, the order of rescue doesn't change. The savings-and-loan cleanup cost $160.1 billion, direct and indirect, by GAO's audit, paid by taxpayers and the industry. FDIC TARP disbursed $443.5 billion and cost $31.1 billion after repayments. GAO / Treasury In both rescues, the first and largest checks went to lenders and asset holders.
The scare words come from both sides too: "socialism" from one, "greedy landlords" from the other. Both are labels, and the labels are moving: 55% of Americans view capitalism positively and 43% now say the same of socialism, the first time it has topped 40% in Gallup's trend. Independents moved from 37% to 45% in a year. Gallup Vice President JD Vance put it to his own party in August: "don't blame young people for being sympathetic to socialism. We've got to blame ourselves." Fox News (Will Cain Show) The ledger is the thing to read.
Workers produce more each hour. Pay has not kept pace, and rent comes out of pay.
Since 1979, output per hour in the nonfarm business sector is up 141.9%. Average real hourly compensation is up 54.8%. The median full-time worker's real weekly pay is up 12.8%. Labor's share of output fell 15.6%. BLS; BLS (Output and pay are measured with different price indexes, so the exact size of the gap is debatable. The direction isn't.)
The 1968 federal minimum wage was $1.60, or $15.63 in today's dollars. Today it's $7.25. DOL / BLS
RAND estimates the bottom 90% earned $2.5 trillion less in 2018 alone than they would have if income had grown as evenly as it did in the two decades after the war (67% more), and $47 trillion less from 1975 to 2018. RAND (WR-A516-1)
Where it went: since 1989 the top 1%'s share of US wealth rose from 22.8% to 32.5%, and the top 0.1%'s from 8.6% to 15.0%. The 50th to 90th percentiles fell from 35.7% to 28.8%; the bottom half from 3.5% to 2.3%. Federal Reserve (DFA via FRED) The top 1% now hold more than the middle 40%.
Two economies, one ticker. The top 10% own 88.1% of household stocks and funds Federal Reserve (DFA via FRED); the ten largest companies are about 38% of the S&P 500 ChartRow / Apollo. Consumer sentiment set an all-time low of 44.8 in May 2026. University of Michigan One honest caveat: the spending split is less extreme than the headlines. The Census survey puts the top 10% at 23% of spending, well under the 45%+ often quoted. Minneapolis Fed (citing Moody's); Minneapolis Fed The wealth gap is the real K.
Some of the missing raise didn't go to profit at all. The average family health premium paid through an employer hit $26,993 in 2025, up 6% in a year, against 4% for wages. KFF (Employer Health Benefits Survey) And some went to shareholders: S&P 500 companies bought back a record $1.02 trillion of their own stock in the twelve months to September 2025. S&P Dow Jones Indices
And rent sits on top of it: shelter prices are up 123.8% since 2000, against 96.3% for all consumer prices, through April 2026. BLS
The national accounts say the same thing in one quarter. After-tax corporate profits hit 13.1% of GDP in Q2 2026, the highest since the series began in 1947, up 26.6% in a year. BEA The same quarter, labor's share of nonfarm business output hit its lowest on record. BLS
Rates, tariffs, immigration, property tax, pricing software, insurance and credit. Seven levers set most of a renter's bill, and none of them makes a stump speech.
Rates: the Fed and the bond market. The effective fed funds rate is down 0.47 points in a year to 3.75%, yet the 10-year Treasury rose 1.12 points to 5.24% and the 30-year mortgage to 7.28%. Treasury / Freddie Mac / Fed Then on September 16 the Fed raised its target to 3.75% to 4.00% and projected 4.1% by year-end. Federal Reserve Board The refi wall from Issue 04 (where we examined maturities breaking assets before operations do) just got steeper.
Tariffs: Congress and the White House. Construction materials are up 10.1% in a year and 61.1% since January 2020; lumber up 7.6%. BLS
Immigration: Washington. Net international migration fell 53.8% in a year, from 2.73 million to 1.26 million: Texas −52.8%, Georgia −50.8%, North Carolina −43.4%, Tennessee −62.5%. U.S. Census Bureau That means fewer renters at the bottom of the market and fewer trades on the job site; specialty-trade employment grew just 1.1%. BLS
Property tax: statehouses. Texas spent $18 billion on relief in 2023 and budgets $51.0 billion for 2026 to 2027; the headline piece raised the school homestead exemption from $40,000 to $100,000, and it now stands at $140,000. Texas Comptroller Apartments got rate compression; the 20% circuit breaker stops at $5 million properties. State of Texas Renters pay property tax through rent, and got the smaller share.
Rent-pricing software: DOJ and Congress. In November 2025 DOJ reached a proposed settlement with RealPage that would end its software's use of competitors' nonpublic, competitively sensitive data to set prices. U.S. DOJ DOJ is now settling with landlords one at a time, including Dallas-based Willow Bridge in July 2026, the fifth defendant to settle. U.S. DOJ Our position: ban it. Software that pools rivals' private rents distorts the real submarket dispersion we tracked in Issue 03; pooling data so everyone prices off the same sheet is a cartel with a login. Operators should compete on price and product, building by building. That puts us on the same side of this one as a Democratic Senate candidate who would ban it outright. Newsweek The ledger doesn't care about the label.
Insurance: statehouses and the market. Insurance cost per apartment rose from $502 in 2021 to $777 in 2024, +55%, with 2023 alone up 25%. National Apartment Association A Minneapolis Fed survey of 35 owners in four Upper Midwest states found premiums up 45% from 2023 to 2024, and higher deductibles and exclusions moved more of the risk onto owners. Federal Reserve Bank of Minneapolis The market has started to turn: commercial property premiums fell 6.3% on average in the second quarter of 2026. CIAB via Insurance Journal
Credit: FHFA. Washington raised Fannie Mae's and Freddie Mac's apartment-loan caps to $88 billion each for 2026, $176 billion in all, from $146 billion. FHFA The ceiling on what they can buy rose 20% in the year rates rose.
CRC raises money and borrows to buy buildings. Here is where that puts us.
Coral Reef Capital raises capital and uses leverage. The line we draw runs between capital that improves the building and capital that only extracts from it.
A renewal beats a turn: no make-ready, no vacant days, no concession. Better units at rents people can pay is the business. A rent cap lands on good and bad operators alike. The fix for the extractor is competition: more supply, so tenants can leave, and no shared pricing software, so operators actually compete for them.
Who owns the buildings? Individual investors own 37.6% of US rental units, most of them the small ones. LLCs and partnerships own 40.4%, and two-thirds of units in properties with 150 or more apartments. REITs and real-estate corporations own just 4.3%. Congressional Research Service The 50 largest apartment owners in the country hold 11% of the stock between them. NMHC Extraction is a way of operating, and no one type of owner has a monopoly on it.
Most renters would rather own. In the New York Fed's 2026 survey, 64.5% of renters said they would prefer owning if they could afford it, and on average they put their odds of ever owning at 34.7%. Federal Reserve Bank of New York That gap is who is left holding the rent.
The tenant's paycheck is our underwriting. If pay keeps up, the value-add premium is affordable. If it doesn't, no amount of renovation fixes the rent roll.
Three ways the AI trade can end, and what each one means for our tenants.
Four companies plan about $725 billion of capital spending in 2026 company guidance via AI Weekly, chasing a model lead of 2.7% Stanford HAI, AI Index 2026. Big Tech laid out $294.8 billion in the first half of 2026 alone, up 84%. SEC filings (XBRL) In an Atlanta Fed survey of nearly 6,000 executives, about nine in ten reported no AI impact on employment or productivity over three years, while expecting just 1.4% productivity growth ahead. NBER / Atlanta Fed Cheaper AI is a demonstrated purchasing option in our workflow; its effect on tenant employment remains an underwriting risk to monitor, not a foregone conclusion.
AI is already cited in 120,136 announced job cuts through September, about 21% of the total. Challenger, Gray & Christmas In our markets, selected white-collar occupations make up 38.3% of jobs in Austin, 34.4% in DFW, and about a third in Raleigh and Atlanta. BLS OEWS Rather than assuming a single chain of events, we underwrite to three distinct scenarios:
Firms aggressively substitute cheap models for white-collar staff to protect margins as debt rolls over.
Signals to watch: Tech and finance layoff acceleration, rising professional continuing claims, rising Class A concessions.
Underwriting response: 0% untrended rent growth, assume 4 to 6 weeks concession on turns, increase bad-debt reserve to 3.5%, underwrite exit cap +50 bps.
Models augment workers; output expands; companies retain talent and real wages continue beating rent growth.
Signals to watch: Sustained real wage growth, stable or falling white-collar unemployment, sub-30 days on market.
Underwriting response: Standard baseline rent growth tracking local wage growth (2% to 3%), execute premium unit renovations, maintain standard 1% bad-debt reserve.
Big Tech slashes data-center and hardware capex; cloud prices collapse; broad economy absorbs the hit without broad job cuts.
Signals to watch: Tech capex guidance cuts, data-center construction slowdown, semiconductor inventory build.
Underwriting response: Isolate exposure in peripheral data-center submarkets; core Sunbelt infill apartments remain insulated; benefit from cheaper software tools.
And our industry is in the chain too. In a survey of 280 multifamily executives, 82% expect AI to replace several traditional property roles within a year. EliseAI survey via Multifamily Executive (The survey was sponsored by a vendor selling AI leasing bots; read it accordingly.)
The proposals on the table range from an "AI dividend" that gives every citizen a stake Texas Tribune to taxing billionaire wealth El-Sayed campaign to nothing at all.
Whoever builds the smartest machine, the matchup that matters to operators is who shares in the gains and who absorbs the cuts. A tenant who keeps a paycheck renews.
Eight real policies. Guess the country, then check.
United States: Intel, August 2025. Bloomberg Law
China: Ant Group, November 2020. China Daily
United States: U.S. Steel, invoked September 2025. WSJ via Adnkronos
China: the government's own figures. China Daily / Caixin (official line)
United States: Texas, 288,676 windstorm policies plus 118,486 more properties. TWIA / Texas FAIR Plan
United States: 50% of $2.32 trillion. Mortgage Bankers Assoc. via Scotsman Guide
United States: the Pentagon and MP Materials, July 2025. Yahoo Finance
China: Evergrande, January 2024. BusinessDay (wire)
Score fewer than six? That's the point. The jerseys say Capitalists vs. Communists; the plays say both teams run a mixed offense.
All 75 figures in this issue, grouped by topic, each with its source.
| Figure | Source & date | Why it matters |
|---|---|---|
| 10-Yr Treasury: 5.24% | Treasury via FRED, Oct 1, 2026 Treasury / Freddie Mac / Fed | Up 1.12 points in a year; the 2026 high (5.29%, Sep 30) is the highest since May 2002. The rate the refi wall rolls into. |
| 30-Yr Mortgage: 7.28% | Freddie Mac PMMS, Oct 1, 2026 Treasury / Freddie Mac / Fed | Up 0.98 points in a year. Keeps would-be buyers renting. |
| Fed Funds: 3.75%–4.00% | FOMC, Sep 16, 2026 Treasury / Freddie Mac / Fed; Federal Reserve Board | Raised 12–0; median 4.1% at year-end and 4.1% in 2027. Effective rate still down 0.47 in a year. |
| 10-Yr Term Premium: 1.02 pts | Fed Board via FRED, Sep 25, 2026 Treasury / CBOE / Fed Board | Highest since April 2010, up from 0.50 a year ago. Bond buyers want paid for deficit and inflation risk. |
| 2s10s Spread: +0.45 pts | Treasury via FRED, Oct 2, 2026 Treasury / CBOE / Fed Board | The curve is steepening: long money costs more than short money again. |
| VIX: 16.39 | CBOE via FRED, Oct 1, 2026 Treasury / CBOE / Fed Board | Stocks are calm while long rates sit above 5%. |
| Core PCE Inflation: 3.0% | BEA via FRED, Aug 2026, y/y BEA / BLS | Why the Fed hiked: a point above target. |
| Maturing CRE Debt in 2026: $875B | MBA, Feb 2026 Mortgage Bankers Association | 17% of $5.0T outstanding; 13% of multifamily mortgages mature this year, into a 5%+ 10-year. MBA counts all lenders, a wider net than the institutional-debt count we cited in Issues 01 and 04. |
| Multifamily CMBS Delinquency: 7.69% | Trepp via CRE Daily, Aug 2026 Trepp via CRE Daily | Up from 6.86% a year ago. |
| Freddie Mac Apartment Loans 60+ Days Late: 0.64% | Freddie Mac, Aug 2026 Freddie Mac | Up from 0.48% a year ago and 0.60% in July. Agency loans are slipping as well as CMBS. |
| Multifamily CMBS Distress: −123 bps y/y | KBRA, Aug 2026 KBRA | The other read: distress down from 11.6% a year ago, and −73 bps in August alone. |
| Fannie + Freddie Apartment Caps: $176B | FHFA, 2026 FHFA | $88B each, up from $146B: a 20% higher ceiling on agency apartment-loan purchases. Half must be mission-driven. |
| Figure | Source & date | Why it matters |
|---|---|---|
| Apartment Permits, Our 7 Metros: −24% to −60% | Census BPS, 2025 vs. each metro's peak U.S. Census Bureau | Nashville −60%, Austin −54%, Atlanta −48%, Raleigh −44%, Charlotte −43%, Houston −40%, DFW −24%. |
| 5+ Unit Completions: −57% from Peak | Census via FRED, Aug 2026 U.S. Census Bureau | 302K annualized vs. 705K in Aug 2024; −35.7% in a year. The delivery wave that held rents down is ending. |
| Apartments Under Construction: 666K | Census via FRED, Aug 2026 U.S. Census Bureau | −33% from the July 2023 peak of 1.0M. Fewer units coming; whether rents tighten depends on demand. |
| 5+ Unit Starts: 344K, −15.5% y/y | Census via FRED, Aug 2026 U.S. Census Bureau | The pipeline is still shrinking. |
| 5+ Unit Permits (US): 467K | Census NRC, Aug 2026 U.S. Census Bureau | Total permits 1.394M, +3.5% y/y: national permits are turning up while our metros sit far below peak. |
| All Housing Completions: −27.1% y/y | Census NRC, Aug 2026 (±8.9%) U.S. Census Bureau | 1.128M annualized, down from 1.548M. |
| Data Centers +73.2%; Apartments −0.6% | Census construction spending, Aug 2026, y/y U.S. Census Bureau | Data centers $85.0B annualized vs. $49.1B a year ago; new multifamily $115.8B, −19.1% from its June 2023 peak. |
| Factory Construction: −19.8% y/y | Census construction spending, Aug 2026 U.S. Census Bureau | $168.2B annualized. The CHIPS-era factory boom is fading. |
| Total Construction Spending: −1.7% y/y | Census, Aug 2026 U.S. Census Bureau | $2.203T annualized. Outside AI, building is shrinking. |
| Construction Materials: +10.1% | BLS PPI, Aug 2026, y/y BLS | +61.1% since January 2020; lumber +7.6% in a year. |
| Steel Mill Products +23.4%; Copper Mill Shapes +20.9% | BLS PPI via FRED, Aug 2026, y/y BLS | Tariffed metals in every building's frame, wiring and plumbing. |
| Tariffs: ~$30B Added to Housing Construction | Brookings / Tax Policy Center Brookings / Tax Policy Center | About 90% falls on new homes, including apartments. |
| Construction Jobs: 8.36M, a Record | BLS via FRED, Sep 2026 BLS | +1.3% y/y; job openings +17.6%; hourly pay $41.69, +4.2% vs. +3.0% for all private workers. |
| Foreign-Born Labor Force: +0.6% y/y | BLS CPS via FRED, Sep 2026 (not seasonally adjusted) BLS | 32.28M, 1.44M below the March 2025 high. Fewer new renters and fewer framers. |
| Net International Migration: −53.8% | Census Vintage 2025, 2024 → 2025 U.S. Census Bureau | 2.73M to 1.26M. Texas −52.8%, Tennessee −62.5%. |
| New-Home Supply: 8.5 Months | Census NRS, Aug 2026 Census / HUD | Median new-home price $393,700. Builder incentives compete with our renewals. |
| Figure | Source & date | Why it matters |
|---|---|---|
| Wages Beat Rents 38 Months Straight | Dallas Fed conference (JPI / RealPage data), 2026 Dallas Fed conference (JPI / RealPage data) | The supply wave gave renters a raise. The best case for supply over control. |
| National Rent: $1,388, −0.4% y/y | Apartment List, Sep 2026 Apartment List | Vacancy 7.0%; 34 days on market. Renters have leverage. |
| Asking Rent: $1,773, +0.4% y/y | Yardi Matrix, Aug 2026 Yardi Matrix (vendor data) | First August gain in years. Austin −2.8%, Houston −1.7%. |
| Effective Rent: +0.9% y/y | RealPage, Aug 2026 (vendor data) RealPage (vendor data) | Eight straight monthly gains in 2026; Charlotte and Houston ~−2% y/y. |
| Austin: +1.8% q/q After 12 Down Quarters | Northmarq via CRE Daily, Q2 2026 Northmarq via CRE Daily | $1,316 effective; vacancy 12.3% from a 15.8% peak; still −3.9% y/y. |
| Rental Vacancy: 7.3% | Census HVS, Q2 2026 U.S. Census Bureau | Highest since Q3 2017. Homeownership 65.0%. |
| CPI Rent: +2.75% vs. All Items +3.35% | BLS via FRED, Aug 2026, y/y BEA / BLS | Rent is rising slower than everything else; tenants' other bills are the squeeze. |
| Unemployment: 4.2% | BLS via FRED, Sep 2026 BLS / DOL | Low but drifting. Collections track it. |
| Payrolls: +0.3% in a Year | BLS via FRED, Sep 2026 BLS / DOL | +29K in September; +496K over twelve months. |
| White-Collar Services Jobs: +0.55% y/y | BLS via FRED, Sep 2026 BLS / DOL | Professional and business services, the sector AI targets. |
| Job Openings: 7.08M | BLS JOLTS via FRED, Aug 2026 BLS / DOL | Down from 12.3M in March 2022. |
| Initial Jobless Claims: 197K | DOL via FRED, week of Sep 26, 2026 BLS / DOL | Layoffs aren't in the claims data yet. |
| AI-Cited Job Cuts: 120,136 | Challenger, Gray & Christmas, YTD through Sep 2026 Challenger, Gray & Christmas | About 21% of announced cuts, the leading reason year to date. |
| Wells Fargo Headcount: −7% y/y | Wells Fargo Q2 2026 supplement Wells Fargo | 197,466, down 15,338 from 212,804 a year earlier. |
| Consumer Sentiment: 51.7 | University of Michigan, Aug 2026 University of Michigan | The all-time low was this year: 44.8 in May. |
| Median Household Income: $87,460 | Census, 2025 U.S. Census Bureau | +2.6% after inflation, a record. The median did rise. |
| 2025 ACS Rent-Burden Data: Delayed | Census Bureau advisory, Aug 4, 2026 U.S. Census Bureau | A Commerce Department order is holding back every metro's rent-burden tables. The scoreboard is dark. |
| Figure | Source & date | Why it matters |
|---|---|---|
| Private AI Investment: US $285.9B vs. China $12.4B | Stanford AI Index 2026, calendar 2025 Stanford HAI, AI Index 2026 | More than 23 to 1, for a 2.7% model lead. |
| Big Four Capex, H1 2026: $294.8B, +84% | SEC filings (Microsoft, Alphabet, Amazon, Meta) SEC filings (XBRL) | Up from $160.1B in H1 2025. The expensive machine, in dollars. |
| AI's Payoff So Far: ~9 in 10 Firms See None | NBER / Atlanta Fed survey of ~6,000 executives, 2026 NBER / Atlanta Fed | No impact on jobs or productivity over three years; executives expect −0.7% employment ahead. |
| Credit Agreements: Claude 73.2%, DeepSeek 65.4% | Vals AI CorpFin, Aug 2026 Vals AI | Claude Opus 5 vs DeepSeek V4 Pro on long loan documents; DeepSeek costs about a sixth as much per output token; Inkling Small (US) 69.6% at under a twentieth. |
| Requests: DeepSeek 21.8%, Anthropic 2.5% | OpenRouter, week of Sep 28, 2026 OpenRouter | The cheap machine wins on volume. On data-extraction spend, Claude leads 14.9% to 7.4%. |
| Cheapest Chinese Model: $0.60 per 1M Output Tokens | DeepSeek price page, Oct 2026 (off-peak) DeepSeek | Flash model $0.60–1.20 output; V4-Pro $1.98–3.96. The cheap machine's sticker price. |
| US Chips to China: Licensed, Capped at 50% | BIS rule, Jan 13, 2026 (Covington) Covington (BIS rule summary) | H200-class exports case by case, volume capped at half of US shipments. |
| US Imports from China: −19.4% YTD | Census via FRED, Jan–Jul 2026 U.S. Census Bureau | $156.4B; −34.6% vs. 2024. Exports to China flat at $65.2B. |
| US–China Truce: Held to Jan 10, 2027 | Treasury Secretary, via Supply Chain Dive, Sep 24, 2026 Supply Chain Dive / FreightWaves | Not yet documented by the White House. Under the 2025 deal China paused rare-earth export controls; trade runs on executive orders and mineral licenses. |
| China New-Home Prices: Down in 65 of 70 Cities | NBS, Aug 2026, y/y China National Bureau of Statistics | Shanghai is one of five risers. The planned housing bet, still unwinding. |
| China Real-Estate Investment: −19.9% | NBS, Jan–Aug 2026 China National Bureau of Statistics | New floor space sold −12.1%. |
| China Subsidizes Mortgage Interest | AP, Sep 29, 2026 AP via BNN Bloomberg | PSL rate cut to 1.5% plus homebuyer subsidies. Demand-side money into a supply problem. |
| China Youth Unemployment: 18.9% | NBS via Trading Economics, Aug 2026 NBS via Trading Economics | A one-year high, with 12.7M graduates. Urban rate 5.3%. |
| China CPI: +0.8% y/y | NBS, Aug 2026 China National Bureau of Statistics | Near-deflation in China vs. 3.0% core inflation here. Opposite problems. |
| Billionaires: Depends Who Counts | Hurun Global Rich List 2026; Forbes, Mar 2026 Hurun Report; Forbes; World Bank | Hurun: US 1,000, China 1,110. Forbes: US 989, China 539. Per person the US leads by 3.7 to 7.6 times; China has 4.1x the people. |
| Figure | Source & date | Why it matters |
|---|---|---|
| Corporate Profits: 13.1% of GDP, a Record | BEA via FRED, Q2 2026 BEA | $4.28T after tax, annualized, +26.6% y/y. Highest share since 1947. |
| Labor Share: Lowest on Record | BLS via FRED, Q2 2026 BLS | Nonfarm business labor share index 93.4, −3.4% y/y; lowest since 1947. |
| Pay's Share of National Income: 50.9% | BEA via FRED, 2025 BEA | Lowest since 1935. Wages and salaries alone 42.0%, tying the 2013 record low. |
| Top 0.1% Share of Wealth: 15.0%, a Record | Federal Reserve DFA, Q2 2026 Federal Reserve (DFA via FRED) | 8.6% in 1989; +0.9 points in a year. |
| Top 1% Share of Wealth: 32.5% | Federal Reserve DFA, Q2 2026 Federal Reserve (DFA via FRED) | 22.8% in 1989. The middle 40% hold 28.8%. |
| Top 10% Share of Stocks: 88.1% | Federal Reserve DFA, Q2 2026 Federal Reserve (DFA via FRED) | When the market is up, most of our tenants own almost none of it. |
| Forbes 400: $8.0T, +21% | Forbes, Sep 2026 Forbes | Up from $6.6T; a record $4.4B to make the list. 590 billionaires didn't. |
| Income Gains: 90th Percentile +1.7%, 10th Flat | Census, 2024 → 2025 U.S. Census Bureau | The K-shape in the government's own income data. |
| Figure | Source & date | Why it matters |
|---|---|---|
| NYC Rent Freeze: Live Oct 1 | CRE Daily; PoliticsNY, 2026 CRE Daily / PoliticsNY | ~1M stabilized units at 0%; seven landlords' suit heard Sep 2, ruling pending. |
| Texas Homestead Exemption: $140,000 | Texas Comptroller Texas Comptroller | School-district exemption for owner-occupied homes. Renters' buildings get none of it. |
| Apartment Insurance: $502 → $777 per Unit | NAA benchmarking, 2021–2024 National Apartment Association | +55% in three years; 2023 alone +25%. |
| Apartment Insurance Premiums: +45% | Minneapolis Fed survey, 2023 → 2024 Federal Reserve Bank of Minneapolis | Higher deductibles and exclusions shift more risk to owners. |
| Capitalism 55%; Socialism 43% | Gallup, Aug 3–24, 2026 Gallup | Share of Americans with a positive view. Socialism tops 40% for the first time since Gallup began asking in 2010; independents 37% → 45% in a year. |
| Texas Senate: Talarico 50, Paxton 44 | Marist, Sep 17–20, 2026 (n=1,139 RV, ±4.4) Marist Poll | A poll measures the race. We read the plans. |
| Michigan: 51% of Renters Cost-Burdened | Michigan Statewide Housing Needs Assessment via Newsweek MSHDA via Newsweek | ~1.1M renter households. The stakes in El-Sayed's state. |
What it means for an operator in Dallas or Charlotte.
Capitalists vs. Communists is the poster. On the field, both teams run state capitalism, one with private capital in front and one with the state in front. Both are racing to build, or sell, the machine that does a white-collar job for less.
For an operator in Dallas or Charlotte, the flag matters less than two questions. Do our tenants share in the gains or absorb the cuts? And do we compete for them, or price them off a shared sheet?
Build more. Pay more. Rescue last the ones who need it least.