Last week we showed you that July's housing decline wasn't national — the West fell 7.1% year over year while the Midwest posted a gain. We named the split. We didn't name what's driving it.

It's AI, and it's doing two opposite things to two cities 800 miles apart.

Per GeekWire, Seattle pending home sales fell 15.6% year over year in July — the sharpest drop of any major US metro — with brokers pointing directly at tech layoffs and job insecurity.

Per Team Price's August 21, 2026 Austin market update, the Austin-area MLS is carrying 17,482 active listings against 4,055 pending, 5.9 months of inventory, and 55.93% of active listings have already taken a price cut. Buyers there are now paying 26.4% less per square foot than at the peak.

And per HousingWire, San Francisco's active inventory is down nearly 20% year over year at a median list price around $1.2 million, with relatively few sellers cutting price.

Three tech metros. One technology. Opposite results.

For thirty years the rule was simple: a technology boom meant headcount, headcount meant households, households meant housing demand. Every tech metro rose together because they were all buying the same input — people.

This boom buys different inputs. Per HousingWire, AI investment is flowing into chips, data centers, and life sciences rather than headcount-driven residential demand. We spent the last two editions on where that money goes — utility rate bases, interconnection queues, county tax abatements. None of those line items rent an apartment.

So the boom is real and the housing demand it used to generate is not automatic anymore. That's the structural break, and it explains why "tech metro" stopped being a single category.

Per tech-sector layoff tracking, 78,557 tech workers lost jobs in Q1 2026 — about 2.6 times the Q1 2025 figure. Those cuts land in cities where tech is the employment base but not the capital destination. Seattle and Austin are both.

San Francisco is the exception because it's where the capital itself is concentrated. The equity from AI companies is being created there, and a relatively small number of people holding a large amount of newly liquid wealth buy houses in a market with 20% less inventory than a year ago. That doesn't require broad employment growth. It requires a handful of very large outcomes.

Which is why the same force produces a 15.6% pending-sales collapse in one metro and an inventory squeeze in another. One city employs the industry. The other one owns it.

Worth flagging, because it's the kind of thing that decides a listing price. Seattle's pending sales drop is well documented at 15.6%. Seattle's prices are not. One current read has active listings up 14.2% year over year with the median sale price down 1.8%; another has prices up 1.8% over the same period.

Down 1.8% and up 1.8% cannot both be right, and the difference between them is whether you cut your price this month. When two market reports disagree by that much, the answer isn't to pick the friendlier one — it's to price off the transaction data in your own submarket, because the metro-level number is currently unsettled.

The reason this matters isn't the ranking. It's that almost everyone is holding the wrong map.

A seller in Seattle is pricing against comps set before the layoff cycle, in a market where the buyer pool contracted 15.6% in a year. A buyer in Austin who "waited for the market to come to them" is now shopping in a metro down a quarter from peak, which sounds like a win until they price the insurance and the tax bill — and until they ask whether the local employment base is stabilizing or still contracting. And a national investor running one Sunbelt-or-bust thesis is averaging Austin's decline against Midwestern scarcity and getting a number that describes no market that exists.

"Tech metro" is not a diversification strategy anymore. It's three different bets that used to move together and no longer do.

If you're selling in Seattle, Austin, or a similar tech-employment metro: your comps are lagging indicators of a buyer pool that has already shrunk. Price to the pending-sales trend, not to closed sales from the spring, and take the offer that's in front of you rather than waiting for the one your neighbor got in March.

If you're buying in a corrected metro: a 26% decline from peak is only a bargain if the employment base has stopped shrinking. Before you write the offer, check the metro's Q2 and Q3 layoff filings — state WARN notice databases are public and searchable — and ask whether the biggest local employers are hiring headcount or buying compute. That's now a housing question.

If you own rentals in a tech metro: underwrite tenant credit and vacancy on the layoff data, not on the metro's five-year rent history. A market with 78,557 sector-wide cuts behind it and more coming has a different renter than the one your model was built on.

If you're an investor with a national thesis: stop treating tech metros as one exposure. Seattle, Austin, and San Francisco are now three separate positions responding to opposite ends of the same trend, and a portfolio holding all three isn't diversified — it's long and short the same variable at the same time without knowing it.

Every previous technology boom widened the base of people who could afford housing in the cities that hosted it. This one is narrowing it in most of those cities while concentrating it violently in one.

That's the part that should stay with you. AI isn't raising or lowering housing — it's sorting it, by whether your city sells labor to the industry or owns equity in it. A metro on the wrong side of that line doesn't get a slow decline. It gets the sharpest pending-sales drop in the country in a single July.

What to watch from here: whether Seattle's decline spreads to the metros in its supply chain, and whether San Francisco's inventory squeeze survives the first quarter in which AI equity stops appreciating. Both are testable within two quarters, and neither is in the national numbers yet.

Claim: By November 30, 2026, the year-over-year pending-home-sales gap between Seattle and San Francisco will remain at 10 percentage points or wider — the AI split persisting as a durable divergence between tech-employment metros and tech-capital metros rather than converging back to a single "tech market" trend.

Stated confidence: 66% · Verification date: November 30, 2026 · Status: OPEN

This newsletter is the free version of what I do: take a number everyone is repeating, find out whether it holds, and show the work.

The paid version is that same check on your numbers. For this week's question, send me your metro and I'll come back with the local layoff filings, the pending-sales trend, and the employer concentration behind them — so you know which side of the AI split your market is actually on.

You get a written answer with the sources, the arithmetic, and an explicit list of what I could not verify. Same discipline as the tracked prediction above: a claim, the evidence, and no confidence I haven't earned.

cokas.io — written scope back in 48 hours. No sales call, ever.

Every prediction in The Pattern Brief carries a verification date and is revisited in a future edition — including this one.