New York took the tech talent crown from San Francisco this week. The move is sectoral rather than geographic, and it decides who signs the next office lease.
CBRE’s Scoring Tech Talent 2026 report puts the New York metro area at 394,300 tech workers against the San Francisco Bay Area’s 375,730, the first time New York has led in the 13 years of the study. The cause sits in the industry column rather than on the map. Since 2022, the finance, insurance and real estate sector added 90,530 tech jobs across the US while the high-tech industry shed 21,262. San Francisco kept the number one scorecard rank and 30% of its office leasing anyway, because a developer hired by a bank works in space the bank already leases.
Two hiring managers post for a machine learning engineer this month. One works at an AI company in San Francisco and one works at an insurer in Midtown Manhattan. Their salary bands overlap. Their effect on the buildings around them does not. The insurer seats its new hire on a floor it signed for in 2019. The AI company needs a floor that nobody has built yet.
Both hires count the same in the headline everyone ran this week. Only one of them puts a broker in a room.
What Happened
CBRE published the 13th edition of Scoring Tech Talent on August 18, and CNBC picked it up on August 21 through Diana Olick’s Property Play newsletter. The finding that traveled: New York’s tech talent workforce reached 394,300 in 2025, ahead of the San Francisco Bay Area’s 375,730. New York added 30,640 tech jobs between 2022 and 2025. The Bay Area lost 23,900 over the same three years.
The crown is not part of CBRE’s actual index. The scorecard, which weights 13 metrics covering talent depth, concentration, wages, rents and graduate supply, still ranks the San Francisco Bay Area first, then Seattle, Toronto, New York, Austin and Washington DC. That top six has not changed since last year. Boston moved up two places to seventh. Nashville, Montreal, Jacksonville and Pittsburgh climbed the most.
The AI numbers underneath are large and moving fast. Tech workers with AI skills across the US and Canada grew 45% year over year to 751,000 as of mid-2026. The San Francisco Bay Area, New York, Seattle and Washington DC hold 37% of the US total. Toronto, Montreal and Vancouver hold 60% of Canada’s. Both New York and the Bay Area added more than 20,000 AI specialists in the past twelve months, which is the same story in both cities.
The divergence lives in the jobs that are not AI. AI roles made up 31% of available US tech talent listings in June 2026, up from 11% at the mid-2022 hiring peak. In the Bay Area the AI share hit 57%, up from 20%. Both markets posted about a third more AI jobs than they did at that 2022 peak. Non-AI tech postings fell 60% nationally and 73% in the Bay Area.
Colin Yasukochi, who runs CBRE’s Tech Insights Center, told reporters the Bay Area should remain “the central location for the AI industry and for innovation” even after losing the headcount lead.
The Backstory
Rewind to 2022 and the picture inverts. US tech talent employment grew 7.3% that year, adding 405,000 jobs. Growth fell to 1.1% in 2024 and recovered to 1.8% in 2025, worth 108,760 jobs. The 2025 gain is roughly a quarter of the 2022 gain.
Underneath that flat-looking aggregate, employers ran a swap. They cut 1.21 million jobs in 2025, up from 761,358 in 2024. Through June 2026, the tech industry accounted for 31% of the 443,604 cuts announced so far this year, against 13% of cuts for all of 2025. Cuts that employers blamed on AI directly reached 101,743 through June, or 22.9% of the total, after totaling 54,836 and 4.5% for the whole of last year. The share went up fivefold in six months.
What replaced those roles pays better and sits in different companies. Data scientist positions grew 12.4% in 2025, adding 29,000 jobs, and finance led that growth rather than tech. Computer and information systems manager roles added 24,600. Technology and engineering jobs added 15,030, and 95% of the high-tech industry’s share of those came from manufacturing rather than software.
The sector table is where the story lives. Since 2022, finance, insurance and real estate added 90,530 tech jobs. Professional services added about 66,000. Transport, warehousing and wholesale added about 66,000. The high-tech industry subtracted 21,262.
New York did not out-recruit San Francisco. Banks and insurers out-recruited software companies, and New York holds more banks and insurers than anywhere else. Finance alone added nearly three times as many US tech jobs as New York’s entire metro gain.
The Plan
CBRE is a brokerage, and Scoring Tech Talent is a lead-generation asset for two of its business lines. The chapters carry links to its Technology Practice Group, its Plan Lease and Occupy service, and its Labor Analytics consulting practice. The report exists to help an occupier pick a city and a landlord price a floor.
Read the scorecard methodology with that in mind and the weighting makes sense. CBRE weights labor costs above office rents, in its own words, “because companies allocate more capital to labor than to real estate”. Concentration metrics carry the highest weights, because clustering is what produces new leases.
The commercial payoff shows up in chapter six. Tech companies took 21% of all US and Canada office leasing activity in the first half of 2026, counting deals of 10,000 square feet or more, up from 13% in 2023. AI companies drove 30% of San Francisco leasing since 2023 and require staff on site full time, which is why fully remote postings in the Bay Area collapsed to 7% by April 2026 from 24% in mid-2022. The national figure is 18%.
New York has its own version. Commercial Observer reported in April that Anthropic was taking all 465,630 square feet of 330 Hudson Street, and in June that Airbnb had struck an $82 million deal for 281 Park Avenue South as an East Coast hub. Colliers counted 800,000 square feet of Manhattan office demand from AI businesses in the second quarter alone, more than the whole of 2025.
Those are AI-company leases in New York, not bank-hired-a-developer leases. They are also a small fraction of the 30,640 jobs that won New York the crown.
The Business Model Angle
Three things separated this year, and only one of them is about geography.
Tech talent became an input, and inputs get bought by whoever has the budget. More than 38% of tech workers in the US and Canada still work inside the tech industry, but the spread by market is wide. CBRE puts the San Francisco Bay Area above 50%; Commercial Observer, reading the same chart, reports 61% for San Francisco against 34% for New York. Software engineers concentrate harder still, at 69% inside the tech industry in the Bay Area against 48% nationally. San Francisco’s tech workforce sells to one customer industry. New York’s sells to banks, insurers, hospitals, media and retail. Any operator who has run a business with 60% of revenue in one account knows what that does to the volatility of the P&L. The Bay Area’s labor market is levered to a single sector’s capital cycle. New York’s is diversified by construction.
Headcount does not convert to square feet at a constant rate, and the two crowns have separated. A developer hired by JPMorgan occupies a desk that already exists inside a lease signed years ago. A developer hired by a Series B AI company triggers a new lease, a buildout and a broker commission. That difference explains why New York can take the headcount lead while San Francisco keeps the number one rank and 30% of its market’s leasing activity. It also explains the strangest number in the report: San Francisco office rents in the fourth quarter of 2025 sat 14% below where they were in the first quarter of 2020, in the middle of an AI leasing boom, while South Florida, Nashville, Austin and Dallas posted rent growth above 20%. The demand landing in San Francisco is real and it has not yet repriced the market it is landing in. New York’s tech talent win is the least real-estate-productive form of tech job growth in America, which is why the brokerage that published the number left it outside its own index.
The AI labor market is a hiring flow sitting on an installed base that has barely moved. In the Bay Area, AI roles are 57% of tech job postings against 26% of the actual tech workforce carrying AI skills, a ratio of better than two to one. Across the US and Canada, the 751,000 AI-skilled workers amount to under 10% of the 7.6 million tech talent total. Pricing follows the flow and then reverts toward the stock. The tech industry wage premium over the US average already fell to 15% in 2024 from 18% in 2023, and CBRE attributes the compression to non-tech employers bidding for the same people. Talent that once cost a premium because only one industry wanted it now costs less because everyone wants it, which sounds backwards until you remember that supply moved too.
The cost picture is worth carrying into any siting decision. CBRE models a 500-person tech company in 60,000 square feet at $91 million a year in the San Francisco Bay Area and $36 million in Quebec City. That works out to $182,000 per head against $72,000. The $55 million gap is mostly payroll, not rent, which is the entire reason the scorecard weights labor first.
The Risk
The strongest objection is methodological, and it lands on the AI numbers rather than the crown. CBRE builds its AI-specialty counts from LinkedIn members who self-report their occupation and their machine learning skills. BMA covered what happens to that measure when a Stanford and Revelio Labs team compared archived pre-2023 profiles with live ones and found 30% more AI references in the same work histories, some of them inserted retroactively into job titles that never carried them. If profiles carry 45% more AI skills this year, some unknown share of that is people typing a word that pays. The 751,000 figure measures claims. The job posting data, which comes from Lightcast rather than LinkedIn, measures employer behavior and deserves more weight.
The finance number has a reclassification risk of its own. Banks that moved contractors onto payroll, or that reclassified existing staff into tech occupation codes, would produce the same 90,530 without adding a single new job to the economy. CBRE’s lens is occupational rather than industrial, so a bank employee who moves from operations to a data engineering title becomes tech talent without changing seats. Nobody has published the split.
The Challenger cut figures are announcements, they are gross rather than net, and they cover every role at a tech employer instead of only tech occupations. Comparing them against CBRE’s occupational employment counts mixes two measures, and the comparison holds only as a rough scale check.
The crown itself is thin. New York leads by 18,570 workers, under 5% of San Francisco’s base, on estimates built from survey data. One revision closes it.
Then there is the counterargument that San Francisco is winning the part that matters. The Bay Area drew 80% of US AI venture funding since 2020, holds a sixth of the country’s AI-specialty talent, and posted the highest AI concentration of any market at 26%. Venture capital funds the companies that sign leases, and it clusters harder than labor does. Europe has spent a decade discovering that a talent pool without local growth capital exports its scale-ups rather than its problems.
The last risk cuts against both cities. Only nine of CBRE’s 50 markets created more tech jobs than they graduated tech students. Washington DC and the San Francisco Bay Area were the top two net tech graduate markets, meaning the number one ranked tech market in North America now runs a labor surplus. That is the same supply-side squeeze colleges are already pricing into tuition, arriving at the other end of the pipeline.
Quick Questions
Did tech workers move from San Francisco to New York? The report does not show a migration. It shows New York adding jobs while the Bay Area cut them, in different industries. A laid-off Meta engineer who takes a job at a Manhattan bank shows up in both numbers, but so does a New Yorker who never left.
Is San Francisco losing the AI race? No. AI is 57% of its job postings, the highest share of any market, and it holds a sixth of US AI-specialty talent plus 80% of AI venture funding since 2020. It is losing the non-AI base faster than anywhere else, at a 73% drop in postings against 60% nationally.
Why does San Francisco still rank first? The scorecard measures concentration, wages, education and quality rather than headcount. Tech talent makes up more than 10% of total employment there. Across the 50 markets the average is 5.5%.
Where should a company put an engineering team? Depends on what you are optimizing. CBRE’s cost model spans $36 million to $91 million a year for the same 500-person team. If you need people who have shipped frontier AI products, you pay Bay Area rates because the pool is there. If you need competent engineers for enterprise work, Toronto added 75,000 tech jobs since 2022 and Dallas added 37,230, both at a fraction of the cost per head.
What does this mean for office landlords? Tech is back to 21% of leasing, but it is concentrated in AI companies that need full-time in-person space. Markets where tech jobs grew inside banks and insurers get the wage income and the apartment demand without the office absorption. Manhattan apartment rents already average $3,653 a month, the highest in the study, at 29% of the average local tech salary.
The Business Model Analyst Take
Every wire story this week ran the same frame: New York beat San Francisco. The frame is wrong in a way that matters to anyone making a siting, hiring or leasing decision. Finance beat tech. New York happens to be where finance sits.
The lesson generalizes past cities. When a specialized input becomes a general one, the places and companies that win are the ones with the most different buyers, not the ones with the most concentrated expertise. San Francisco built the deepest tech labor pool on earth and sold it to one industry. That worked for two decades and then the industry started buying compute instead of people. New York never had the deepest pool. It had the widest customer list, and that is what showed up in the count.
For operators, two things follow. If you employ specialists, check the industry mix of the market you hire in before you check the salary data, because a single-industry labor market moves with that industry’s capital cycle and yours might not. Cheap talent in a downturn is only cheap if your own demand holds up.
If you own or lease space, stop treating job growth as a demand signal without asking who signs the paychecks. Employment counts and square footage counts diverged this year and the gap is going to widen. The buildings get built where companies expand, not where workers live, and those two maps came apart the moment tech talent stopped being a tech-industry asset.
