Bay Area drew 52% of U.S. venture dollars in 2025 as leading AI companies cluster here

The San Jose–San Francisco–Oakland area took 52.4% of all U.S. venture capital dollars in 2025, a record share, according to the PitchBook-NVCA Venture Monitor, a quarterly report by the research firm PitchBook and the National Venture Capital Association.  Artificial intelligence drew 65.4% of all U.S. venture dollars that year, the report found.

The pull has not faded.  From July through September 2026, Bay Area startups raised $40.7 billion, about 41% of the U.S. total and more than Los Angeles, New York and Austin combined, the latest Venture Monitor shows.  Venture capital is money that investors put into young private companies in exchange for a share of ownership.

Who is here

Most of the companies building the most capable AI systems are within an hour's drive of San Jose:

• OpenAI (ChatGPT) and Anthropic (Claude), both in San Francisco.  PitchBook's latest estimates value them at about $1.02 trillion and $1.19 trillion, the two most valuable private startups it tracks.
• Nvidia, in Santa Clara, which designs the graphics processing units, or GPUs, that most AI systems are trained on.  It reported $215.9 billion in revenue for the fiscal year that ended in January and about 42,000 employees, according to its annual SEC filing.
• Google (Mountain View), whose DeepMind lab makes the Gemini models; Meta (Menlo Park); and Apple (Cupertino).
• xAI, Elon Musk's AI company and maker of the Grok chatbot, which hires mainly in Palo Alto and was acquired by SpaceX in February.  Its Colossus supercomputer is in Memphis, Tenn.
• Scale AI, a San Francisco firm that prepares the labeled data used to train AI, in which Meta made an investment in June 2025 that valued Scale at more than $29 billion, Scale said, and Databricks, a San Francisco data-software company PitchBook values at about $198 billion.

The concentration shows up in jobs, too.  The San Francisco and San Jose metro areas held 2.7% of jobs in U.S. metro areas but 13% of job postings asking for AI skills in early 2025, the Brookings Institution, a Washington research group, found.  Brookings ranked them as the only two AI "superstar" metros among 387 it studied.

How the technology works, and why that matters for location

Today's chatbots run on large language models.  A model is trained by feeding it huge amounts of text and having it predict the next word billions of times, adjusting its internal settings after each guess.  The method most labs use, the transformer, was described in a 2017 paper by Google researchers.  It is the "T" in ChatGPT.  Unlike older software, which follows rules that programmers write, or a search engine, which returns a list of links, a language model learns patterns from examples and answers in plain sentences.  Many readers already use such tools, often for free:  Stanford's AI Index estimates Americans got $172 billion a year in value from generative AI by early 2026, up from $112 billion a year earlier.

Building and running leading models takes vast computing power, mostly rented from cloud providers.  OpenAI's and Anthropic's computing bills rose sharply from 2024 to 2025, the AI Index reported, and Google alone reported more than $150 billion in capital spending in 2025.  That is why AI rounds are so large, and why access to computing power acts as a moat, a hard-to-copy advantage, that few companies outside the giant cloud providers have, according to a J.P. Morgan analysis in the Q3 2026 Venture Monitor.

Why here

• A long head start.  Fairchild Semiconductor, founded in 1957, helped make the region the center of chipmaking.  Chips led to personal computers, then the internet, then cloud computing.  Each wave left behind money, skilled workers and investors who had seen a boom before.
• Universities.  Stanford and UC Berkeley train AI researchers and spin off companies.  Google itself started as a Stanford research project.
• Big companies that seed new ones.  The 2017 transformer paper came from Google.  Researchers trained at Google and OpenAI later founded or joined rivals; Anthropic, which raised its first funding in 2021, was started by a team whose earlier research included OpenAI's GPT-3.
• Money nearby.  Many of the largest venture firms are based here.  PitchBook credits the 2025 record to the region's established network, the return to in-person work, San Francisco's revival and nearby research institutions.
• Network effects.  Each new company makes the region more useful to the next one: more engineers to hire, more customers and more investors who understand the work.
• No noncompete agreements.  California's Business and Professions Code Section 16600 voids most contracts that bar workers from joining a competitor.  SB 699, signed in 2023 and in effect since 2024, made those contracts unenforceable even when signed in another state.  Engineers can switch jobs or start companies freely, and ideas spread with them.
• Immigrant talent.  Immigrants founded or co-founded 65% of the U.S.-based companies on Forbes' 2023 list of top AI startups, according to the National Foundation for American Policy, a research group that favors more immigration.  But Stanford's 2026 AI Index found the number of AI researchers and developers moving to the United States has fallen 89% since 2017, including an 80% drop in the last year alone.

Why other places trail

The United States as a whole dominates.  U.S. private AI investment reached $285.9 billion in 2025, more than 23 times China's $12.4 billion, according to Stanford's AI Index.  The U.S. had 1,953 newly funded AI companies; the United Kingdom, next, had 172.

• Seattle is home to Microsoft and Amazon, which run two of the world's largest clouds, but its startups raised $3.5 billion in the third quarter of 2026, PitchBook data show.  CBRE, a commercial real estate firm, found AI companies there have leased less than half as much office space as in Manhattan or Boston since 2019.
• New York ($11.4 billion in the quarter), Austin ($6.4 billion) and Boston ($5.7 billion), home to MIT and Harvard, have deep talent and money but fewer frontier labs.  CBRE counts 4.0 million square feet of AI leasing in Manhattan and 3.7 million in Boston since 2019, against 21 million in San Francisco and Silicon Valley.
• London is home to Google DeepMind, which is moving its teams into a new King's Cross building this year, Google said.  U.K. private AI investment was $5.9 billion in 2025, according to the AI Index.
• Paris has Mistral AI, which raised €3 billion in September at a post-money value of more than €21 billion, the largest round ever for a European tech company, Mistral said.  France's 2025 private AI investment was $4.36 billion.
• Toronto is where Geoffrey Hinton and fellow University of Toronto faculty laid the research foundation for the Vector Institute.  The city hosts that AI research center and Cohere, a model maker founded there in 2019.  Canada drew $4.28 billion in 2025.
• Beijing and other Chinese cities are the strongest rival.  The AI Index found the performance gap between top U.S. and Chinese models has effectively closed, and China leads in AI research papers and patents.  Private figures likely understate China's spending because of government-backed funds, the report says, but Chinese firms face U.S. licensing requirements to buy advanced Nvidia chips, Nvidia says in its annual report.

What it means for jobs and pay

Pay here is far above the national average.  In the first quarter of 2026, the average weekly wage was $4,446 in Santa Clara County, up 7.8% from a year earlier, and $4,650 in San Mateo County, compared with $1,654 nationally, the U.S. Bureau of Labor Statistics reported.  BLS wage figures include bonuses and stock options, generally counted when they are paid.  San Francisco's average fell 9.6%, the largest drop among large U.S. counties.

The boom is filling offices.  AI companies are 62% of tech tenant demand in San Francisco, according to CBRE.  The city's office vacancy rate was still 27.7% in the third quarter of 2026, but tenants moved into 1.28 million more square feet than they left.

It also brings money to Sacramento.  Income tax withheld on stock paid to tech workers is about $10 billion a year, roughly 10% of all state income tax withholding, the Legislative Analyst's Office, the Legislature's nonpartisan fiscal adviser, estimated in December 2025.  That leaves the state budget tied to tech stock prices.

Housing costs remain high.  The median single-family home in Santa Clara County sold for $1.92 million in September, more than twice the state median of $899,720, the California Association of Realtors reported.

Who is left out

The gains are not spread evenly.  Silicon Valley lost 13,100 jobs from mid-2024 to mid-2025, according to Joint Venture Silicon Valley's 2026 Index, which also counted about 410,000 local jobs that include tasks AI can do.  The group said that exposure signals potential for jobs to be augmented, restructured or, in some cases, displaced, not measured job losses.

Entry-level workers appear hit first.  Employment for workers ages 22 to 25 in jobs most exposed to AI is about 19% below where it would be had it kept pace with less-exposed jobs, Stanford's Digital Economy Lab reported in August, using payroll data through June 2026.  The researchers said the drop comes mostly from less hiring, not layoffs, and that they cannot yet prove AI caused it.

Layoffs continue alongside the hiring.  Employers filed notices covering 2,736 Santa Clara County workers from July 1 to Oct. 28, including PayPal (251), eBay (229), Netflix (227) and ServiceNow (156), state WARN records show.  WARN notices, required before large layoffs, do not list a reason.  Unemployment in the San Jose metro area was 4.2% in September, up from 3.5% in May, according to preliminary BLS figures.

What it means for San Jose

AI runs on data centers, buildings full of servers, and they need huge amounts of electricity.  Santa Clara's city utility, Silicon Valley Power, says data centers use about 55% of its power.  Its peak demand of about 740 megawatts is projected to reach about 1,300 megawatts in five to 10 years, its director told state lawmakers in January.

San José signed an agreement with PG&E in July 2025 to speed power hookups after the utility received requests for nearly 2,000 megawatts of new data-center demand in the area.  The city estimated each project could bring $3.4 million to $6.8 million a year in utility and property taxes.  PG&E estimated that each 1,000 megawatts of new data-center demand could lower all its customers' bills by 1% to 2%, because big users share the fixed cost of the grid, the announcement said.

Source:  PitchBook-NVCA Venture Monitor Q3 2026; Q4 2025; Brookings; Stanford AI Index 2026; Stanford Digital Economy Lab; BLS county wages; BLS San Jose metro; CBRE; LAO; C.A.R.; Joint Venture Silicon Valley; EDD WARN report; Nvidia 10-K; City of San José; American Public Power Association; BLS QCEW Q&A; California Legislature SB 699; Mistral; NFAP; Scale AI; xAI; Google; Computer History Museum; Anthropic; Vector Institute; arXiv.