San Francisco recorded 24% growth in venture-backed company formation since 2022, while every other major U.S. tech hub contracted. Boston fell 34%. New York dropped 32%. Los Angeles declined 48%. Austin slid 47%. Three data sets reveal how dramatically San Francisco has diverged from its competitors—and how quickly that advantage could disappear.

[Chart: Venture-Backed Company Formation 2022-Present. Bar chart comparing SF +24% vs. Boston -34%, NY -32%, LA -48%, Austin -47%. Source: a16z/SVB State of Market 1H '26]

The Bay Area now captures roughly 40% of all early-stage venture dollars, ranking first in six of seven major verticals according to a16z's latest State of Market analysis. Those verticals include artificial intelligence, fintech, enterprise software, consumer technology, healthcare technology, and climate tech. Only in biotechnology does Boston maintain a narrow lead, driven by its concentration of research universities and pharmaceutical infrastructure.

Big Tech companies guided combined 2026 capital expenditures at approximately $650 billion, with the bulk targeting AI infrastructure: data centers, GPUs, networking equipment, and power systems. Amazon allocated approximately $200 billion. Alphabet set guidance between $175 billion and $185 billion. Meta targeted $115 billion to $135 billion. Microsoft estimated $97 billion to $105 billion, materially above the prior year's $88 billion. The Financial Times independently reported Big Tech AI spending at approximately $660 billion for 2026, corroborating the aggregate figure.

[Visualization: Big Tech 2026 Capex Breakdown. Stacked bar showing Amazon $200B, Alphabet $175-185B, Meta $115-135B, Microsoft $97-105B. Total: ~$650B. Source: Company guidance reports, Financial Times analysis]

This infrastructure buildout is concentrating in San Francisco. Historical data from a16z shows that more than half of net‑new jobs created since 1940 emerged in occupations that didn't exist in 1940. The AI wave follows the same pattern. Yet two recent city proposals threaten to reverse this momentum: an 800% gross receipts tax increase and research lab permitting restrictions that would reshape cost structures and regulatory environments precisely when startups gain traction.

Gross Receipts Tax: 800% Increase Targets Revenue, Not Profit

The proposed measure increases the Administrative Office Tax by nearly 800% for companies with CEO-to-median worker pay ratios exceeding 100:1. Despite being branded as a CEO tax, the levy applies to gross receipts—not executive compensation or corporate profits. Companies at the highest proposed tier would face a rate of 1.12%, nearly nine times the current 0.13% rate, according to analysis by GrowSF.

Gross receipts taxes charge revenue before expenses, meaning low‑margin businesses absorb disproportionate impact. A startup with 250 employees and $750 million in gross receipts would pay $10.4 million annually in San Francisco. The same company would pay $4 million in Oakland, $17,000 in San Jose, and $3,600 in Sunnyvale—a 600‑fold difference between San Francisco and Sunnyvale for identical operations.

[Chart: SF Tax Burden Comparison. Heat map showing $10.4M (SF) vs. $4M (Oakland) vs. $17K (San Jose) vs. $3.6K (Sunnyvale) for identical 250‑employee, $750M revenue startup. Source: GrowSF analysis]

San Francisco tested this approach in 2018 when Proposition C created a gross receipts tax structure that counted all revenue flowing through a company. Stripe relocated its headquarters ten miles south to South San Francisco within months. Other companies dissolved their headquarters structures or shifted operations outside city limits. The new proposal replicates the same mechanism at significantly higher rates.

[Timeline: SF Gross Receipts Tax History. 2018: Prop C passes with 0.5% top rate; Q1 2019: Stripe relocates HQ; 2019‑2024: Multiple companies restructure; 2026: New proposal at 1.12% top rate. Source: City records, company announcements]

Research Lab Permitting Converts from Automatic to Discretionary

Late in 2025, Supervisor Jackie Fielder introduced zoning language that would convert research and development facilities from permitted uses to conditional uses requiring case‑by‑case approval. The measure applied broadly: artificial intelligence research, life sciences, climate technology, and cancer research labs would all require political review before opening in the Mission District.

Fielder's office characterized the proposal as regulatory oversight, but conditional use permitting creates structural uncertainty. Lab approval depends on discretionary political decisions rather than codified standards, making project timelines unpredictable and costs higher. Labs seeking to establish operations face approval processes with undefined timelines, subjective criteria, and potential for political negotiation over unrelated demands.

Supervisors modified the measure after opposition from multiple colleagues. The margin was narrow. The near passage reveals how close the city came to fundamentally changing its approach to research infrastructure at a moment when AI research investment is accelerating faster than any previous technology wave.

District 5 Vote Raises Questions About Cost Pass‑Through

Supervisor Bilal Mahmood ran for District 5 in 2024 on a platform emphasizing housing production and technology sector support. The technology community contributed significant campaign resources based on those commitments. When the gross receipts tax increase came to a vote, Mahmood supported it.

GrowSF noted particular concern about impact on District 5 residents. The tax structure increases costs for grocery stores and pharmacies operating on thin margins. Those businesses typically pass increased costs to consumers through higher prices. Residents using SNAP benefits to purchase food would see their purchasing power decline as the same benefit amount buys less. The measure taxes business operations, not personal wealth, but the economic incidence falls partly on consumers.

Venture Ecosystem Concentration Accelerates Network Effects

The 24% growth figure stands out in context: every comparable city is shrinking. The venture capital ecosystem has consolidated around San Francisco at a moment when AI infrastructure spending accelerates dramatically. The Bay Area leads in six of seven major verticals and captures four dollars of every ten invested at the early stage nationwide.

This concentration creates network effects. Talent clusters near companies. Specialized service providers locate near clients. Knowledge transfer happens through informal channels: coffee meetings, conference hallways, chance encounters at events. A machine learning engineer in San Francisco can change jobs without changing neighborhoods. A startup can hire a specialized AI ethics consultant within days. An investor can visit three portfolio companies in a single afternoon.

These dynamics don't replicate easily in distributed environments, but they're fragile. Company formation responds quickly to cost structures and regulatory friction. The 2018 Stripe headquarters relocation occurred within months of Proposition C's passage. The current tax proposal creates an even steeper cost gradient.

Historical Pattern Shows Technology Creates New Job Categories

Technology employment growth over the past 80 years followed a consistent pattern: new technologies created entirely new job categories that didn't exist before. The punch card operator. The database administrator. The mobile app developer. The cloud infrastructure engineer. Each wave generated occupations previous generations couldn't have imagined.

According to a16z's longitudinal employment analysis, occupations that didn't exist in 1940 account for more than 50% of jobs created between 1940 and 2020. (Methodology note: Analysis draws from U.S. Census Bureau occupation classifications, Bureau of Labor Statistics employment data, and economic research on technological job displacement and creation. Full methodology available in a16z's American Dynamism research series.)

AI infrastructure spending at $650 billion annually will follow the same trajectory. Someone will need to train models, audit algorithm decisions, manage AI ethics compliance, design human‑AI interaction interfaces, maintain inference infrastructure, and solve problems not yet identified. Those jobs will cluster where the technology is being built.

San Francisco is currently where those jobs are forming. The 24% company formation growth translates directly into employment: new companies hire people, grow teams, spin off adjacent businesses, and create demand for specialized services. That growth is happening against the national trend in a city that has struggled with job creation in other sectors.

Policy Choices Will Determine Whether Growth Continues

The combination of near‑record tax increases and near‑passage of research restrictions reveals how narrow San Francisco's growth margin has become. The city is winning despite its policy environment, not because of it. Every startup choosing to incorporate in Sunnyvale or San Jose represents jobs, tax revenue, and ecosystem density that won't materialize in San Francisco.

The venture data makes the stakes concrete. When Boston, New York, Los Angeles, and Austin all contract simultaneously, there's no backup hub waiting to absorb displaced growth. If San Francisco's policy mix drives companies away, they'll form elsewhere—and the ecosystem advantages the city currently enjoys will erode rapidly. Network effects work in both directions.

A startup evaluating where to incorporate faces clear financial incentives. Locating ten miles south reduces tax liability by more than 99%. The $10.4 million tax differential between San Francisco and Sunnyvale represents a stark choice: maintain operations in the only American city posting positive startup formation growth at the beginning of the AI infrastructure era, or reduce costs dramatically with a short move.

What Happens Next

Three groups face immediate decisions. Founders incorporating new companies should model tax liability across jurisdictions before filing paperwork—the $10 million annual difference compounds over a company's lifetime. Policymakers need cost‑benefit analysis that accounts for lost formation, not just revenue from existing companies; historical data from the 2018 Prop C experience provides a natural experiment. Investors managing portfolio companies should evaluate whether headquarters location creates unnecessary tax drag on portfolio returns.

The data shows San Francisco holds a unique position. No other major U.S. tech hub is growing. The AI infrastructure wave is concentrating investment in ways that create lasting advantages for cities that capture it. But advantages erode when policy creates financial incentives to leave. The question isn't whether San Francisco can maintain its lead forever. The question is whether current policy choices will end it faster than necessary.

What happens when enough companies run that calculation?