Connecticut’s economic condition is easy to misread. By the measures that make headlines, we look enviable: among the highest household incomes in the nation, and a state government in its strongest fiscal position in decades. The rainy-day fund is full, and the state has paid down more than $10 billion in pension liabilities.
But those headline rankings –and the tax revenues behind the state’s surpluses— are disproportionately the result of capital gains, dividends, interest, and rents flowing to a small number of wealthy households. Our high median income is real, and it is largely the legacy of high-wage jobs created decades ago.
What the rankings do not show is a healthy, growing Connecticut economy underneath. The deeper numbers tell the truer story. Since 1989, the United States has expanded its employment by nearly 50 percent. Connecticut’s payroll growth over those 37 years is below 3 percent. We needed 18 years just to regain the employment peak set in March 2008, a peak virtually every state had passed, most a decade ago. And the quality of our jobs has been eroding alongside their quantity, with high-wage technical and corporate roles steadily replaced by positions paying far less.
Recent quarters have brought genuine momentum, including stronger growth, new arrivals, a bond-rating upgrade. Most welcome, and, against a 37-year hole, a beginning rather than an arrival.
Ours is a state with a healthy balance sheet and a sick economy. The balance sheet will not stay healthy if the economy stays sick. The surpluses funding the pension paydown ride on volatile capital income generated outside the state; if our own growth does not return, the liabilities will.
This condition is structural, and it is old; it predates every current officeholder by a generation. And the essential first response has already been delivered: enforcing guardrails a bipartisan legislature built, Gov. Ned Lamont has restored the state’s fiscal foundation, faithfully, against every pressure to abandon the discipline. That foundation is the signal achievement of this era; it will pay dividends for decades. The question this essay takes up is what it now makes possible.
Why growth left
Let’s be honest about why Connecticut got left out. The modern data-driven economy rewards concentration: deep pools of talent, capital, and firms feeding on one another in a handful of giant metropolitan centers. Connecticut is a small state with no major metro, sitting between two of the most powerful metropolitan economies on earth. That is a structural fact, not a policy failure we can reverse with a marketing campaign. On conventional terms, whether hunting elephants, bidding for footloose tech firms, trying to out-Boston Boston, we cannot compete. We should stop pretending otherwise.
But economic history, including Connecticut’s own, teaches a critical lesson. Regions rarely win by imitating the leader. They win by differentiating, by driving deep into specialized capabilities the leaders do not have and cannot quickly acquire. Connecticut was never an economic power because of its size. It became the patent capital of 19th-century America through the precision manufacturing of the armory tradition. It made Hartford the insurance capital of the continent through accumulated expertise no rival could replicate. Specialized depth, not scale, is this state’s original playbook. Three words summarize three centuries of Connecticut economic success: differentiate, don’t imitate. The question before us was never how Connecticut becomes another Boston. It is how Connecticut becomes the world’s best Connecticut.
The layer Connecticut can win
Artificial intelligence hands us the chance to answer it. The AI economy has two layers. In the platform layer, the frontier models, massive compute, and elite research talent, every advantage concentrates: Connecticut will not win. New York’s $500 million Empire AI consortium and Massachusetts’ $100 million AI Hub target that platform layer. Matching them dollar-for-dollar would buy us an inferior copy of what our neighbors already have.
We do not need to build the AI brain. We need to become the best place in the world to apply it.
In the application layer, where AI is embedded in real industries, the scarce inputs flip. Value flows to proprietary data, deep domain knowledge, and physical assets. Those are precisely what Connecticut holds.
The market is already pricing this logic: when Salesforce’s latest results confounded predictions that AI models would swallow the software incumbents, analysts led by Yale’s Jeffrey Sonnenfeld concluded that value in the AI economy is migrating to those who hold the scarce data intelligence cannot function without. AI underwriting is worth little without decades of claims data; this is the actuarial talent concentrated in Hartford.
AI-augmented design and production of submarines and jet engines must happen where the factories, security clearances, and controlled processes physically sit. Groton, East Hartford, and Stratford cannot be relocated. Neither can the thousands of precision manufacturers that supply them.
AI-driven genomic medicine requires sequencing capacity, clinical depth, and translational research at scale, precisely what the consortium of Yale, UConn, and the Jackson Laboratory was built to deliver, and an asset whose potential we have barely begun to realize. And these assets sit unusually close together: world-class universities, insurers, hospitals, defense plants, and the state government itself, all within an hour of one another. What Connecticut lacks in metropolitan scale it holds in convening proximity. Applied AI is, above all, a coordination problem among institutions.
What it looks like in practice: A state-brokered data commons, letting our insurers, hospitals, and manufacturers give vetted Connecticut startups secure, controlled access to the industry data they cannot get anywhere else —without that data ever leaving its owners’ hands— turning Connecticut’s institutional memory into its recruiting pitch; a compact among our anchor employers to open real operational problems, each year, to Connecticut startups and university teams.
The scarcest assets in any startup ecosystem are not accelerators and mentors but hard problems and sophisticated first customers. Connecticut has both in abundance. For example, innovation vouchers that let a 60-person machine shop in Waterbury hire Connecticut talent to bring AI onto its shop floor.
And an AI extension service is an idea Connecticut invented. The nation’s first agricultural experiment station was established in this state in 1875, decades before the federal government followed. It was research at the Windsor station that developed shade-grown tobacco and built a world-leading Connecticut industry from public science carried onto private land. The AI extension service applies that same Connecticut invention to the factory floor and the claims office: practitioners who walk the floor and connect firms to Connecticut tools and training, not consultants who bill for slide decks.
These are not hypothetical needs. Under new federal cybersecurity certification rules now attached to the Navy’s latest submarine contracts, hundreds of small Connecticut defense suppliers must meet demanding security standards within the next year or watch decades-old work, and thousands of new hires, flow to better-prepared states. The vouchers, the extension service, and the training pipelines above have their first mission already waiting, with a hard deadline.
The workforce ambition should be equally sharp. Connecticut cannot out-produce Massachusetts or New York in generic AI engineers and should not try. The fastest-growing demand nationally is for AI fluency, for people who can work with, direct, and audit AI inside a domain they already master. That is our opening, because domain mastery is what Connecticut’s workforce already has.
The aspiration: America’s most AI-fluent insurance workforce. America’s most AI-fluent aerospace and defense manufacturing workforce. One of America’s most AI-fluent biomedical workforces. Fluency compounds on expertise: our machinists, adjusters, and lab technicians can be elevated into directing the tools entering their trades. And the graduates our universities produce every year will finally see careers worth staying for, here rather than in Boston or New York.
Be clear-eyed about the alternative
Our insurers, manufacturers, and health systems will adopt AI regardless of what Hartford does. The projections are no longer abstract: a new Tufts University index puts expected AI-driven job displacement over the next two to five years at roughly 16 percent in finance and insurance and in professional and technical services, the sectors where Connecticut’s white-collar employment is most concentrated, in a state that has already lost some 25,000 insurance jobs.
The strategic question is not adoption; it is where the value is captured. If the tools, the vendors, the startups, and the specialized workforce are built elsewhere, Connecticut absorbs the job losses, imports the productivity gains, and watches the domain knowledge that is currently our advantage migrate into other states’ firms. If they are built here, disruption converts into a new export industry. That window will not stay open long. Connecticut must act, and act now.
The quantum stakes
The stakes are no longer hypothetical. Last month Connecticut won one of just 12 NSF Regional Innovation Engine awards nationwide for QuantumCT. The first tranche is $15 million, with up to $160 million over a decade, if Connecticut delivers. That federal money must be earned, not banked: it is released in stages, only as quantum technologies demonstrably become embedded in defense, biotechnology, and financial services.
Those are exactly the markets now being rebuilt around AI. Critically, AI software will design, optimize, validate, and carry quantum’s capabilities into use. If Connecticut sits out the AI economy, it will struggle to hit the milestones its quantum award requires and will likely watch the largest federal innovation investment in its history slip away. AI and quantum are not two initiatives. They are one strategy.
The rules are written; now build
Connecticut has, to its credit, already done something its neighbors have not: written the rules. The bipartisan AI legislation signed in June gives firms here what founders elsewhere still lack: regulatory certainty and a sandbox in which to build. For entrepreneurs creating AI for regulated, knowledge-intensive industries, specifically insurance, healthcare, aerospace and defense, bioscience, that is exactly the environment that matters most. But rules are not a strategy.
None of this asks Connecticut to abandon its fiscal discipline; it asks us to use what that discipline has built. Act one is complete. Act two is what foundations are for. A one-time, tightly capped investment of $100 to $150 million from the current surplus, less than 2 percent of what the guardrails have already delivered to the pension funds, deployed only alongside at least two private matching dollars for every public one, would mobilize $300 million or more for precisely the industries where Connecticut can actually win.
These are investments in which the taxpayer shares the upside. And unlike the defensive $500 million the State has prudently set aside to buffer federal funding cuts, this is offense: an investment in growing the economy that ultimately secures the pension paydown itself, because only a broadening tax base keeps those liabilities funded through the next market correction.
Even the physical foundations favor us. Connecticut is a substantial exporter of electricity, with nearly half the power generated here serving customers beyond our borders. Where computing infrastructure is needed, Connecticut can host it selectively, on its own terms, with ratepayers protected rather than burdened.
In the coming weeks, the Connecticut Center for Economic Analysis will publish the full blueprint behind this argument, a granular policy brief with the specific programs, the financing, and the governance, along with a commitment to provide an annual public assessment of whether the initiatives are working. Connecticut has heard promises before; this strategy comes with a scorecard.
For 37 years Connecticut has lacked a credible answer to its stagnation. The AI revolution has flipped the odds: for the first time, the structure of an emerging technology aligns with the structure of our strengths. We will not get that alignment twice. The world’s best Connecticut is still ours to build.
Fred Carstensen is a Professor of Finance and Economics at the University of Connecticut and Director of the Connecticut Center for Economic Analysis.