New Jersey’s AI Moonshot: Governor Phil Murphy on Partnerships, Progress, and Preparedness
Summary
Murphy is treating AI as a statewide economic “realm,” not a narrow technology vertical. New Jersey’s case rests on talent—Princeton, Rutgers, the world’s high concentration of scientific researchers, its research base and pharmaceutical sector—and the NJ AI Hub’s partnerships with Microsoft and CoreWeave. His broader objective is to build “the very best catcher’s mitt” for whatever capabilities arrive next.
The clearest proof point is already in government operations: 13,000 of roughly 60,000 state employees use New Jersey’s Microsoft Copilot-based AI assistant. Murphy reports a 50% increase in call-resolution rates and, in one program, a reduction in average waits from more than 40 minutes to 90 seconds. “All government is basic—one foot in front of the other, make it work.”
Murphy currently sees AI primarily as an augmentation engine, but he does not dismiss eventual headcount reductions. Employees are beginning to escape “mundane weeds” to address neglected work, while any contraction could initially occur through attrition and forgone hiring rather than layoffs. He does not expect the tradeoff “tomorrow or even 2027 or ’28,” though he calls the options for people never hired “very, very tricky.”
New Jersey’s AI opportunity is a barbell: talent-rich technology companies at one end, compute-heavy data centers at the other. Murphy concedes that data centers are not major employment engines, making their value proposition principally about infrastructure and energy; frontier companies, by contrast, produce a talent-driven “employment buzz” he hopes Princeton can attract.
Energy is central to the compute buildout—and nuclear could benefit if AI demand persists. New Jersey still targets 100% clean energy by 2035, but its offshore-wind program is paused pending common ground with the Trump administration. Murphy sees existing nuclear capacity, possible data-center siting near nuclear campuses and reuse of former industrial sites as potential solutions, while stressing that the state has not yet “tripped a wire” on dirty energy.
The values problem is unresolved at both the domestic and international levels. Murphy prefers federal AI regulation and will enforce existing anti-discrimination laws, but his position on state-specific rules is “to be determined.” On China, he sees fundamental differences over privacy, intellectual property and the individual versus the collective—and offers the sobering judgment that institutionally there may be “little constructive way forward” for private bridge-builders.
Both speakers expect today’s assumptions to age unusually fast. Labenz argues that DeepSeek was “more on trend than off trend” and highlighted how simple some advance-driving techniques can be; Murphy concludes that “the distance that we’ve traveled from ChatGPT to today will be dwarfed” by the next two years. He is not claiming he will deliver superhuman intelligence—he is assuming it might arrive and preparing the state accordingly.
Deep dive
1. New Jersey’s AI thesis begins with talent, not technology
Murphy traces the strategy to a Bay Area dinner roughly three years earlier. After AI executives complained about costs, regulation and the business environment, he asked why they remained; their answer was talent. New Jersey’s pitch followed: its No. 1 public-education system, Princeton, Rutgers, the world’s highest concentration of scientists and research professionals per square mile, and an innovation legacy spanning Einstein, Edison and Bell Labs.
His central reframing is that AI is “a realm, not a technology.” In a pharmaceutical and life-sciences state, drug discovery alone could be transformed, but the larger error is assuming AI belongs in “a narrow technology lane.” Even without an industry-development strategy, he says government would still need to confront service delivery, ethics, equity and access.
Murphy contrasts generative AI with his skepticism about crypto, whose “so-called use case does not jump out” at him. He suggests that crypto and other past technology pitches helped poison the well; generative AI nevertheless looks different because “the use cases are overwhelming” and its impact is increasingly visible through direct experience.
His adoption model is a lead-lag story: a snapshot makes public discourse look badly behind the technology, while a video would show the curves beginning to converge. Labenz’s pushback—worth keeping—is that convergence may still be far too slow if developers’ forecasts of superhuman systems around 2027 prove directionally right.
2. The state is building a catcher’s mitt for faster capabilities
Murphy’s answer to forecasts from Sam Altman and Anthropic’s Dario Amodei is deliberately pragmatic: “I’m not going to be the guy that’s going to take us to superhuman intelligence, but I’m assuming it’s going to happen.” His job, as he defines it, is to give New Jersey “the very best catcher’s mitt” for what arrives.
The institutional centerpiece is the NJ AI Hub with Princeton University, announced just over a year earlier. Microsoft and New Jersey-born CoreWeave subsequently joined as two founding corporate pillars. Murphy wants the broader New Jersey ecosystem to serve startups, residents seeking better services and thought leaders working through ethical or regulatory questions.
The economic premise is broad optionality rather than one predetermined application. New Jersey should become a place where a founder, technical expert or policy researcher concludes, “I can find a lot of that inside of the four walls of New Jersey.”
Murphy’s humility is explicit: he does not claim deep technical expertise, and demonstrations since that dinner have repeatedly blown him away as “staggering” and “crazy stuff.” The policy response is therefore preparedness, partnership and capacity-building rather than confidence in a fixed forecast.
3. Government AI has moved from pilots to measurable service gains
About 13,000 of New Jersey’s roughly 60,000 state employees regularly use the state’s version of Microsoft Copilot, with training provided free. Murphy says New Jersey also offers its AI-assistant package to other states and cities, framing public-sector learning as “we’re all in this together.”
Murphy’s reported operational numbers are the episode’s strongest evidence: a 50% increase in call-resolution rates through tools including self-service, and one program’s phone wait falling from more than 40 minutes to an average of 90 seconds. Residents calling Murphy’s monthly Ask the Governor programs mostly ask about unresolved unemployment benefits, property taxes and other basic services.
Murphy’s workforce case is that employees spend too much time in “mundane weeds,” reaching important work only 20% of the time when it should occupy 80%. He cautiously calls the shift early-stage, but says “upskilling” is becoming concrete: residents receive faster service while workers regain time for neglected, higher-value tasks.
Applications are widening beyond call centers. The Office of the Public Defender is creating an AI-powered brief bank to help defenders draft legal briefs more efficiently, while business certifications, permitting and notary-public processes are additional targets. Asked whether defendants should be entitled to AI assistance alongside counsel, Murphy calls it a fair question he had not considered.
4. Job displacement is deferred, not dismissed
Labenz presses the core public-sector conflict: if AI can eventually answer every call better with far fewer people, should government prioritize service efficiency or preserve jobs? Murphy says New Jersey is “nowhere close” to that choice because unmet government needs still provide substantial headroom for redirected effort.
That judgment is capability-dependent, not categorical. Murphy currently sees work that remains beyond generative AI, but concedes that the boundary “may be sooner than we all think.” If government ultimately needs fewer employees, he expects the first adjustment might be gradual attrition and forgone hiring rather than a single-day layoff decision.
His stated horizon is “not tomorrow or even 2027 or ’28,” yet he identifies the harder second-order problem: what opportunities will exist for people government no longer hires? New Jersey currently has more employees and operational businesses than ever before, he says, but that future inflection point will be “very, very tricky.”
Autonomous driving makes the distributional issue tangible. Murphy considers it here to stay, but the New York–New Jersey port, one of the world’s largest, supports major trucking and related livelihoods; roughly 90% or 91% of its traffic is on the Jersey side. He argues that Waymo systems seen throughout Arizona should also face Jersey City or Hoboken: “If you really want to make sure this is right, you got to test this” amid dense, real traffic.
5. Compute growth turns the AI race into an energy problem
Murphy divides AI employment into a barbell. Hyperscalers and data centers require enormous compute and power but are “not a big employment spinner”; technology companies offer a different, talent-led environment resembling Goldman Sachs bullpen days, with highly skilled teams working together in open offices. Princeton could become the East Coast magnet for that side.
Scale makes the private sector the likely center of impact: Murphy estimates New Jersey GDP at roughly $700 billion to $800 billion, versus a state budget in the high $50 billions. CoreWeave is the local champion building both globally and in-state, while other companies are also exploring data centers and compute power in New Jersey.
The buildout collides with a 2035 goal of 100% clean energy. New Jersey’s offshore-wind program—described by Murphy as previously the nation’s largest—is paused while the state seeks common ground with the Trump administration. Existing nuclear generation is licensed in some cases into mid-century; Murphy considers it carbon-neutral while acknowledging that parts of the environmental community disagree.
Dense-state constraints favor creative reuse. A refurbished former pharmaceutical campus in Kenilworth has reached an arrangement with an unnamed large AI player after its previous occupant relocated elsewhere in New Jersey. Murphy also floats siting compute near nuclear campuses and, assuming DeepSeek does not materially alter industry trajectory, predicts AI demand could produce “a renaissance of nuclear technology and nuclear power.”
6. American values and regulation remain the least settled layer
Drawing on his experience living and doing business in China, Murphy sees incompatible defaults: privacy “is not an issue,” the collective prevails over the individual, and intellectual-property security remains a serious concern. AI will reinforce those tendencies, while the United States cannot assume its own values and norms will prevail despite its technological leadership.
Labenz resists framing the competition as a simple winner-take-all race, warning that escalation could end badly for everyone, like layering AI onto the nuclear-arms problem. Murphy agrees it will not be “Eagles 34-nothing in the third quarter”—there may be first and second place, but probably no shutout—yet offers little optimism about institutional cooperation or private bridge-building with China.
Domestically, Murphy strongly prefers federal regulation but has become less certain about what federal action will look like. He worries Europe may have moved too far too early and notes California came close to doing something consequential before pulling back. New Jersey-specific AI rules remain “to be determined,” though existing anti-discrimination and other foundational laws will apply regardless of the technology.
The closing consensus is acceleration. Labenz says DeepSeek was notable because it highlighted how simple many techniques driving advances can be; Murphy expects the conversation to look stale within three or six months. With 11 months left as governor, he wants to remain involved because “the distance that we’ve traveled from ChatGPT to today will be dwarfed by the distance we travel from today, say, two years from now.”