Megadeal subsidies show up in other firms' patents. That changes the negotiation.

A forthcoming study in the Journal of Accounting Research reports that firms which received nothing from a nine-figure subsidy package go on to patent more once such a deal lands in their county, compared with firms in neighbouring counties that never got one ^[11][17][16]. That matters because the megadeal fight has been argued almost entirely over jobs promised versus jobs delivered, and this shifts part of the ledger toward an outcome states can measure but have never been asked to price ^[10][13].

The context is scale. Micron broke ground in early 2026 at Clay, New York, on what is set to be among the largest semiconductor manufacturing sites in the United States, and expects to invest more than $250 billion there over the next decade against up to $5.5 billion in state tax credits, one of the largest economic development grants ever awarded to a single corporation ^[1][2][3]. The credit is roughly 2.2 percent of the announced capital ^[4]. Other packages read differently: Indiana provided Amazon more than $8.2 billion in tax breaks, according to Business Modern Analyst as cited by Forbes, and Georgia's data centre sales-tax exemption will cost the state $2.5 billion in 2026 alone, according to Construction Owners ^[5][6].

Legislatures are moving the other way. The Center on Budget and Policy Priorities counts more than 300 bills across 41 states introduced in 2026 alone to rein in these deals, usually on the grounds that the jobs and investment bought with public money have not necessarily materialised ^[9][10]. A megadeal, in the study's definition, is a package of $100 million or more negotiated between a government and a single company for a factory, headquarters, or data centre ^[7]; critics call the practice deal shopping and a race to the bottom ^[8].

The method is the interesting part. Yoojin Lee of California State University, Long Beach, Shaphan Ng of Singapore Management University, and Aruhn Venkat of the University of California, Riverside built a dataset from Good Jobs First plus their own searches, matched to patent filings by unrelated firms in the same county ^[11][14]. The sample runs to 183,574 cohort-firm-county-year observations tied to 115 megadeals between 1990 and 2014 ^[15]. Venkat says prior academic work concentrated on tangible outcomes such as employment, mostly at county level; this design looks at an intangible outcome at the individual firm level ^[13]. Patent counts are not the only evidence offered: citations from local firms' patents also rise after an award, which the authors read as knowledge diffusing outward from the subsidised site rather than staying inside it ^[19][18].

Three cautions belong in any board pack that cites this. The sample stops in 2014, twelve years before the current legislative wave, and covers about 4.6 deals a year, so it describes a smaller, slower era of packages than Micron's ^[20][21]. The account published so far gives the direction of the effect, not its size, so nobody can convert it into a return on $5.5 billion ^[22][3]. And spillover patenting is a benefit, not a verdict; it sits on the same page as the foregone revenue, not in place of it ^[17].

Watch whether the published version reports magnitudes, whether the 300-plus bills end up capping awards or merely adding disclosure, and whether nearby firms in Onondaga County start filing.