What we’re reading, March 6, 2026
Semiconductor lawsuits, energy transmission costs, and AI-generated research
Happy March! Here’s what caught our attention this week:
If policymakers in the US can agree on anything, it’s that they want more semiconductor plants (“fabs”) built in the US. The bipartisan CHIPS & Science Act promised hundreds of billions of dollars in federal subsidies toward this purpose, and President Biden even signed legislation exempting the factories from federal environmental reviews. Problem solved, right? Lol, of course not. Micron, a US-based chip company, is building a fab near Syracuse, NY, costing $100 billion and employing some 9,000 workers in an area that’s lost a lot of manufacturing jobs. Naturally, for this they are getting sued by a foundation/union-funded group, Jobs to Move America, for not doing sufficient review under New York’s environmental review statute. The best summary comes from the trade pub SemiAnalysis, which notes, “The project has already taken an absurd 1200 days between announcement and groundbreaking. Competitors overseas who began at the same time now have built and working fabs.” — Dylan Matthews
Last month we posted a Kiesling/Macey paper, which argued that perverse incentives for incumbents help keep energy supply constrained. A paper published last week by Dasom Ham, Owen Kay, and Catherine Hausman dovetails nicely with that post, providing more quantitative evidence for how those incentives are created. The authors find that eliminating some of the current constraints to moving energy would have reduced electricity generation costs in the continental US by roughly $3 - 5 billion in 2023 and $6 - 7 billion in 2022, when natural gas prices spiked. Better connections between energy supply (particularly in the renewables-rich interior of the country) and demand centers (primarily on the coast) would also help equalize prices, trimming revenues for coastal incumbent generators currently benefiting from their locational “advantage.” — Willow Latham-Proenca
When novel technologies get shoehorned into ill-fitting permitting systems, the results are rarely enviable. In a recent piece, Samuel Roland tours the byzantine landscape of state-level geothermal regulation. While some states consider geothermal a “surface estate” - where the landowner holds the rights - others allocate it as a mineral right to a separate - sometimes difficult to locate - entity. Attempts to map geothermal permitting onto existing technologies like oil & gas drilling have also led to permitting processes that don’t match actual risks for each project phase, and a dizzying patchwork of agency jurisdictions. To top it off, unclear application of western water law can create enough legal risk to kill project financing. The piece proposes several practical fixes, including a voluntary model code, statutory safe harbors for non-consumptive water systems, and risk-tiered permitting. — Willow Latham-Proenca
Around 23% of the terrestrial territory of the Netherlands is a protected area; to secure a permit to build near these areas, environmental review about nitrogen impacts are particularly important. The country used to use a system called the Programma Aanpak Stikstof to determine when the nitrogen impact of new construction was compliant with regulations, but a 2019 court ruling determined this program was invalid. Thus began the Dutch Nitrogen Crisis, which has constrained building in the country since. Inferring the impact of this policy change is hard because construction in the following years was probably impacted by covid-19; but a new paper makes some headway by comparing building rates across sites nearer and farther away from the protected areas. It finds the policy change reduced construction by 2.7 permits per quarter, against a mean of 41.4 permits per quarter. This paper is interesting in other ways too - see item #5 for more. — Matt Clancy
Another interesting thing about the paper discussed in item #4, about Dutch nitrogen regulation and residential building, is that it was written completely by AI! It’s part of the Autonomous Policy Evaluation Project (Project APE), an initiative from the Social Catalyst Lab at the University of Zurich. They’ve used AI to generate 241 papers so far, each evaluating the various impacts of some policy with real code and data pulled from official sources, that you can download from github to audit. The Dutch nitrogen paper has some serious issues - many of the figures and citations have been improperly processed and don’t show up correctly in the paper. But AI continues to improve. There’s been a lively debate playing out online about how the ability of AI to churn out papers could break peer review and the journal system; but I found project APE a good reminder of the potential upside: for us, transparent quantitative analysis on every policy for which data is available. — Matt Clancy
The main thing Alex has been reading this week is the text of the Senate Banking Committee housing reform bill, for which he is deep in the details. It’s encouraging to see bipartisan movement on federal housing policy, though there’s a lot to get right as it moves forward. And Jordan is on holiday. We’re excited to hear about his beach reads next week! — Nisha Austin
Here are a few other updates from our grantees and team:
Ben Schifman discusses the NEPA litigation “doom loop” - and its potential fixes - in a recent long-form piece at IFP.
Ruxandra Teslo argues that AI alone won’t speed up clinical trials. The real bottlenecks are regulatory and operational, not technical, and solving them requires institutional reform.
Finally, Saloni wrote a new post on the history of clinical trial reforms and what we should do next. She discussed how practices like randomization, preregistration, and results reporting were once seen as radical, and why greater data transparency is the natural next step.