Leverage analysis

Grid is the binding constraint, not capex or regulation. Chips ship, datacenters get sited, but interconnection queues and generation capacity cap the actual deployable compute on a 2-5 year horizon. The 0.9 enterprise score is misleading here --- frontier-lab compute expansion doesn't need enterprise absorption to justify itself, it needs megawatts. Public backlash is a second-order risk that activates only if a visible grid failure gets pinned on a hyperscaler.

No camp in this registry actively opposes alignment funding, which is why robustness is 0.82. The real friction is talent absorption rate --- interpretability researchers are a narrow pipeline and money past a threshold buys headcount that doesn't exist. Enterprise (0.7) reflects that alignment work doesn't translate into deployable product fast enough for non-frontier firms to fund it directly; it stays concentrated in 3-4 labs. Palantir is not a contester, just indifferent.

Regulation at 0.3 is the veto layer. Permitting reform, interconnection queue reform, and new generation (especially nuclear and transmission) are bottlenecked at FERC, state PUCs, and NEPA --- none of which move on frontier-AI timelines. Capex at 0.6 is secondary; money is ready, the pipes are not. Public backlash at 0.6 shows up on transmission siting, not generation. X-risk contests because more grid capacity directly enables more capability scaling, which is the thing they want paused.

Leverage is low (0.35) because retraining programs have a thirty-year track record of poor outcomes relative to spend --- the intervention polls well and delivers little. Binding constraint is that most AI-displaced roles don't have a coherent retraining target; you can't retrain a mid-career knowledge worker into a role that AI isn't also compressing. Enterprise at 0.5 reflects that firms doing the displacing have no incentive to fund retraining at scale. No camp contests it because it's politically cheap; no camp fights for it because it doesn't move the needle.

DoD obligated AI-related contract spending rose substantially 2022-2025, driven by JWCC, Project Maven, and CDAO-managed pilots; precise totals are hampered by inconsistent AI tagging on contract line items.

No other pure-play US defense-AI software vendor has matched Palantir's contract backlog or combatant-command integration depth; cloud-provider primes (AWS, Microsoft, Google, Oracle via JWCC) supply infrastructure, not mission-software integration.

Credible 2030 forecasts for US datacenter share of electricity consumption diverge by more than 2x --- from ~4.6% (IEA/EPRI conservative) to ~9% (Goldman Sachs, EPRI high scenario) --- reflecting genuine uncertainty, not measurement error.

Frontier-lab and big-tech employees have episodically resisted DoD contracts (Google Maven 2018, Microsoft IVAS 2019, Microsoft/OpenAI IDF deployments 2024), producing temporary pauses but no sustained shift in vendor willingness.