Open Questions

Does AI augmentation or displacement determine whether the Forward Labor Problem intensifies or dissolves over the next decade?

If AI augments, the Forward Labor Problem intensifies — each worker must be more skilled, training pipelines must produce higher-order skills faster. If AI displaces at the aggregate level, the constraint may dissolve for existing skill categories while concentrating in a narrow AI-oversight workforce. The two regimes produce structurally different forward labor markets.

New Research:AI's net effect on labor demand — empirical evidence on augmentation vs. displacement from early adoption data, historical SUPERSEDES chain analysis across prior technology waves, and the structural indicators that would distinguish the two regimes before the full employment effect is observable.

→ Economics & Labor › Labor Markets › AI & Workforce

Can the German dual apprenticeship model be transplanted to the US, or are the institutional preconditions — sectoral bargaining, compressed wage structures, long employer-worker tenure — structurally unavailable in the US context?

The German model works because sectoral wage compression mitigates the hold-up problem. US labor markets have compressed wages in two sectors where apprenticeship works (unionized trades) and high wage variance everywhere else. The question is whether US policy can create the coordination structure the German system requires, without collective bargaining.

Is the nursing shortage a credentialing governance failure, a training capacity failure, or a retention failure — and does the answer change the intervention point?

The nursing shortage has persisted for 40+ years despite multiple interventions, suggesting the interventions are addressing the wrong failure mode. Three competing diagnoses: credentialing governance failure (scope-of-practice restrictions), training capacity failure (clinical placement constraints), and retention failure (post-pandemic burnout and exit rates).

Does geographic immobility compound skills scarcity in high-demand markets in ways that wage signals alone cannot resolve?

The standard labor market response to a regional shortage is wage premium sufficient to attract mobile workers. This mechanism fails when the cost of relocation exceeds the wage premium available: a nurse relocating from Ohio to San Francisco faces a $1,500-2,000/month housing cost differential that partially or fully offsets the wage premium available. For trades workers with established families and social networks, geographic mobility is functionally unavailable even at substantial wage premiums. This means that in high-cost-of-living markets — the same markets where AI data center buildout is most concentrated (Northern Virginia, Silicon Valley, Austin) — wage signals cannot clear the market because the relocation cost is above the equilibrium wage premium. The shortage in these markets is a housing governance failure compounding a training pipeline failure, not simply an inadequate wage signal.

Does the FERC final rule produce a uniform national co-location framework, or does state-level ZEC program jurisdiction create a patchwork where co-location is available in some states and effectively blocked in others?

FERC's authority under the Federal Power Act covers wholesale transmission and interstate commerce. State PUC authority over retail service and ZEC program terms creates independent veto points. Illinois, which has the largest ZEC program by subsidy volume and the most active consumer advocate intervention, is the highest-risk state for post-rule blocking. If federal co-location permission cannot override state ZEC conditions, hyperscalers face a state-by-state regulatory negotiation that may take longer than the queue-based interconnection process they are trying to avoid.

New Research:State-level co-location viability — ZEC program terms and amendment procedures, state PUC jurisdictional arguments, the preemption litigation risk in each ZEC state, and a practical state-by-state map of co-location availability.

→ Policy & Governance › Energy Regulation › State Utility Regulation

Which nuclear operators are commercially positioned to offer co-location, and at what price, scale, and contract structure?

Federal permission to co-locate does not create a market — it creates the legal space for a market. Constellation (the largest US nuclear operator), Talen, and a small number of others with plants in data center markets are the realistic near-term suppliers. The commercial question is whether their existing capacity commitments, ZEC program obligations, and physical plant configurations allow them to carve out co-location capacity at prices and terms that work for hyperscalers.

Does the regulatory pincer — co-location restricted by RM24-9's deliverability standard AND large-load grid interconnection years away — leave hyperscalers with no near-term path to firm low-carbon power at scale?

The Firm Power Gap document identified Primitive B as the missing load interconnection framework. If FERC's co-location rule imposes a restrictive deliverability standard that limits co-location availability, and the load interconnection rulemaking is still forming, hyperscalers face two closed doors simultaneously. The only remaining option is gas-backed grid power — which contradicts 24/7 CFE commitments — or waiting for the regulatory and interconnection timelines to resolve, which takes 3–5 years. This is the timing failure that makes the co-location regulatory outcome load-bearing for the near-term firm power gap.

Can small modular reactors deliver at the pace AI infrastructure requires, or is the timeline too long to close the near-term firm power gap?

SMRs are the most discussed solution to the firm power problem: factory-built, deployable at 50–300 MW scale, firm and low-carbon. The timing problem is not structural — it is calendar. No US SMR design has an active NRC design certification for a commercial power unit. The minimum regulatory path from application to construction start is 6–12 years. First realistic US commercial SMR deployments: 2030–2033 for demonstration-class units, 2033–2035 for commercial-scale designs. The AI data center power demand crisis runs 2024–2028. The windows do not overlap. The near-term firm power gap closes before any commercial SMR turns on. The question this document asked — can SMRs close the near-term gap? — resolves to no. SMRs are the structural solution for the 2030s, not the near-term bridge. NuScale's 2023 cancellation reset the US commercial timeline; TerraPower's Natrium design received a construction permit at Kemmerer, Wyoming in March 2026, confirming the technology is progressing — on the 2030s timeline this finding predicted. A separate traversal confirmed the calendar finding.

New Research:SMR deployment timeline and commercial viability — first-of-a-kind construction risk, NRC licensing timeline, cost trajectories from first-of-a-kind to nth-of-a-kind, and whether any design achieves commercial scale before 2030.

→ Energy & Climate › Nuclear › Small Modular Reactors

Does the FERC co-location regulatory decision resolve in favor of direct generator-to-hyperscaler contracts, or does it force grid-connected procurement?

The co-location question resolved against pure private-wire co-location in November 2024 when FERC rejected PJM's proposed expansion of the Talen-Amazon Susquehanna arrangement — the deliverability standard blocked behind-the-meter co-location at commercial hyperscaler scale. The market's response revealed the actual solution: Talen restructured the entire arrangement into a 17-year, $18B, 1,920 MW grid-connected nuclear PPA — four times the scale of the co-location proposal. In December 2025, FERC directed PJM to establish new co-location rules (Docket EL25-49) with three service tiers: Network Integration Transmission Service, Firm Contract Demand Service, and Non-Firm Contract Demand Service. The order opened a hybrid path but did not create an implementable framework — PJM compliance proceedings are running through April 2026. The structural outcome is clear: the build path for Primitive A runs through grid-connected nuclear PPAs at hyperscaler-negotiated terms, not private-wire co-location. The regulatory outcome changed the expected answer to this question but did not change the structural finding: the market needs a composable firm power product, and it now has a concrete commercial precedent in the Talen-Amazon structure.

New Research:FERC co-location rulemaking — the pro forma tariff modifications required, state PUC preemption questions, the retail customer protection arguments against co-location, and the regulatory outcome timeline.

→ Policy & Governance › Energy Regulation › FERC

Is natural gas the de facto bridge fuel that closes the near-term firm power gap, or does the methane accounting problem make gas-sourced firm power structurally incompatible with 24/7 CFE commitments?

The only available dispatchable low-cost firm power at the required scale in the required timeline is natural gas combined cycle and gas peakers. New gas plant construction is occurring at scale in data center markets. But hyperscaler 24/7 CFE commitments claim carbon-free power on an hourly basis — gas power does not satisfy this even with offsets. Either the 24/7 CFE commitment is structurally incompatible with near-term AI power requirements (forcing acknowledgment that gas is the interim bridge), or the accounting treatment of gas + offsets + hourly CFE matching is sufficiently flexible to accommodate gas-sourced power. Both outcomes have significant consequences for energy infrastructure investment and for corporate climate commitment credibility.

Does concentrated hyperscaler demand create enough revenue certainty to finance new nuclear builds at commercial scale?

New nuclear construction has failed in the US primarily due to construction cost overrun risk: Vogtle Units 3 and 4 came in at $35B vs $14B estimated, 7 years late. The revenue side of this constraint is now changing. Hyperscalers will sign 20-year contracts at $100/MWh for firm low-carbon power, and the IRA provides $15–27/MWh in production tax credits on top. The Talen-Amazon $18B grid-connected PPA confirms that offtake demand is sufficient. But a separate traversal (The Nuclear Finance Unlock) identified the remaining gap: offtake revenue is necessary but insufficient. Project finance requires that construction cost be bounded, and US capital markets will not lend against nuclear construction without a completion guarantee from a creditworthy constructor. Post-Vogtle, no such guarantee exists from any active constructor. The revenue problem is solved. The construction risk problem is not. Two missing instruments remain: a nuclear construction completion guarantee (the remaining unlock for project finance) and a standard nuclear PPA transferability mechanism (so offtake contracts can be assigned if a project changes ownership before or during construction).

New Research:Nuclear project finance in the AI era — whether hyperscaler offtake contracts structurally change the risk profile of new nuclear builds, the price points required for construction economics, and whether the capital markets will follow.

→ Energy & Climate › Nuclear › Project Finance

Can the UK Hinkley Point C Contract for Difference model be adapted to US new nuclear, and which government entity has the authority and balance sheet to be the counterparty?

Hinkley Point C is being built under a government Contract for Difference — a strike price guarantee backed by the UK government that replaced construction cost risk with sovereign credit risk. The US has no equivalent mechanism. The question is whether DOE's LPO, a new nuclear DFI, or a state-level public utility financing mechanism can provide an analogous instrument.

New Research:Government-backed nuclear financing mechanisms — UK CfD anatomy, US DOE LPO structure and capacity, Defense Production Act nuclear applications, state-level rate-basing mechanisms, and the political economy of federal nuclear construction backstop under current Congressional composition.

→ Policy & Governance › Energy Regulation › Nuclear Finance Policy

Does the FERC co-location ruling resolve in favor of direct behind-the-meter nuclear contracts, and what is the effect on offtake bankability relative to grid-connected PPAs?

The Constellation-Microsoft Three Mile Island PPA is a grid-connected deal. The Amazon-Talen arrangement was a co-location that FERC partially blocked. Co-location and grid-connected PPAs have different bankability profiles: a direct bilateral behind-the-meter contract removes RTO counterparty complexity but adds legal uncertainty; a grid-connected PPA has established legal structure but routes revenue through the wholesale market, adding settlement risk.

Do construction automation and advanced manufacturing reduce the FOAK cost premium for SMRs, or do the same workforce and supply chain constraints that drove Vogtle overruns apply to first-generation SMR construction?

The SMR project finance thesis depends on FOAK costs being lower than Vogtle's $/MW experience — either because factory manufacturing controls quality and schedule, or because smaller module sizes limit per-unit exposure. SMR factories do not yet exist at scale. The first builds will use largely conventional nuclear construction methods.

Does the AI capex cycle represent structural platform investment or a cyclical build that will compress with macro headwinds?

Hyperscaler AI capex (Microsoft, Alphabet, Meta, Amazon) is running at historically elevated levels — collective guidance for 2025 exceeded $300B across the four largest hyperscalers. If AI capex is structural, it confirms and deepens platform permanence. If it is cyclical, the current repricing has not yet priced the forward earnings compression when the build rate normalizes.

New Research:AI capex ROI analysis — the structural vs cyclical determination for hyperscaler infrastructure investment, consumption-based revenue trajectory vs. announced capex commitments, and the earnings power implications for platform businesses at different ROI scenarios.

→ Finance & Capital Markets › Equities › AI Infrastructure

At what point does tariff-driven supply chain disruption convert from temporary headwind to structural constraint on platform economics?

The current thesis is that tariffs are a temporary headwind for hardware-adjacent revenue and do not impair platform earnings power. This holds unless: (a) sustained input cost inflation forces platform pricing above customer willingness to pay, eroding demand; (b) supply chain disruption delays product cycles long enough to open competitive windows; (c) retaliatory trade measures target digital services directly, closing the GATS exemption. Each condition would convert a temporary headwind into structural impairment. The thresholds are unquantified. The risk is not symmetric — impairment events are harder to recover from than temporary earnings compression.

Does AI-enabled abstraction compress platform switching costs, eroding the platform permanence premium over time?

Platform permanence rests on switching costs: customers embed workflows on a platform and find migration expensive. AI-enabled abstraction layers — model-agnostic interfaces, universal API wrappers, AI-assisted data migration — could reduce switching costs over time if they make platform migration easier. The structural test is whether AI moves lock-in from the API layer (where abstraction is feasible) to the data and model layer (where it is structurally harder to displace). Evidence from the AI infrastructure build-out points toward deepening rather than eroding: enterprise adoption of AI co-pilots, workflow automation, and model fine-tuning is creating new platform embeddedness at the data layer — proprietary training data, fine-tuned model weights, integrated retrieval systems — that is more durable than API-layer lock-in, not less. A customer who has fine-tuned a model on their proprietary data within a cloud provider's infrastructure faces switching costs that are orders of magnitude higher than the API-layer switching costs an abstraction layer could reduce.

New Research:AI and platform switching costs — whether model-agnostic interfaces and AI-assisted migration are structurally compressing enterprise software lock-in, and the valuation implications for the platform premium if switching costs compress at scale.

→ Tech & Media › AI & Software › Platform Economics

Is the current repricing structurally more similar to 2022 (rate shock, fast recovery) or 2000–02 (structural overvaluation, multi-year impairment)?

The four-prior-event baseline assumes the current large-cap platform tech cohort is profitable, cash-generative, and structurally sound — making it more like 2022 than 2000. The structural diagnostics that separate the two regimes: FCF generation under pressure, net revenue retention trajectory, and the ratio of platform revenue to concept revenue in the index.

Does the HALEU supply chain become a binding constraint on advanced SMR deployment timelines, independent of licensing and construction readiness?

TerraPower's Natrium, X-Energy's Xe-100, and several other advanced designs require high-assay low-enriched uranium (HALEU, enriched to 5–20% U-235) rather than conventional LEU. US HALEU domestic enrichment capacity is nearly nonexistent: Centrus Energy's DOE-funded demonstration facility in Piketon, Ohio was producing at approximately 900 kg/year as of 2024 against a multi-ton annual requirement for commercial deployment. Russia (TENEX/Rosatom) was the primary global HALEU supplier until sanctions following the 2022 Ukraine invasion. The HALEU fuel supply chain is a sleeper constraint that could independently set the deployment timeline regardless of NRC licensing speed or construction finance readiness.

New Research:HALEU supply chain — enrichment capacity, DOE procurement contracts, domestic enrichment buildout timeline (Centrus, Urenco), allied alternatives, and the timeline risk this introduces for the advanced SMR deployment schedule.

→ Energy & Climate › Nuclear › Fuel Supply Chain

Does the Vogtle cost overrun experience structurally preclude conventional large-scale new nuclear in the US, or do alternative delivery models — fixed-price EPC with South Korean or Japanese contractors — change the construction economics?

Vogtle finished at $35B vs $14B estimated and became the textbook case for US nuclear unbuildability. But South Korea built APR-1400 units at Barakah (UAE) on time and near budget under a Korean EPC model with different workforce and procurement practices. If Korean or Japanese EPC contractors can deliver large-scale nuclear in the US, the SMR premium (smaller, more manageable) may be weaker than assumed.

New Research:Non-Western nuclear delivery models — South Korean APR-1400 construction economics at Barakah and domestic builds, Japanese PWR construction records, NRC design certification for APR-1400 or Hitachi ABWR, and whether the US labor/regulatory overlay fundamentally changes the economics or whether the Vogtle failure was contractor-specific.

→ Energy & Climate › Nuclear › Large-Scale Nuclear

Will the first commercial SMR deployment demonstrate a positive nuclear learning rate, or will FOAK cost data confirm that nuclear construction costs are structurally non-learnable in the Western regulatory context?

The entire commercial logic of SMRs rests on the learning rate assumption: factory fabrication and serial deployment produce declining unit costs. This has never been demonstrated for nuclear in any Western regulatory context. Large nuclear learning rates have been negative since the 1970s. The first 2–3 commercial SMR deployments will produce the first real data. If the learning rate is positive and material (>10% per doubling), the long-term economics are transformative and the SMR sector achieves commercial viability. If it is flat or negative — as it has been for large-scale nuclear — the premise of the sector requires fundamental revision. This is the most commercially load-bearing empirical question in energy infrastructure over the next decade, and it cannot be answered until units are built.

Does wholesale CBDC solve cross-border finality, or does it relocate the governance problem to central bank interoperability?

mBridge requires participating central banks to jointly assert finality for each transaction. If 180 central banks must interoperate, the governance constraint has shifted from bilateral correspondent bank agreements to multilateral central bank protocols — a harder problem, not an easier one. A bilateral Fed-ECB protocol is more tractable than a universal standard, but even bilateral requires treaty-level commitment. Whether the governance problem is meaningfully smaller with wCBDC is unconfirmed.

New Research:Wholesale CBDC governance architecture — bilateral vs multilateral protocols, mBridge geopolitical constraints, and whether a Fed-ECB atomic DvP protocol is achievable within a decade.

→ Policy & Governance › Central Banking › Digital Currency

Does the T+1 same-day affirmation failure rate worsen for US funds holding non-US securities?

US equities moved to T+1 in May 2024; EU and UK securities still settle T+2. A US-domiciled fund holding European equities must fund the cash leg at T+1 (US settlement) against a securities leg that settles T+2 in Europe. The mismatch creates a systematic one-day credit exposure that didn't exist under T+2. The scale of US institutional holdings in non-US equities — trillions of dollars — makes this a material structural risk that wasn't present in the T+2 regime.

Deepen Passes:Add a pass focused on US institutional cross-border equity holdings, the T+1/T+2 mismatch exposure, and whether the fail rate data post-May 2024 confirms the structural exposure.

Is DTCC's effective monopoly on US equity clearing structurally necessary, or a governance artifact that could be unbundled?

The netting argument is real — DTCC's ~98% netting efficiency requires all equity trades to flow through one CCP. A second US CCP would fragment netting and materially increase aggregate collateral requirements across the system. But the GameStop collateral call showed that concentration also concentrates crisis exposure. Whether the netting benefit exceeds the concentration risk — and whether a competitive CCP market could emerge with acceptable netting degradation — is unresolved and structurally important for the next phase of the market.

New Research:CCP market structure — the netting efficiency argument for monopoly, the concentration risk argument against it, and whether the SEC's post-GameStop regulatory response has changed the competitive landscape.

→ Finance & Capital Markets › Securities Infrastructure › Market Structure

Does tokenized collateral on private ledgers solve Primitive B before central bank infrastructure arrives, or does it create a new class of collateral fragmentation?

Broadridge DLR, JPMorgan Onyx, and BlackRock BUIDL are deployed on different ledgers with different legal frameworks. If institutional collateral migrates to tokenized form across multiple incompatible private ledgers, the collateral mobility problem may be replaced by a collateral fragmentation problem — faster to move within a platform, slower to move between them. Whether the cross-platform interoperability layer gets built by a consortium, by DTCC, or by a new entrant determines whether Primitive B is solved or displaced.

Deepen Passes:Add a pass on tokenized collateral interoperability — whether Broadridge, JPMorgan, and BlackRock platforms will connect, and whether DTCC's digital assets infrastructure is the integrating layer or a competing platform.

Does workload attestation survive sophisticated spoofing at scale?

Hardware-anchored latency-based location verification is not easily spoofed — physics of light speed enforces geography. But interpretive ambiguity remains: a chip in Singapore could be legitimate cloud infrastructure or a Chinese shell company. Whether NVIDIA's confidential computing infrastructure can be extended to continuous workload verification at commercial scale is unconfirmed.

New Research:Workload attestation architecture — from confidential computing to continuous custody verification. Can NVIDIA NRAS extend beyond location to prove what a chip is doing?

→ Tech & Media › AI & Software › Confidential Computing

Does the Chip Security Act mandate get written to the right level of the attestation stack?

The Act's current language is vague enough that strapping an AirTag to a GPU's heat sink would satisfy the requirement. Who writes the technical specification — and whether they understand the difference between location ping and custody chain — determines whether the mandate produces Primitive A or a compliance checkbox.

New Research:Chip Security Act technical specification — who is writing the attestation standard, at what level of the stack, and whether the compliance floor enables or forecloses the deep attestation product opportunity.

→ Policy & Governance › Export Controls › AI Chip Regulation

Is the Primitive B market commercially viable before the regulatory mandate arrives?

The commodity counterfeit electronics market is large — annual financial losses exceeding $100 billion. But the buyers who benefit most (defense primes, aerospace OEMs) already have AS9100/AS6081 frameworks and may not fund a new attestation layer until the EU Digital Product Passport creates a broader compliance requirement.

New Research:Primitive B market sizing — defense and aerospace procurement cycles, EU DPP readiness, and whether a pre-mandate commercial case exists independent of the regulatory forcing function.

→ Tech & Media › Computer Hardware › Electronics › Component Authentication

Does continuous hardware attestation constitute warrantless search under the Fourth Amendment as applied to domestic operators?

Carpenter v. United States (2018) established that prolonged location tracking requires a warrant even when data passes through third parties. A mandate requiring chips to continuously report location and workload to a government-accessible registry may be defensible for export-controlled foreign transfers — the foreign intelligence exception is broad — but its application to domestic datacenters and domestic operators is constitutionally untested. The latency-based location method (speed-of-light geography, no active reporting) may be distinguishable from GPS surveillance, but any attestation architecture that requires periodic active reporting to a centralized registry starts to look like a government-mandated tracking device installed on private property without individualized suspicion. If the mandate is written broadly enough to satisfy export control goals, it likely sweeps in domestic operators. That constitutional exposure could delay implementation, invite injunctive relief, or force a bifurcated architecture — one standard for domestic, one for export — that undermines the integrating primitive.

New Research:Fourth Amendment implications of continuous hardware attestation mandates — Carpenter applicability, the foreign intelligence exception boundary, latency-based vs active-reporting architectures as constitutional distinctions, and whether a bifurcated domestic/export standard is legally required or technically feasible.

→ Policy & Governance › Export Controls › AI Chip Regulation