Fusion Joins the Grid Queue. One Humanoid Per Hour.

CFS submitted a Generation Interconnection Request to PJM for its 400 MW ARC fusion plant — the first fusion company ever to enter a major grid operator's queue.

Dominion Energy advised CFS on navigating PJM's process as part of a Joint Development Agreement, and PJM will now run grid simulation models to stress-test the plant's generation systems.

CFS has already secured the world's first Conditional Use Permit for a commercial fusion plant, signed offtake agreements with Google and Eni, and raised nearly $3B since 2018.

Grid interconnection studies take 4–6 years from application to electricity generation — submitting now is a prerequisite for delivering power in the early 2030s, not a nice-to-have.

The application required CFS to document not just fusion physics but the full power delivery system — steam, turbine, and generating subsystems — proving the engineering is beyond conceptual design.

PJM received 811 applications totaling 220 GW in this cycle, its first since 2022. CFS's 400 MW sits alongside solar, wind, and gas projects — treated as a peer, not a science experiment.

CFS is privately held ($3B raised) and the grid filing de-risks the timeline for offtake partners Google and Eni — if PJM approves the interconnection, the commercial contracts become materially more bankable.

Second-order opportunities include high-temperature superconducting magnet supply chains, tritium breeding blanket technology, and the specialized construction firms needed for first-of-a-kind fusion builds.

Risk caveat: SPARC, CFS's prototype, has not yet demonstrated net fusion energy (Q>1). The grid application is running ahead of the physics proof — a bet that the science will deliver on schedule.

Energinet paused all new large-scale grid connection agreements after receiving ~60 GW in requests against 7 GW peak demand — a 9:1 mismatch.

Data centers represent 14 GW of the queue; hyperscalers (Microsoft, Google, Apple) account for 60% of Denmark's existing 398 MW installed data center capacity.

Google's global data center policy director warned that companies will immediately pivot to other markets if the moratorium lacks a clear timeline, and Digital Realty's Nordic MD confirmed that operators will simply relocate AI workloads.

Denmark is the first European country to formally pause data center grid access — setting a regulatory precedent that other Nordic and EU countries are watching closely.

The moratorium reveals a structural tension: countries built the world's cleanest grids to attract the world's most power-hungry industry, and the grid cannot serve both ambitions simultaneously without massive infrastructure upgrades.

The European Commission has warned that future data center capacity growth will be constrained primarily by grid readiness, not capital — making grid access the new bottleneck for AI infrastructure investment in Europe.

The Denmark freeze directly strengthens the investment case for advanced nuclear and fusion — the grid constraint is exactly why your audience is investing in TerraPower, CFS, and Tokamak Energy. Every GW of unmet demand is a future customer for clean baseload power.

Second-order opportunities include grid modernization companies, battery storage for power quality management(IEEE Spectrum reported a new buffer system for 1.5 GW AI data centers in Texas), and behind-the-meter generation providers.

Risk caveat: Denmark's moratorium is temporary and potentially politically motivated by government formation dynamics — it may be lifted or modified once a new administration is in place. But the structural grid constraint underlying it is permanent.

Figure scaled BotQ production from 1 robot/day to 1 robot/hour — a 24x throughput increase in under 120 days — using custom manufacturing software across 150+ networked workstations.

The factory has delivered 350+ Figure 03 units, manufactured 9,000+ actuators, and achieved an 80%+ first-pass yield at end-of-line, with the battery line hitting 99.3% yield across 500+ packs.

Each robot undergoes 80+ functional verification tests including burn-in sessions with thousands of cycles of squats, presses, and jogging to eliminate early hardware failures before deployment.

The transition from prototype to production line is where most humanoid companies stall — Figure has cleared this gap faster than any competitor, with April 2026 output alone exceeding the company's entire 2025 shipment total (~150 units).

Every deployed robot is a data-collection engine for Helix, Figure's proprietary vision-language-action model — the larger the fleet, the faster the AI improves, creating a compounding advantage that widens with every unit shipped.

Figure simultaneously unveiled perception-conditioned whole-body control — robots now use onboard stereo cameras to navigate stairs, ramps, and uneven terrain without task-specific programming, a step-change from the previous "blind" walking policy.

Figure is privately held at a $39B valuation (Series C, September 2025), backed by Nvidia, Microsoft, Intel Capital, Qualcomm, Brookfield, and Parkway Venture Capital. The production ramp is the strongest pre-IPO signal the company has issued — CEO Brett Adcock has not set a fixed IPO timeline but is "monitoring investor appetite."

Second-order opportunities include actuator manufacturers, die-cast component suppliers, custom MES software providers, and the specialized testing and quality assurance infrastructure required for high-volume humanoid production.

Risk caveat: Figure 03 is deployed primarily in internal R&D and data collection, not revenue-generating commercial deployments at scale. The BMW Spartanburg pilot (30,000+ vehicles, 90,000+ parts) validated Figure 02, but Figure 03 commercial contracts have not been disclosed — the gap between production volume and commercial revenue remains the key question.

QuEra demonstrated encoding of 580 logical qubits from 1,152 physical qubits — a ratio just above 2:1, compared to the 100:1 or 1,000:1 ratios typical of current approaches.

The simulation achieved one error per trillion steps in quantum memory, approaching the "Teraquop regime" considered necessary for practical fault-tolerant computation.

The result builds on a theoretical breakthrough by Kasai (2026) revealing quantum error-correcting codes with encoding rates above 1/2, co-designed for neutral atom architectures.

The biggest barrier to useful quantum computing is that error correction consumes most of the available qubits — if you need 1,000 physical qubits per logical qubit, you need millions of physical qubits for useful computation. A 2:1 ratio, if validated in hardware, changes the entire scaling equation.

The result is specific to neutral atom platforms (QuEra's architecture), leveraging their reconfigurable qubit arrays and atom movement capabilities — it doesn't automatically transfer to superconducting or photonic approaches.

QuEra has already shipped its second commercial quantum computer (to Japan's AIST), meaning the company has hardware on which to test these codes — the gap between simulation and experiment is shorter than for a pure theory result.

QuEra is privately held and positioned as a leading neutral atom quantum computing company — if the 2:1 ratio translates to hardware, it could materially compress the timeline to commercially useful quantum systems.

The enabling technology stack includes acousto-optic deflectors, laser cooling systems, and precision optical components — companies in this supply chain benefit regardless of which neutral atom company ultimately wins.

Risk caveat: this is a simulation on paper, not a physical demonstration. The 2:1 ratio applies to memory qubits only, not computation. Carnegie Mellon's Tayur notes the result is at the "research paper level" — production-scale systems remain years away.