The Crossover

For the first time, new solar and wind capacity exceeded new fossil fuel capacity globally. Over 500 gigawatts of *nameplate renewable capacity* were added in 2025 (IEA), predominantly solar PV in China. Science magazine named this "rapid expansion of renewable energy" its 2025 Breakthrough of the Year. The turning point, so long anticipated, arrived. Context matters. New capacity adds to rather than replaces existing fossil generation in most markets. But new-build economics have decisively flipped. Barriers are no longer technological or even primarily financial. Transmission bottlenecks and permitting now dominate deployment timelines. We can build clean energy faster than we can connect it. Generation is only half the challenge. Storage, grid integration, and industrial decarbonization remain unsolved at scale. The year saw substantive progress on all fronts. Fusion control algorithms tested on real tokamaks, perovskite solar cells that last ten times longer, direct air capture plants approaching commercial scale, and electrolyzers producing hydrogen at record efficiency. The pieces are assembling. The race now is against time. ## Fusion Control Advances The path to fusion energy requires not just achieving ignition but controlling the plasma that sustains the reaction, a 100-million-degree maelstrom confined by magnetic fields and shaped by millisecond adjustments. Several advances in 2025 addressed this challenge. Frattolillo et al. demonstrated the first experimental implementation of a *Current Limit Avoidance* (CLA) algorithm on the TCV tokamak[^1] in Switzerland. In large superconducting tokamaks like ITER, reaching coil current limits can cause loss of plasma control and major disruptions. The algorithm solves a constrained optimization problem in real-time: $$\min_u \|u - u_{\text{ref}}\|^2 \quad \text{subject to} \quad I_{\min} \leq I_{\text{coil}}(u) \leq I_{\max}$$ This finds control inputs $u$ that stay as close as possible to reference values while keeping coil currents $I_{\text{coil}}$ within safe limits. Validated experimentally, demonstrating that online optimization can execute fast enough for plasma control. TCV is a medium-sized research tokamak. Results transfer to ITER insofar as the control algorithms scale, though ITER's larger size will introduce new constraints. Mele et al. demonstrated the first *Model Predictive Controller* (MPC)[^2] for plasma shape control on a real tokamak. MPC optimizes control inputs while enforcing constraints on plasma outputs, enabling more sophisticated control strategies than traditional feedback controllers. These advances address engineering challenges that become severe as tokamaks scale toward reactor parameters. ITER aims for first plasma in the late 2020s. The control infrastructure is being built now, one algorithm at a time. ## The Renewable Surge Fusion control advances one algorithm at a time, preparing for reactors that may operate decades from now. But the energy transition cannot wait for fusion. Renewable energy, available today at scale, dominated new capacity additions in 2025. Over 500 gigawatts of new renewable capacity came online, with *solar photovoltaics* accounting for the majority.[^3] Panel costs continued to fall, driven by manufacturing scale in China. *Utility-scale solar* now costs less than $30 per megawatt-hour in favorable locations (BloombergNEF), cheaper than operating existing coal plants in many markets. The economics have flipped so completely that the question is no longer whether to build renewables, but how fast. Implications extend beyond electricity. Cheap renewable electricity makes *green hydrogen* economically viable for industrial processes that cannot be directly electrified, from steelmaking to ammonia production to long-haul shipping. Decarbonization pathways that seemed economically fanciful a decade ago now pencil out. But growth creates new challenges. Solar and wind generate electricity intermittently, requiring either storage or flexible backup to match supply with demand. The mismatch between peak solar generation (midday) and peak demand (evening) creates a duck curve that deepens each year. The sun shines brightest when we need it least. ## Perovskite Solar Cells Stabilize Solar's intermittency creates the duck curve. But another challenge is simpler: efficiency. Every additional percentage point of efficiency means less land, fewer panels, lower costs. *Perovskite solar cells* have achieved laboratory efficiencies rivaling silicon at a fraction of the manufacturing cost. But they degrade rapidly, historically lasting months rather than the decades expected of solar panels. In 2025, that began to change. LONGi announced a crystalline silicon-perovskite *tandem cell*[^4] reaching 34.85% efficiency, certified by NREL. The tandem architecture stacks a perovskite layer atop conventional silicon, capturing different portions of the solar spectrum. More importantly, researchers addressed stability. Solar cells with embedded alumina (Al₂O₃) nanoparticles[^5] maintained high performance for over 1,530 hours (roughly 64 days of accelerated aging), a tenfold improvement compared to 160 hours without the enhancement. The gap between laboratory stability tests and 25-year field operation remains substantial, though the trajectory is encouraging. The nanoparticles form a protective barrier against moisture degradation and reduce iodine leakage, a primary cause of perovskite deterioration. Oxford PV, which has deployed commercial perovskite-silicon tandems, reported that their cells lose only about 1% efficiency in the first year of operation, with designs targeting 25-year lifespans comparable to conventional panels. Laboratory stability and commercial durability are converging. ## Battery Diversification Perovskite efficiency gains address how much energy each panel can harvest. But intermittency remains. When clouds cover the sun or the wind dies, generation drops regardless of panel efficiency. Storage must bridge those gaps, and in 2025, battery technology diversified beyond lithium-ion to meet different needs at different scales. *Sodium-ion batteries*[^6] reached parity with *lithium-ion* on cost and cycle life for stationary applications, though energy density remains 30-40% lower. Sodium is abundant and geographically distributed, avoiding supply chain risks associated with lithium and cobalt. Manufacturing processes are similar enough that lithium-ion production lines can be adapted. CATL, the world's largest battery manufacturer, began commercial production of sodium-ion cells in 2024. Energy density remains lower (approximately 140-160 Wh/kg compared to 200-250 Wh/kg), but this matters less for stationary storage where weight is not a constraint. For grid-scale applications, cost per kilowatt-hour stored matters more than energy per kilogram. ## Thermal Energy Storage Electrochemical batteries store energy in chemical bonds, then release it as electricity. But not all applications need electricity. Industrial processes often require heat directly, at temperatures far higher than batteries can provide. *Thermal batteries* store energy as heat rather than electrochemically, addressing needs that electrochemistry cannot.[^7] Thermal storage is well-suited to solar energy. *Solar thermal systems* can store heat in materials like molten salt or sand, then release it to generate electricity or provide heating when the sun is not shining. A sand-based thermal battery in Finland has demonstrated multi-day storage capability at commercial scale. Thermal storage can be built from inexpensive, abundant materials. Stored heat can reach temperatures exceeding 1,000°C, suitable for industrial processes that require high-grade heat. This addresses industrial decarbonization challenges that electrification alone cannot solve. ## Electric Vehicle Grid Integration Stationary storage, whether electrochemical or thermal, requires dedicated infrastructure. But another form of storage is already parked in garages and parking lots. Electric vehicles carry batteries that sit idle most of the day. A growing fleet represents both a load on and a resource for the electricity grid. Jiang et al. developed a *vehicle-to-everything* (V2X)[^8] value stacking framework that maximizes economic benefits from EVs while maintaining grid voltage stability. The framework includes *vehicle-to-building* (V2B), *vehicle-to-grid* (V2G), and energy trading. Their transformer-based forecasting model predicts building load, photovoltaic generation, and EV arrivals to optimize charging and discharging decisions. Experiments using real data from electricity markets in Australia and the United States showed significant cost reductions. Uncertainty in EV arrival times had the largest impact on system performance, highlighting the importance of accurate mobility prediction. A parked EV fleet with 50 kWh batteries and bidirectional charging capability represents distributed storage that dwarfs dedicated grid batteries, if the coordination challenges can be solved. ## Remote Renewable Energy Hubs Vehicle-to-grid technology assumes EVs are near the demand they serve. But some of the best renewable resources are far from where people live. The Sahara has abundant sun. Patagonia has abundant wind. Iceland has abundant geothermal. *Remote Renewable Energy Hubs* would harvest energy where it is plentiful and transport it to where it is needed, using renewable electricity to synthesize *energy carriers*[^9] like hydrogen, *ammonia*, and synthetic fuels. Dachet et al. proposed a taxonomy classifying hub designs by technology choices (electrolysis vs. direct synthesis), energy carriers produced (hydrogen, ammonia, methanol), and local integration strategies. The framework provides a basis for comparing proposed projects across continents. ## Direct Air Capture Scales Up Remote energy hubs would produce clean fuels at scale. But even complete decarbonization of new energy systems cannot undo the CO₂ already in the atmosphere. *Direct air capture* (DAC)[^10] addresses this legacy, pulling carbon dioxide from ambient air and storing it underground or converting it to products. Three major DAC facilities became operational or entered construction in 2024-2025. The largest, in Texas, is designed to capture up to 1 million tonnes of CO₂ per year at full scale. Iceland hosts a smaller facility capturing 36,000 tonnes annually, storing the CO₂ as mineral in basalt rock. Current costs remain high ($600-1,000 per tonne of CO₂ captured) but are falling. Costs depend on energy price, capacity factor, and accounting boundaries. Gross capture costs differ from net atmospheric removal after accounting for energy-related emissions. Climeworks unveiled Generation 3 technology, which the company claims could follow the same cost-reduction trajectory as solar. One analysis suggests costs of $230-540 per tonne by 2050. More optimistic projections claim sub-$100 costs are achievable with sufficient scale. Economics require either high carbon prices or direct subsidies. The U.S. *45Q tax credit* provides $180 per tonne for carbon stored geologically, closing much of the gap. Microsoft has purchased more than 80% of all DAC credits to date, providing early market demand. ## Green Hydrogen Electrolyzers Direct air capture requires energy, and lots of it. So does every other piece of the decarbonization puzzle. Green hydrogen sits at the intersection: it stores renewable electricity in chemical form and enables processes that electricity alone cannot power. *Electrolysis*, the splitting of water into hydrogen and oxygen, becomes the linchpin. Electrolyzer efficiency improved markedly in 2025. Ecolectro demonstrated *anion exchange membrane* (AEM)[^11] electrolyzers reaching 74% cell efficiency at high current densities, with a path to 80%. Their technology avoids PFAS, iridium-based catalysts, and titanium components, achieving 80% cost reduction compared to conventional *proton exchange membrane* (PEM)[^12] stacks. New catalysts based on ruthenium, silicon, and tungsten (RuSiW) showed exceptional durability in corrosive acidic environments while costing less than conventional platinum or iridium catalysts. The combination of higher efficiency and lower capital cost is essential for green hydrogen to compete with hydrogen from natural gas. Global installed electrolyzer capacity reached 1.4 GW at the end of 2023, nearly doubling from the previous year. Projections call for hundreds of gigawatts by 2030 if hydrogen is to play its expected role in industrial decarbonization. ## The Pieces on the Board Electrolyzers, like solar panels before them, are following the learning curve that makes technologies cheaper as they scale. Each doubling of cumulative production reduces costs by a predictable percentage. The pattern has repeated across solar, wind, batteries, and now electrolyzers. Renewable energy transition is accelerating faster than even optimistic projections suggested five years ago. Perovskite stability improvements bring tandem solar cells closer to commercial deployment. If 35% efficient panels become manufacturable at scale, land requirements for solar energy drop proportionally. Same rooftop, more power. Storage technology is diversifying to meet different needs. Sodium batteries for cost-sensitive stationary applications, thermal storage for high-temperature industrial heat, vehicle batteries for distributed grid services. No single technology solves all problems, but collectively they address most. Direct air capture and green hydrogen address emissions sources that electrification alone cannot eliminate, from aviation to shipping to steelmaking to cement production. Both remain expensive but are on cost reduction trajectories similar to solar a decade ago. The lesson of that decade is simple. Never bet against the learning curve. ## The Missing Moves Sodium batteries may displace lithium-ion in stationary applications, but manufacturing transitions take years even when the technology is ready. Factories must be built, supply chains established, workers trained. Long-duration storage, days to weeks rather than hours, remains unsolved at scale. Current technologies bridge hours to days. The multi-week doldrums demand something else entirely. Fusion control algorithms advance on research tokamaks while ITER awaits first plasma in the late 2020s. Commercial fusion reactors remain decades away, the sun in a bottle always receding to the next horizon. Direct air capture costs must fall by an order of magnitude to matter for climate. At $500 per tonne, removing 10 gigatonnes per year would cost $5 trillion annually. At $100 per tonne, the same removal costs $1 trillion: still enormous, but potentially manageable. That difference is the difference between aspiration and plausibility. Technical components of the energy transition are falling into place. What remains is execution: building, connecting, deploying, at a pace the world has never before attempted. The pieces are on the board. The clock is running. --- **Citations**: [1] Frattolillo, D., et al. "Implementation of an ITER-relevant QP-based Current Limit Avoidance algorithm in the TCV tokamak." arXiv:2502.04338, January 2025. [2] Mele, A., et al. "First experimental demonstration of plasma shape control in a tokamak through Model Predictive Control." arXiv:2506.20096, June 2025. [3] "Science's 2025 Breakthrough of the Year: The unstoppable rise of renewable energy." Science, December 2025. [4] "LONGi Breaks World Record for Crystalline Silicon-Perovskite Tandem Solar Cell Efficiency." LONGi News, 2025. [5] "Scientists crack the code to longer-lasting perovskite solar technology." ScienceDaily, February 2025. [6] "Science's 2025 Breakthrough of the Year." Science, December 2025. [7] "Materials science breakthroughs 2025: Trends to watch." CAS Insights, 2025. [8] Jiang, C., et al. "Dynamic Rolling Horizon Optimization for Network-Constrained V2X Value Stacking of Electric Vehicles Under Uncertainties." arXiv:2502.09290, February 2025. [9] Dachet, V., et al. "Remote Renewable Energy Hubs: a Taxonomy." arXiv:2507.07659, July 2025. [10] "DOE Announces $12 Million For Direct Air Capture Technology." Department of Energy, 2024. [11] "Ecolectro, Green Hydrogen Startup, Successfully Achieves Breakthrough Production and Efficiency Milestones." Ecolectro, March 2024. [12] "Groundbreaking discovery enables cost-effective and eco-friendly green hydrogen production." ScienceDaily, January 2024. **Footnotes**: [^1]: The TCV (Tokamak à Configuration Variable) is a medium-sized tokamak at the Swiss Plasma Center. It is used to test control strategies and plasma configurations for future devices like ITER. [^2]: Model Predictive Control solves an optimization problem at each timestep to determine control inputs, taking into account predictions of future system behavior and constraints. It is widely used in chemical process control and is now being adapted for plasma physics. [^3]: 500 GW of new capacity compares to approximately 8,000 GW of total global electricity generation capacity. Renewables now account for the majority of new capacity additions worldwide. [^4]: Tandem solar cells stack two different absorber materials to capture different portions of the solar spectrum. Silicon absorbs lower-energy photons; perovskite absorbs higher-energy photons that would otherwise be wasted as heat in silicon alone. [^5]: Alumina (Al₂O₃) nanoparticles are chemically inert and thermally stable. Their incorporation creates a protective layer that prevents moisture ingress and reduces iodine migration, both key degradation mechanisms in perovskites. [^6]: Sodium-ion batteries use similar manufacturing processes to lithium-ion but with sodium compounds as electrode materials. Energy density is somewhat lower, but cost can be significantly reduced due to abundant raw materials. [^7]: Thermal batteries can use sensible heat (temperature change of a material), latent heat (phase change), or thermochemical reactions. Each has different cost and performance characteristics for different applications. [^8]: V2X (vehicle-to-everything) includes V2G (vehicle-to-grid), V2B (vehicle-to-building), and V2V (vehicle-to-vehicle) energy flows. Smart charging and discharging can provide grid services while meeting driver needs. [^9]: Energy carriers like hydrogen and ammonia can be transported by ship or pipeline. Ammonia (NH₃) is particularly attractive because it can be liquefied at modest pressure and has established global shipping infrastructure. [^10]: Direct air capture uses chemical sorbents to bind CO₂ from ambient air. The CO₂ is then released by heating or pressure reduction and can be stored geologically or used in industrial processes. [^11]: AEM (anion exchange membrane) electrolyzers use alkaline chemistry but with membrane separators, combining advantages of alkaline and PEM technologies. They avoid expensive platinum-group metal catalysts. [^12]: PEM electrolyzers typically use iridium oxide catalysts at the anode, where oxygen is produced. Iridium is one of the rarest elements on Earth, creating supply constraints for large-scale deployment.