Airy

The EU just banned Chinese inverters from all EU-funded energy projects. Brussels put China in the same risk category as Russia, North Korea, and Iran. Chinese brands Huawei and Sungrow are the world's top two solar inverter manufacturers. China controls 80% of global supply. An inverter converts solar and wind power into grid-ready electricity. Every renewable installation needs one. And the EU just decided that the dominant supplier is a security risk. The stated concern: cybersecurity. Inverters are networked devices connected to the grid. They can be remotely monitored and updated. Brussels argues that a hostile state could theoretically use inverter access to disrupt European electricity infrastructure. The practical concern: dependency. If 80% of your inverters come from one country, that country has leverage over your energy transition. The same lesson Europe learned with Russian gas is now being applied to Chinese clean energy hardware. Financial institutions have until Mid May to report ongoing projects using Chinese inverters connected to the EU grid. Brussels will then decide case-by-case by November whether to grant exemptions for projects too far along to switch suppliers. The tension in this decision is obvious. Europe needs to deploy renewables as fast as possible. The Iran war, the trade balance chart I covered last week (energy imports wiping out manufacturing surpluses), the climate targets. Speed matters. And Chinese inverters are cheaper, available, and proven at scale. Banning them slows deployment. European alternatives (SMA, Fronius) exist but can't fill 80% of the market overnight. Prices go up. Timelines extend. Projects get delayed. But the dependency argument is legitimate. Europe learned what single-source energy dependency means when Russia invaded Ukraine. Applying that lesson to the hardware layer of the energy transition is strategically consistent, even if it's economically painful. This is the choice every country faces with Chinese clean energy supply chains. Speed and cost versus strategic autonomy. Cheap panels and inverters that deploy fast versus domestic supply chains that deploy slow but keep control at home. The EU is choosing autonomy over speed. The question is whether they can build domestic inverter manufacturing capacity fast enough that the choice doesn't cripple the transition it's meant to protect. Ping me if you are working in European inverters.

A trailer full of batteries parked at a highway service area. 8 EVs charging simultaneously. 120 kW per port. $0.24 per kWh. Deployed during holiday traffic peaks when fixed stations are overwhelmed. It's a power bank for an entire highway rest stop. And it solves the exact problem that EV skeptics love pointing to: "what happens when everyone drives to grandma's house on the same weekend?" This happens. You truck in extra charging capacity the same way you'd truck in extra fuel for a gas station expecting holiday traffic. The batteries charge overnight on cheap grid electricity, get towed to the service area in the morning, and serve cars all day. The elegance is in what it doesn't require. No grid upgrade at the service area. No construction. No permitting. No new electrical connection. You drive it in, plug cars in, drive it away when the holiday's over. Porsche already rents out charging trucks for events across Europe. The economics work differently at different scales and price points. The China-specific context matters: only 12% of vehicles on Chinese roads are new energy vehicles and highway charging infrastructure is already experiencing congestion during holidays. As that percentage climbs toward 20%, 30%, 50%, the demand for flexible charging solutions scales with it. Fixed infrastructure handles the base load. Mobile infrastructure handles the peaks. Same principle that governs electricity grids: you don't build for the worst hour of the worst day. You build for average demand and deploy flexible resources for spikes.

Building a brand new wind farm with battery storage in Germany costs $91/MWh. Just paying for the fuel and maintenance of an existing gas plant costs over $100/MWh. New clean energy built from scratch is cheaper than keeping an already-paid-for fossil plant running. The capital cost of the fossil plant is already sunk. You've already built it. And it's still more expensive to operate than building its replacement from zero. That's the definition of structurally bankrupt. These IEA charts show LCOE and firm LCOE (including storage costs) for solar and wind sites around the world in 2025 and projected for 2030. Solar highlights: Hebei Province, China: $54/MWh firm LCOE today, dropping to $37 by 2030 Central Oman: $69 today, $47 by 2030 Bahia State, Brazil: $65 today, $44 by 2030 Rajasthan, India: $79 today, $54 by 2030 Nevada, US: $113 today, $77 by 2030 Wind highlights: Inner Mongolia, China: $59 today, $49 by 2030 Schleswig-Holstein, Germany: $91 today, $73 by 2030 Western Australia: $94 today, $75 by 2030 Rio Grande do Norte, Brazil: $88 today, $73 by 2030 The "firming premium" (the orange segment, the cost of adding storage to make renewables dispatchable) is the part that used to kill the economics. Storage costs have fallen 93% since 2010. That orange bar keeps shrinking. By 2030 it's a fraction of the total in most locations. China at $54/MWh firm solar and $59/MWh firm wind is already at price points that make new fossil generation absurd. By 2030, Hebei Province solar hits $37/MWh including storage. There is no fossil fuel on earth that competes with that. Every year these numbers drop, another category of fossil plant becomes uneconomic to operate. First it was coal. Then inefficient gas. Now even efficient combined-cycle gas plants in high-price markets are losing to new-build renewables plus storage. The energy transition used to be a policy choice. Now it's a spreadsheet result. And spreadsheets don't care about politics.

"Claude can rip a PR/FAQ in 30 seconds. What you don't get is the friction." AI makes production cheap. The hard part was never writing the document. It's the thinking the document forces you to do. The uncomfortable moment where you realize your customer isn't who you assumed, or your pricing won't work, or your moat is thinner than you thought. We built this program at Climate Drift because too many climate founders burn millions learning things they could have learned on paper. Free info session linked below.

Vienna heats 60% of its homes through district heating. One of the heat sources: the city's iconic waffle-cookie factory. The Manner factory. If you've ever been to Vienna, you've seen the pink packaging. They've been making hazelnut wafer cookies since 1890. The factory runs ovens at high temperatures all day. That heat used to be wasted. Now it feeds into the district heating network that warms hundreds of thousands of apartments. District heating works by circulating hot water through insulated underground pipes to buildings across the city. The heat can come from anywhere: waste incineration, industrial processes, geothermal, data centers, or a cookie factory. You just need a source of heat and a pipe to move it. Vienna has been building this system for decades. 60% of homes connected. The network is the asset. Once the pipes are in the ground, you can plug in any heat source. A factory closes? Connect a data center instead. Summers too warm? Add a cooling function. The infrastructure is flexible because it moves thermal energy, and thermal energy is everywhere. The contrast with how most cities handle heating is stark. Individual gas boilers in every apartment. Each one burning fuel independently. No waste heat recovery. No shared infrastructure. The building next to a factory that produces enormous amounts of excess heat still burns its own gas to stay warm. The boring infrastructure wins again. No breakthrough technology. No AI. No blockchain. Insulated pipes in the ground connecting heat sources to heat sinks. Planned decades ago. Still working. Still expanding. And somewhere in that network, the heat that bakes your Manner wafers also warms your apartment. Vienna at its finest, oida.

A tractor drives across a field and shoots weeds with lasers. No chemicals. No herbicides. No soil disruption. The AI identifies the weed, the laser kills it, the crop stays untouched. Carbon Robotics built this. Nvidia is promoting it. The system uses computer vision to distinguish crops from weeds in real time at field speed, then fires precision lasers to eliminate the weeds without touching anything else. No glyphosate. No Roundup. No chemical residue in the soil or the food. Herbicides are one of the largest input costs in commercial farming. They're also one of the most controversial. Glyphosate alone is the subject of thousands of lawsuits. Herbicide-resistant superweeds are spreading across US farmland, requiring stronger chemicals in a losing arms race. Organic farming avoids chemicals but relies on manual weeding, which is expensive and labor-intensive. Laser weeding bypasses the entire debate. No chemicals means no resistance, no residue, no litigation, no regulatory risk. The laser doesn't create resistant weeds the way herbicides do. And it works at the speed of a tractor pass. Let's talk access: Farm automation is powerful, but access decides the impact. If only rich farms can afford it, the productivity gap gets bigger and small farmers are left carrying the same old cost pressure. This is the access question for every piece of agricultural automation. FANUC tomato-picking arms in Dutch greenhouses. Autonomous tractors in Chinese fields. Laser weeders in California. Each one works. Each one is also expensive. The benefits accrue to farms large enough to afford the capital investment. The leasing and service models will matter as much as the technology itself. If Carbon Robotics sells laser weeding as a service per acre instead of a $500,000 machine, the technology reaches the farms that need it most.

Nvidia is putting data centers on the walls of houses. 16 Blackwell GPUs, 4 AMD EPYC CPUs, and 3TB of RAM mounted outside your home like an HVAC unit. Span, a California startup that started with smart electrical panels in 2018, is now building mini data center nodes called XFRA units. They mount on the side of residential homes and small businesses. They tap unused electrical capacity on local grids. AI cloud providers connect to the network remotely and run inference workloads. PulteGroup, one of the largest homebuilders in the US, is already installing them in newly built communities. The math Span is pitching: 8,000 of these units can be deployed 6x faster and at one-fifth the cost of building a comparable 100 MW centralized data center. Homeowners get a flat fee for power and Wi-Fi, plus compensation based on how much compute and energy the network uses. This is the data center version of distributed solar. Instead of building one massive facility that needs its own power plant, gas turbines, and a seven-year grid interconnection queue, you spread the compute across thousands of homes that are already connected to the grid. The parallels to energy are striking. Centralized power plants had the same problems data centers have now: hard to site, expensive to build, community opposition, grid bottleneck. Distributed solar solved many of those by putting generation on rooftops. Distributed compute could do the same by putting processing on house walls. Homeowners become infrastructure. Your house generates solar during the day, runs AI inference with the excess capacity, stores energy in a battery at night, and earns revenue from all three. Whether this actually works at scale depends on questions that haven't been answered yet. Latency for distributed inference. Heat management on a residential wall. Noise. Maintenance. What happens when the homeowner's electrical demand spikes and the compute needs power at the same time. But the concept is worth watching closely. The data center industry's biggest constraint is power and permitting. Thousands of homes with spare electrical capacity and existing grid connections bypass both constraints simultaneously.