Why do moths fly into flames?

Biological Insight

Let us be precise about what happened to the moth. For most of evolutionary history, the brightest thing in the night sky was the moon. Moths evolved transverse orientation - they fly at a constant angle to the light source, which, when the source is infinitely far away, produces a straight line. Efficient, elegant, and for 190 million years, utterly reliable. Then, approximately 400,000 years ago, humans discovered fire. From the moth's perspective, the moon suddenly appeared in clearings, in caves, eventually inside buildings. The moth's response was unchanged: maintain constant angle, fly straight. Except when the light source is nearby, maintaining a constant angle produces not a straight line but a spiral. The moth corkscrews inward until it hits the flame. This is not a malfunction. The moth's navigation system is working exactly as designed. The problem is that the design assumed an environment that no longer exists. Biologists call this an ecological trap: a scenario where an organism's evolved preferences lead it toward lower fitness or death because the environmental cues have decoupled from the outcomes they once predicted. Sea turtles evolved to crawl toward the brightest horizon - which, for millions of years, was the moonlit ocean. Now hatchlings crawl toward beachfront hotels and die on coastal roads. Mayflies lay eggs on asphalt because it reflects polarised light the way water does. Jewel beetles attempt to mate with beer bottles because the amber glass mimics the colour and texture of female beetles, only larger and shinier. The environment changed. The signals didn't. The instincts that ensured survival became the behaviours that ensured extinction. Three years ago, the global automotive industry flew into a flame. Mesmerised by Tesla's trillion-dollar valuation, traditional carmakers convinced themselves - and their investors - that they were one battery factory away from similar multiples. Volkswagen promised 80 percent electric European sales by 2030. Stellantis pledged 100 percent. General Motors set a 2035 deadline to abandon the internal combustion engine entirely. The signal was clear: subsidies, mandates, zero interest rates, trillion-dollar market caps. Every indicator that had historically meant 'invest here' was pointing at electric vehicles. The industry followed its evolved instincts. It spiralled inward. The flame, it turns out, was policy rather than physics. In September 2025, the US withdrew its $7,500 consumer tax credit. Emissions regulations were rolled back. Across the Atlantic, the European Commission diluted its 2035 combustion engine ban from 100 percent to 90 percent - the single largest walk-back of green policy in the bloc's history. The subsidy floor collapsed. And suddenly, vehicles that had looked like the future looked like what they actually were: products that cost more, depreciate faster, and require more planning to refuel than the cars they were meant to replace. Consider the depreciation. For most of automotive history, a vehicle was the second-largest purchase a household would ever make. Buyers accepted that it would lose value the moment it left the lot, but they relied on that depreciation being predictable and gradual. Electric vehicles broke this unwritten contract. In the UK used car market, a one-year-old Audi e-tron traded for 27 percent less than a comparable model had just twelve months earlier - while its diesel equivalent held value. Buying a new EV in 2022 turned out to be the financial equivalent of setting a pile of cash on fire to verify that it was flammable. It worked, but it was an expensive lesson. The technology moves so fast that today's cutting-edge vehicle is tomorrow's obsolete gadget. Buyers discovered that these machines age less like a vintage Porsche and more like an iPhone. Consumer Reports recently ranked Tesla as the least reliable used car brand in America. Hertz dumped 20,000 EVs from its rental fleet, explicitly citing high repair costs and lack of customer interest. When the professionals whose entire business model relies on managing fleet costs and residual values flee an asset class, it is worth paying attention. Ford announced a staggering $19.5 billion write-down as it scrapped plans for its flagship all-electric F-150 pickup. The Lightning, which CEO Jim Farley had hailed in 2021 as 'the truck of the future,' saw sales collapse 72 percent year-on-year before being consigned to the past. General Motors booked $1.6 billion to scale back its own EV production. Volkswagen is preparing to close a German plant for the first time in 88 years. These are not pivots. These are capitulations. The fundamental problem is physics. In the internal combustion era, the manufacturing math was compelling: big cars cost only marginally more to produce than small ones but could be sold for significantly higher prices. As Ford's CEO explained, 'the bigger the vehicle, the higher the margin.' Electrification inverts this logic entirely. The bigger the vehicle, the bigger the battery needed to move it. Since the battery is the single most expensive component, scaling up destroys margins rather than creating them. Americans like trucks. Trucks and SUVs now constitute 80 percent of new US vehicle sales. To achieve respectable range, an electric truck needs a battery so heavy that it erodes payload capacity and efficiency. To keep the price affordable, the manufacturer has to absorb a loss on every unit. Boston Consulting Group reported that automakers generally lose around $6,000 on every EV sold in America. For pure-play EV startups, the numbers were more terrifying still. Lucid was estimated to be losing over $300,000 per vehicle. The electric pickup, once heralded as the 'killer app' that would win over Middle America, turned out to be an engineering contradiction. The next F-150 Lightning will not be fully electric. It will carry a small combustion engine solely to recharge the battery - a tacit admission that for the heavy, aerodynamic bricks Americans insist on driving, the battery-only solution is a dead end. There is another biological phenomenon at work here: island biogeography. When a species colonises an island, it often thrives. The competition is limited. The predators are absent. The ecological niches are open. The dodo flourished on Mauritius for millions of years, evolving into a plump, flightless bird with no fear of anything - because on Mauritius, there was nothing to fear. Then humans arrived. Then rats. Then pigs. The dodo was extinct within a century. Early adopters are an island. According to Bloomberg, electric vehicles remain overwhelmingly popular among the wealthiest Americans, with interest dropping off a cliff as you move down the income bracket. This first wave of buyers treated EVs like the latest iPhone - a status symbol and a piece of cool technology. They were, critically, forgiving. If the panel gaps were uneven, if the software had problems, if the doors wouldn't quite close, they shrugged it off as the price of being on the cutting edge. These are people who view a car door that doesn't shut properly not as a manufacturing defect but as a quirky conversation starter. They treat their vehicles with the forgiving attitude a parent reserves for a toddler's terrible drawing. Eighty-four percent of them had access to home charging. Most owned a second, petrol-powered car for long trips. The industry thrived on this island. It assumed the mainland would be the same. The mainland is not an island. The mainstream buyer is not a tech enthusiast looking for a conversation starter. They are a pragmatist looking for a tool. They are cost-conscious, sceptical of 'tech for the sake of tech,' and utterly unforgiving of inconvenience. They generally own one vehicle. They park it on the street. They expect it to work seamlessly for fifteen years. When this buyer sees a $50,000 car that takes forty minutes to 'refuel' and might lose 30 percent of its range in winter, they do not see the future. They see a downgrade. Because the island was so hospitable, automakers convinced themselves they had solved the puzzle. In reality, they had just picked off the low-hanging fruit in an environment with no predators. The dodo seemed perfectly adapted - until it wasn't. Crossing from island to mainland requires a product that is cheaper and more convenient than what already exists there. Right now, the electric vehicle is neither. The mainland has competition. The mainland has standards. The mainland does not forgive panel gaps. But here is what makes this an ecological trap rather than mere miscalculation: the signals were real. The subsidies existed. The mandates were law. The zero-interest money was flowing. If you were a traditional automaker following the cues that had always indicated where to invest, you would have spiralled toward EVs too. The instinct was correct. The environment had changed. Germany withdrew its EV purchase subsidies, and sales collapsed 40 percent. The subsidies returned, and sales recovered. Italy introduced incentives offering up to €20,000 per vehicle and ran out of funds almost immediately. The demand was not organic. It was purchased. When the free money stopped, the cars stopped selling. Manufacturers were asked to invest billions in multi-decade factory projects based on demand that could evaporate overnight if a finance minister tightened a budget. China, meanwhile, controls 85 percent of global lithium-ion cell manufacturing capacity. For critical minerals like graphite and processed lithium, their dominance is nearly total. Western automakers, having followed the light, discovered they had spiralled into dependency on a foreign supply chain they could not replicate. The most valuable and complex component - the battery, accounting for 40 percent of vehicle cost - came from somewhere else. The 'German' electric car was German only in final assembly. Even the high priest of the revolution has rewritten his gospel. Tesla promised 20 million cars per year by 2030 - twice Toyota's current volume. Sales have hovered below 2 million following two consecutive years of decline. The exponential growth story is dead. Faced with a shrinking car business, the company pivoted to science fiction: humanoid robots, some of which will be shipped to Mars to work in colonies, and 'Full Self-Driving' software that is perpetually 'coming next year.' In the modern stock market, a robot in the bush is worth significantly more than two cars in the hand. A looming threat makes the economics worse still. Big Tech has entered the energy market with an appetite that makes the auto industry look small. OpenAI is planning data centres that will consume 23 gigawatts of power in the next five years - the output of 23 nuclear power stations. If electric vehicles have to compete with AI data centres for grid capacity, the era of cheap home charging may end, dismantling the last remaining economic argument for going electric. The sea turtle instinct - crawl toward the brightest horizon - worked for millions of years because the brightest horizon was always the ocean. The mayfly instinct - lay eggs on polarised surfaces - worked because polarised surfaces were always water. The automotive industry instinct - follow the money - worked because the money was usually connected to genuine demand. The environment changed. The signals didn't. And billions of dollars spiralled into a flame that turned out to be policy masquerading as physics. The moth, of course, cannot update its navigation. Its genome encodes assumptions about the world that were true for 190 million years and will be wrong for the foreseeable future. It will continue flying into flames until either the flames go away or the species evolves. Organisations are supposed to be different. They can observe, learn, and change within a single competitive cycle. But ecological traps are seductive precisely because the signals feel right. The moth doesn't know it's spiralling. From inside the spiral, constant angle feels like straight-line navigation. The destination looks exactly like the moon. There is a final lesson here, and it concerns the organisms that created this trap. Governments are ecosystem engineers - organisms that modify the physical environment, creating conditions that other species must navigate. Beavers build dams that change water flow for everything downstream. Earthworms aerate soil, creating conditions that determine which plants can thrive. Coral builds reefs that define the habitat for thousands of species. Governments construct regulatory environments, subsidy structures, and mandate timelines that shape how companies and industries evolve. Policy is not the organism. Policy is what the organism builds. The 2035 ban on combustion engines was not a law of physics but an environmental modification - a dam built by government ecosystem engineers to redirect the flow of capital and innovation. Like all such modifications, it was subject to erosion. When reality pushed back, the structure adapted. The 100 percent target became 90 percent. The hard ban softened into a system of offsets and exemptions. The dam developed leaks. This is how ecosystem engineering works in nature too. A beaver dam doesn't last forever. It requires maintenance. When the beaver dies or moves on, the dam degrades, the pond drains, and the meadow that formed behind it becomes forest again. Companies that built their strategies around the permanent presence of the dam - the subsidy floor, the mandate timeline, the regulatory certainty - discovered they had adapted to an environment that was itself temporary. The mistake was not in governments attempting to engineer the ecosystem. Ecosystem engineering is what governments do. The mistake was in other organisms treating a beaver dam as if it were a mountain - a permanent feature rather than a maintained structure. When you build your burrow in the pond behind the dam, you are betting that the beaver will keep maintaining it. When the beaver changes priorities, you drown. This has implications far beyond automotive manufacturing. Countries, cities, and regulatory bodies are all ecosystem engineers operating at different scales, each constructing environments that shape how organisms beneath them can compete. China's dominance in battery manufacturing is itself ecosystem engineering - decades of policy constructing conditions where domestic battery production could thrive. The EU's emissions regime is ecosystem engineering. So is America's tariff wall. Each creates an environment, and each environment selects for different adaptations. As policymakers approach artificial intelligence regulation, the EV experience offers a cautionary template. Top-down mandates that assume a predictable technological trajectory may create the same trap: massive capital flows toward compliance with targets that will be rewritten when they collide with reality. The alternative is ecosystem engineering that acknowledges its own impermanence - that sets direction without assuming omniscience, that creates feedback loops between outcomes and structures, that treats the dam as a hypothesis to be tested rather than a permanent feature of the landscape. The question is not whether your industry has ecological traps. Every environment that changes faster than instincts can update will generate them. The question is whether you can distinguish the moon from the flame before you've completed the spiral. The answer, for several trillion dollars of automotive investment, appears to be: no. The question for AI regulation is whether we will learn from the moths - or simply build a bigger flame.