Sophia and I first collaborated in 2023, building technology partnerships at the intersection of AI and sustainability. When she told me she was leaving BCG to join Macrocosm’s leadership team, I immediately wanted to hear more about what Macrocosm is building.
I recently sat down with Macrocosm’s founder, Professor Doyne Farmer, and Sophia to learn more about the ‘supersimulator’ Prof. Farmer discussed in a recent Guardian article. While still early on, this new class of technology shows real promise in helping climate policymakers and institutions better navigate the complexity of the energy transition.
This conversation explores how modeling the real economy – firm by firm, shock by shock – could finally give policymakers and institutions the tools to navigate and steer the climate transition.
Q1. Your Guardian piece opens with a striking claim: economics has failed on the climate crisis. That's a bold charge. What's the evidence?
Doyne: The track record is genuinely terrible – and I can speak to it personally. In 2010, I predicted that solar would be cheaper than coal by 2020. We were right. The integrated assessment models, meanwhile, were off by a factor of six in their estimate of how fast costs would decline. For two decades, they consistently predicted renewables would roll out slowly and costs would fall gently. Solar costs collapsed by more than 90% in a decade. Wind followed close behind. If you had used those forecasts to set climate policy, you would have dramatically underinvested in the transition.
These models failed because they ignored what the data clearly showed: that as deployment grows, costs fall predictably because experience accumulates. This is Wright's Law, and it goes well beyond simple economies of scale. But correcting for that raises a deeper question: what actually drives deployment? That depends on the decisions of millions of individual firms, each responding to prices, competitors, and policy. Look at what happened when China stepped into the solar market. Deployment scaled so fast that costs plunged far beyond what anyone predicted, and the same is now happening with batteries and EVs. But standard models can't see any of these dynamics. And that's before you even get to how shocks propagate through those firms. That's where agent-based models become essential.
Q2. You've spent decades developing a complexity economics and agent-based modeling approach. What does that actually mean, and why does it produce a more honest picture of how economies work?
Doyne: The core idea is simple: stop pretending economies reach equilibrium and model what actually happens. Millions of firms and households making boundedly rational decisions, interacting with each other, producing outcomes nobody planned.
Counterintuitively, this becomes more tractable, not less. Agent-based models simulate each agent individually with realistic behavioral rules and let aggregate behavior emerge. Agents learn and adapt as the world changes – and because the whole system is represented from the bottom up, you can see how a shock in one part propagates through every other. The cascading impact becomes visible.
Covid proved it. Standard models assume equilibrium – supply equals demand. The pandemic shattered both simultaneously: demand collapsed as consumers retreated, supply collapsed as industries shut down, and the two shocks collided and amplified through supply chains. No equilibrium model could track that. We mobilized a team overnight, built a model that could, and predicted a 21.5% GDP contraction. The actual figure was 22.1% – and we significantly outperformed the Bank of England and several others.
Sophia: Whether you’re an infrastructure investor assessing transition risk or an insurer pricing supply chain exposure, it means you're finally working with a model that represents the firms you're actually exposed to, not sector averages. And that produces something the existing tools can't: traceability.
When this approach shows that a critical mineral bottleneck delays battery production across three countries, you can trace every assumption behind that number: the supplier relationships, the inventory buffers, the cascade pathway. That audit trail is what regulators and boards are increasingly demanding – and it's something neither statistical models nor LLMs can produce. The decisions that will define the transition are too consequential to rest on black boxes.
Q3. Give us a concrete example of what Macrocosm’s platform can answer that today's tools simply cannot?
Doyne: Our first step toward modeling the whole economy is the energy system: all 30,000 energy companies and their 160,000 assets. Every power station, every wind farm, every generator is a separate agent making decisions about output, investment, and pricing. We're literally modeling the adaptive decision-making of every energy company in the world to see how much each delivers and at what price. It matters because the transition isn't about swapping one technology for another. When a utility invests in solar, it changes the economics for every other player. But standard models can't see that, and so they optimize one imaginary company standing in for the whole sector.
The same applies to supply chains, and the Hormuz crisis is showing that in real time. A standard model gives you a sector-level GDP impact – a single number, weeks later. It can't tell you which firms exhaust critical inventories first, how the shortage cascades through downstream production, or how much larger the total loss becomes than the original shock. Whether it's a power grid or a global supply chain, our platform simulates the dynamics week by week, sector by sector. It tells you not just how bad, but when.
Sophia: That visibility changes what's possible for sustainability decision-makers. Scope 3 reporting, energy procurement, transition planning. They all depend on understanding how disruptions travel through the deeper tiers, where hidden dependencies and emissions reside. Today, most companies can't do that. An automotive manufacturer can't trace which Tier-3 components become the bottleneck when a logistics corridor is disrupted – and the insurer covering them can't see the accumulation risk across their book. Macrocosm makes those pathways visible and quantifiable for the first time.
Q4. There's a gap between a powerful academic model and something a Chief Sustainability Officer, risk manager, or policymaker can actually use on a Monday morning. How do you bridge that gap?
Doyne: These models only make a real-world impact if they're embedded in the environments where high-stakes decisions policymakers, investors, and firms are making. Taking complexity economics from the research frontier into practice is what Macrocosm is doing. Critical to building that trust is empirical validation: we backtest our work against real events, which is rarely standard practice in economic modeling. It's the same reason you trust a weather forecast: not because the data is richer, but because the underlying physics is right.
Think about what weather forecasting requires: not just better atmospheric science, but an architecture that could ingest real-time observations, update continuously, and express uncertainty in a form people could act on. That's the gap we're closing for economics.
Sophia: And closing that gap is exactly what drew me from BCG to Macrocosm. Today, most sustainability teams work from scenario analysis that's already out of date by the time it reaches the board. The world moves; the analysis doesn't. What we're building is a tool you can interrogate as conditions change. When a new tariff shifts where your supply chain's transition risk concentrates, or grid bottlenecks delay a renewable buildout, you need to see what that means for your exposure now, not next quarter. Not a static report, but a capability you use every week.
Q5. What would it look like — practically — if policymakers and institutions had access to this kind of simulation? What decisions change?
Doyne: We want to do for economic planning what Google Maps did for traffic planning – giving people a dynamic picture of the whole system so that they can better navigate the constantly changing world. For policymakers, that matters more now than ever, because the transition isn't happening in a vacuum. Governments are simultaneously managing energy security, industrial competitiveness, and decarbonization, and the interactions between those priorities are where the biggest policy mistakes get made.
Imagine a government stress-testing an industrial strategy and seeing not just GDP projections, but the predicted response of every firm: who invests in renewables, who holds on to fossil assets, where grid bottlenecks or critical material dependencies emerge. They can test whether a policy intervention accelerates the transition or triggers unintended knock-on effects through supply networks. They finally see who bears the cost, who adapts, and where the transition stalls.
Our research shows that a fast transition likely saves the world around $12 trillion. But that requires getting a huge number of interconnected decisions right, across energy, supply chains, and investment. No one can steer a system they can't see.
Sophia: And no institution can navigate a transition it can't map. Whether it's a border adjustment, an energy shock like Hormuz, or a critical mineral constraint – the impact rarely arrives where you expect, or when. Today, most organizations see their direct exposure but not the connections behind it. Macrocosm makes those connections visible. You can stress-test your exposure before something breaks, re-run it when policy or market conditions shift, and defend the analysis to your board or regulators. That transforms how institutions reason about the transition: not from static assumptions, but from a living model of the economy they're actually operating in.
