Saving the World With Fluid Dynamics

Above: a bombastic headline. Below: some banal introspection, but maybe also a startup pitch? Or maybe a request for grant proposals. I’m not sure.

I sometimes wonder, I’m sure like many other washed up ex-tech folks with time and money to spare, about what I could be doing to help mitigate the existential risks to humanity of climate change. Not in the sense of recycling more or flying less (though, that, too), but in the sense of contributing to some human enterprise that stands a chance of moving the needle, even a little bit, on a global scale.

Like Liam Neeson, I have a very particular set of skills, skills I have acquired over a long — well, medium long — career. At least as importantly, there is an overwhelming number of things I’m not much good at and never will be. So, on the one hand, there’s the question of what humanity should be doing about the climate crisis, collectively. But what I want to know is: where are my particular skills, interests, experiences, and connections best applied? What is my comparative climate advantage?

I was surprised recently to come across a specific, plausible answer. Unsurprisingly to my friends, it involves boats.

Commercial shipping burns an immense amount of fossil fuel, and there’s not a super obvious path to transitioning it to renewable energy. Batteries are literally 100x heavier than fuel as a form of energy storage, and so it doesn’t seem realistic to have plug-in electric cargo ships without miraculous advances. Maybe we’ll end up using nuclear (steamships!) but I’m not sure that’s geopolitically palatable. Maybe something involving fuel cells, though that hasn’t been successful in other domains yet. It seems most likely to me that we will end up with battery electric vessels being sustained over long crossings by onboard renewable power generation, both wind (plus ça change) and solar.

To be clear, that would require some radical new designs. Some napkin math: let’s take a Very Large Crude Carrier as an example, because irony. In very round numbers, they’re 1000’ long, and can fit 10MW of solar panels, which would generate 50MWh over the course of a day. But they consume about 500MWh per day, so we need to find a factor of 10 somewhere. Electric drivetrains are more efficient, which helps, and I’m certain you could increase the available surface area with a different design. Add some wind into the mix (but don’t shade your panels with the sails). If you can stomach going slower that helps immensely, because power consumption is cubic with regard to speed - going to 80% of the speed should be able to halve the daily power needs, but increase the time and so the cost of shipping. So it doesn’t seem impossible, but it also doesn’t seem easy.

At any rate, nuclear power research is one of those things I’m not going to be much use at. Ditto battery and fuel cell technology. And I am not a naval engineer, so I will not be designing the next generation of electric cargo ships.

However, for self-indulgent reasons, I have been spending a lot of time recently talking to naval engineers.  They know this transition is coming, and that regardless of exactly how it plays out, power efficiency is going to be an even bigger deal in the future than it is now. What determines power efficiency? How well the shape of the hulls move through the water, at different speeds, trim, and loads; how well the shape of the propeller translates rotation to forwards motion; how well the shape of the superstructure moves through the air, in different wind conditions, how well the shapes of the sails or wind turbines capture wind power; in other words, it’s all fluid dynamics.

The eye opening moment was when one engineer was explaining to me his cutting-edge workflow, which involves using parametric CAD software to programmatically generate a series of candidate shapes for whatever he’s designing, and run each of them through intensive multi-day numerical simulations of fluid flow on a very large compute cluster to figure out which one is the most efficient.

Optimizing generative models using large-scale batch compute? That, as they say, is in my wheelhouse.

So, here’s the pitch: the electric boat and ultimately electric ship designs of the future will put a much higher premium on efficiency, and will be subject to novel design constraints (like maximizing the horizontal surface area on which you could place solar panels). Adapting the tried and true designs of the past will only get us so far. The best new designs will come from human engineers guiding automated searches through a constrained space of possible shapes, evaluated by computational fluid simulations.

The more expressive the tools are for expressing those parametric designs, and the faster the turnaround on those fluid simulations, the better the resulting designs will be - and the sooner we will be able to transition shipping off of fossil fuels.

I think there are some clear startup opportunities here. In particular, this is extremely well aligned with the huge amount of investment going into AI infrastructure right now. One way to frame it is that we need models that can generate, say, hull shapes the way that ChatGPT generates text and Midjourney generates images, but where the prompts are CAD models and physical requirements instead of words, and where the objective function includes power efficiency. Another thing to point out is that Silicon Valley is currently sprinting up the learning curve of how to run very large GPU clusters, and those lessons could be pivoted into running fluid simulations that are cheaper and have faster turnarounds than ever before. If I were planning a venture-funded startup it would be tempting to offer the design tools for free, to make them as widely available as possible to the next generation of engineers, but charge for a well-integrated cloud-based simulation and optimization product.

Alternatively, given more capital, you could try to follow Tesla, with a somewhat similar vertically-integrated master plan”. First, you design, build and sell a small number of renewable-energy luxury yachts: I’m imagining a 60’ electric catamaran that can outrun a sailboat from San Francisco to Hawaii, in greater comfort, without burning a drop of fuel (vs solar” boats like the Silent 60 that rely on diesel generators for longer crossings).  This lets you develop and prove some of the technology and build brand awareness. Then, you scale up the experience to more consumers and scale down the costs, though in this case that probably doesn’t mean selling high volumes of boats, but instead something like offering zero-emission small-ship cruises to Alaska. Over time, you expand the scope, ambition, and capital needs until, somehow, you get to cargo ships.

Do I want to do this? I will at least say this: I want to do this more than I want to do any other startup. If you would want to join it, or fund it, by all means reach out at

But there are other options. For example, I could putter along doing independent research. I can also fund others doing research — if you’re doing a PhD in this space, for example, please get in touch.

I can contribute to OpenFOAM, the main open source tool for fluid dynamics, either with code or with money.

I could try to work for companies already in this space. CAESES is doing the optimizable-CAD side of things, and looks like a really great product. SimScale is doing the cloud-based simulation. They’re both based in Germany, which isn’t ideal, but you never know.

Or… I don’t know. There’s no big reveal here, just me getting some thoughts in order. Maybe you also have thoughts. Send them! This post has already been greatly improved by feedback from PvH, Marc Hedlund, and Sean Taylor.