Late last year, I wrote an article describing a startup I would like to exist, namely a company (or research organization) focused on advancing capacity to both target where cloud seeding operations should be conducted and to assess, measure, and monitor their efficacy. Both are likely more important than creating the deployment architecture to deliver a seeding agent into clouds: Making predictions about which are the right parts of clouds and the right conditions and making it possible to disambiguate which results are attributable to the operation versus not are necessary functions to improve over time.

I’m under no illusion that deployment is necessarily easy, considering the types of conditions in which cloud seeding can be most effective (inclement weather). But there are a lot of people in the world who are well-positioned to build better drones for specific ends. Plenty of people’s lives depend on it doing so every day in Ukraine and Russia. 

There are not as many people who are well-positioned to materially improve cloud data. Hence, I’m excited about a company that came out of stealth on Tuesday this week: Recast Systems. I’ve known some team members for a while and am glad they’re working on the problems and opportunities laid out in my original post. Here’s how I described the crux of the unlock back in late 2025:

In plainer words, in addition to seeding clouds, if you can form progressively better hypotheses about where and when to seed, and can then track, evaluate, and attribute to what extent you were successful, you have a workflow that is structurally set up to drive recursive improvement. To my understanding, this is a large part of what Recast Systems is focused on. Full disclosure, I am a small investor in Recast, and have, on a handful of occasions, perhaps offered them some moderately useful advice, so I had an inside scoop on some of the state of play when I penned the first piece (winks). I am occasionally prescient, not psychic. 

In the absence of this closed and reinforcing loop, there are many failure modes for any cloud seeding business or operation. The first is being stuck in a “spray and pray” modality forever. Another is struggling to scale a business. If you can’t prove that what customers are paying for is actually producing more rain, scale will be a struggle. Same goes for not being able to credibly claim that seeding efforts are as parsimonious as possible (focusing on the best, rather than all, opportunities for seeding). Cloud seeding has enough failure modes as is, ranging from the technical and scientific to the consternation of conspiracy-peddlers. 

I doubt I have said anything yet that anyone will disagree with. Rainmaker, the cloud seeding startup that effectively re-launched the category after decades of dormancy, certainly also emphasizes the importance of validating its results and forms and tests hypotheses about where to seed (as this article in Harper’s—written from an outsider’s perspective and also well worth a read—references). 

That said, there are two lines of inquiry I want to introduce that are much more debatable.

The first is whether making rain will necessarily be the economically most impactful and/or environmentally most meaningful use case for cloud seeding. As Recast alludes to in their launch video, the use cases of cloud seeding aren’t confined to making it rain more. Cloud seeding can theoretically also facilitate interventions to disrupt and avert extreme weather. There’s making it rain when you need it. And making it not rain when you need it. Droughts aren’t the only rain-related drag on GDP; floods can be quite damaging as well. 

Of course, there’s a lot to prove before anyone is going to claim to have successfully used cloud seeding to intervene to avert bad weather, the same way there is a lot more still to prove for operators claiming to use cloud seeding to make more rain. Recast’s coming-out-of-stealth posts tout hundreds of cloud seeding flights for the Texas and New Mexico governments, but (understandably) doesn’t offer a ton of detail on the results thereof. Still, suffice to say, if, longer-term, they can play the weather both ways, that certainly expands the market.

The second is to what extent the insights from cloud seeding operations will be useful to other domains and disciplines. Recast calls itself a weather company, not a cloud seeding company, which is a positioning choice I imagine is quite deliberate. From my vantage point, the capacity to recursively improve understanding of clouds, rather than the capacity to recursively seed them more effectively, is the most interesting component of this discussion.

In general, viewing clouds as principally interesting for the rain they might produce is a bit myopic. Clouds impact the entire Earth system in all kinds of highly leveraged ways. For instance, clouds are one of the largest sources of reflectivity (albedo) on Earth and play a massive role in regulating Earth’s climate. Earth’s reflectivity has declined markedly since ~2000. Overall, the IPCC identifies losses of reflectivity as the second-largest human-made contributor to climate change after greenhouse gasses, and diminishing global cloud cover is thought to be the largest contributor to the losses of reflectivity (larger than losses of snow and ice). A lot also depends on the type of cloud, though; some types of clouds can cause rather than reduce warming under certain circumstances and conditions. Which just speaks to the complexity of the topic. Moreover, some of the largest uncertainties in modern climate science in general, like aerosol-cloud feedbacks, are, well, also related to clouds!

To be sure, those examples are orthogonal to cloud seeding. But the point remains: Better cloud data are useful for many fundamental scientific and climate-related challenges (and, again, opportunities). For a squarely relevant example, in April, I wrote about a non-profit program, the Arctic Stabilization Initiative (ASI), that is assessing whether it’s possible to responsibly dissipate mixed-phase clouds in the Arctic to allow more heat to escape. The goal would be to reduce pressure on various Earth systems in the Arctic that are at risk of destabilizing due to accelerated warming in the region (the Arctic is warming four times faster than global averages). The potential approach, mixed-phase cloud thinning, would use cloud seeding to dissipate clouds rather than to draw more precipitation out of them, cementing points we established earlier about cloud seedings' versatility as well as the versatility and complexity of how clouds impact Earth systems.

2006 was the moment of lift for the cloud computing industry, when AWS launched Elastic Compute Cloud and Simple Storage Service. Hopefully, in another twenty years, we’ll look back at 2026, with the launch of Recast, ASI, and Rainmaker’s and related players’ progress, as another moment of lift for a cloud industry. Not just for cloud seeding, but for what I will lovingly refer to as “big cloud” (for lack of a better term), and which I intend to capture many different efforts to better understand clouds and to evaluate to what extent it may be possible to intervene in them.