What happens to software moats when building software gets much cheaper?

One thing I have been thinking about from a value investing perspective is what AI does to the durability of software moats.

Historically, part of the advantage of a strong software company was that reproducing the product required significant engineering talent, time, and capital. AI is reducing some of those barriers quickly.

That does not mean software companies suddenly have no moat. If anything, I think it makes the distinction between the product and the actual moat more important.

The things that seem more durable to me are distribution, switching costs, proprietary data, customer relationships, regulatory position, network effects, and deep integration into customer workflows.

There is an interesting hardware parallel too. Manufacturing ecosystems, particularly in China, made many physical products cheaper and faster to produce. The strongest companies did not stop having moats. Their advantages shifted toward scale, supply chains, brands, distribution, proprietary processes, and other things that were harder to reproduce.

For investors, I think the question becomes less about whether a company has great software and more about what remains difficult to copy once producing the software itself becomes much easier.

I wrote a longer essay exploring this idea and the hardware comparison here: https://mehmetmhy.com/posts/modern\_moat/

Curious how people here are thinking about this when evaluating software companies today.