Do "chip design" and "tape out" costs on older nodes reduce over time?

[Xebec]

Curious -- I always see reports about chip design costs going up for each new bleeding edge node. Do these design costs actually decrease over time for older nodes, perhaps due to improved techniques, software, or AI/ML design tool enhancements?

Likewise, does it become cheaper (over time) to 'fab that first chip' after a design is completed, on a given mature node?

If yes to either of these, are they substantial reductions? for example, is 65nm more accessible now to small customers than it was 5-6 years ago? (or 28nm vs 3 years ago?)

Thanks!

 

[Paul2]

Mask costs go down for sure

 

[Staf_Verhaegen]

With recent development in open source EDA tools making them usable for older nodes should also decrease the design cost

 

[Barnsley]

Mask costs go down for sure

That some hard truth.

Though there has been some reversal due to capacity constraints and the FABS are having to spend some if that extra revenue instead of pocketing it all themselves