The traditional VM-per-developer approach is indeed prohibitively expensive.
Since developer workloads are bursty, we encourage resource sharing when possible. We've seen a Coder pattern where developers share a box, through Kubernetes or Docker, to cut costs by an order of magnitude.
The Coder team dogfoods on a single 32-core dedicated machine that costs us $40/m per developer while building our code 40% faster than my M1 MacBook Pro.
I had the same problem of slow builds like you, but instead of throwing hardware at it, I moved from Typescript to modern Python 3.10 with all new features and type hints.
I enjoy a really fast development feedback loop (build times passed from minutes to milliseconds. I organised Jinja2 templates in "components", etc.
The only drawback so far is that the output is not a proper SPA (I'm using Flask), which is fine for what I'm doing, and where needed I bridge the usability gap, 2022 vanilla JS was surprisingly practical. If one feels extra fancy, they would drop in Vue.js I guess.
I'm really fast with this stack, I coded a whole self service subscription management system, product activation codes, support tickets management, etc, for my SaaS alone in 6 weeks.
Excited to try the newest cool features and higher speeds of Python 3.11.
On GCP, I'd have to pay $2400/y for an 8-core machine as powerful as my M1 MacBook.
I get the portability and reproducibility arguments, but it's hard to justify that expense.