In what cases should a CS grad consider doing a PhD if not to go into academia?

Depends on your TC

It’s very common and helpful to take a few years to work then return to PhD, I would take the offer now and keep getting to know folks in research groups you might be interested in and asking about outcomes (how many go back to industry) and really learning about why you might want/not want to commit 5 year to get one . Bottom line I think there is no rush and I would take the offer

How about passion?

If you want to be a applied or resreach scientist then a phd is mandatory

There is no reason to do a Ph.D if you don't intend to go into academia. If you're going private sector, then those YOE you're losing are much more valuable than the letters after your name.

Faster green card lmao & TC or GTFO

If you go for a PhD, you will lose money (~6 years of slave wages), but it may let you do research rather than engineering later (provided you won’t drop out like many people starting PhD), which may be more enjoyable depending on your attitude. At the same time, you may consider MLE kind of a role if you like to be in between research and SW.

Was on a research fellowship for a year in grad school during master’s, and my advisor still asks me to date to go do a PhD program with him whenever we get a meal together or randomly text. I’m posting some things that were useful to me and hopefully can be to you too. Here’s my thought process: I chose to go to industry after the master’s was up mostly because I thought I could do what I was doing/interested in doing in grad school with greater impact (and compensation) in industry. I told myself that I would only want to go do a PhD program if I was interested in a specific topic and can learn things related to it well mostly in an academic setting. My advisor told me that folks who have left academia to go to industry typically don’t come back (he had many friends do that) because the money is honestly great (esp with your background) and after a few years they start thinking about having or actually having a family, and that weighs heavily towards not doing a PhD program. After being in industry for a few years, I strongly agree on both points. The gravity to stay in industry will be strong from a financial perspective. I have friends that jumped between the two and usually they’re quite financially well-to-do and don’t worry about the net worth aspect. I talk to many SWEs at work and lots of “kids these days” really care about TC, so you may not worry about it now but you may get affected by it over time in industry. That being said, it’s usually in a more research-oriented role such as in X, G brain, or deepmind that you’ll find your PhD useful. Most ML SWE or applied scientist roles are going to be straightforward applications of ML (think classification) and not researchy at all.

There can be lot said but my 2 cents. Don't do PhD if you are not determined. Don't just weigh in your interests, interests can dry up during many phases of your PhD and thus resulting in negative thoughts unless you are determined. You are probably aware of yourself to decide on that. If you are determined and interested, go work on your PhD. Dont think about losing TC. Apart from the degree, PhD teaches more than what you can expect about yourself.