Why did Borland ignore the Macintosh market?
It didn't. There was a short lived Turbo Pascal 1.0 (as UncleBod mentions). Even before Turbo Pascal, Borland ported Sidekick as well as their Reflex database for Mac.
(Taken from the back side of Turbo Pascal for the Macintosh)
But yeah, development stopped by 1988 and it all comes down to size, size and size. Plus product type and to a lesser extent environment.
Size, Size and Size
Back in the 80s the computer market was unbelievably tiny compared to today, and within the professional segment (read: non-home/hobby), the Mac was at best third or fourth in size, trailing way behind MS-DOS, CP/M-86 and even CP/M-80 (*1). Best illustrated by the fact that Apple needed 4 years ('84-'87) to sell the first million units. The 'old' 8 Bit Apple II sold a good 3.5 Million during the same period (*2).
During the same period IBM PC and compatible clones sold more than 15 Million (*3) units. This means Borland would have to invest a similar amount in development and marketing for a market only 6-7% the size. Doesn't sound great for a sales-oriented company.
Product type
Turbo Pascal, the main product of Borland in the mid 1980s (introduced Christmas 1983), was CP/M software. Meant to run on a rather frugal environment using a character-based frontend with terminal type control (in-band signalling). Getting that to DOS (*4) wasn't a big deal, as it didn't offer any notable improvement over CP/M (*5). Pascal did cover those basics quite well, great for cross development between CP/M (*1) and DOS (*6).
Mac applications work completely different. No CLI, no text-based graphics or even cursor keys, plus Mac screen-handling being all about proportional text, something alien to character-based terminal programs. Any port requires a complete rewrite of the UI - which as we all know can easy reach 90% of all application code. This completely kills any dream of portability. Mac development was a different world done in separate departments, just ask Microsoft.
It was no real goal for Borland to invest in developing GUI features into (Turbo) Pascal as there was no GUI for either of their markets (*7). It took until 1987 to include graphics (BGI) as a core part (*8,*9) and it wasn't until 1990 that Turbo Vision added (somewhat *10) serious support for GUI.
Environment
The different front end structure was already mentioned, but the (original) Mac was even more restricted due its hardware. With just 128 KiB, the memory situation was worse than on an 8-bit computer. At the same time, it couldn't do 80 characters per line - something every DOS program would expect.
I would like to think that the often-cited 'shortcomings' are only about Turbo Pascal, not Sidekick or Reflex, as both were ported before Pascal and to the original 1984 Mac. 1985 brought the 512 KiB Mac, removing the biggest obstacle, for Borland to port Turbo Pascal. A good indication is also that the 1.0 Version required 256 KiB minimum RAM (*11).
During the 80s and 90s, Borland developed several amazing, cutting edge developer products.
Not so much, as they were buying most.
Their Turbo Pascal and Delphi, an Object Pascal development environment,
Delphi might sound old today, but it only debuted in 1995, at a time when the Mac/PC quota was even worse than in 1987 and at the same time the total Apple numbers were still not high enough either.
were very popular on DOS and Windows.
You may want to add CP/M to that. It was the important kickstart for Borland to cover the vast existing base of CP/M - and support the move from CP/M to MS-DOS that happened exactly at the time when Turbo Pascal was introduced. Not to mention the great public success in Europe with systems like the CPC.
Slowly over time, however, Borland lost the fight against Microsoft, even though their products were technically superior to Microsoft's offerings.
Which might be debatable. They'd been a leader on the CP/M to DOS move, but missed the transition from DOS to Windows. Sure, Delphi was great, but it was way too late. When Win 3.x became a hit after 1990, Borland's main feature was still text-based. Windows support in version 6 and 7 was quite clumsy. So any early and not so early developer for Windows looking for a compiler had to use MS - possibly even MS-Pascal. By the time Delphi came in 1995, only a few professional developers hadn't switched to a more Windows enabled environment - which usually meant C.
It seems they preferred charging absurdly low prices just to stay afloat than to expand into the Macintosh market.
Low price was always the main argument to buy Borland. It undercut any other supplier by far. 50 USD was about a quarter of what any other compiler would cost. It was so low that people chose it who otherwise would not have bought anything.
It would have been the ideal market given Pascal was the lingua franca for Macintosh development for years.
Not least that the Mac itself was done in Pascal.
Oddly, enough, while Borland was slowly losing the market to Microsoft, other vendors prospered selling Pascal compilers exclusively for the Macintosh.
The same way as language developers flourished in other niche markets.
*1 - While CP/M systems might've no longer been top-selling in the mid 1980s, it's the installed base applications get programmed for. Same reason why there's still a good market for COBOL even today. New machines sell with new OSes, but all machines are served by later application development. So having a way to offer Software for CP/M-80 customers and MS-DOS with little or no variation gives a great incentive to using TP.
*2 - Not to mention the installed base, which adds >1M for Apple II and zero for the Mac.
*3 - According to the same numbers, which to me look a bit US-specific. So while the PC:Mac ratio might be similar in other parts of the world, those areas may have had some variation.
*4 - After all, DOS always was just a somewhat modernized CP/M clone.
*5 - Better file handling system calls aren't really a point once file handling is wrapped (abstracted) by a HLL.
*6 - In fact, I earned some change on side projects in 1984, programmed for PC users, but developed under CP/M-80 on my Apple II :))
*7 - Well, by 1987 it was MS-DOS only, as 3.0 was the last published for CP/M-80.
*8 - There was a graphics toolbox added separate product with 3.0.
*9 - 1987's Version 4 also marks the point when it first enabled production of EXE files, allowing more than 64 KiB of code - easy to forget how primitive TP was compared to 'real' compilers :)
*10 - Turbo Vision was primarily text-based, but allowed transition to a GUI by working with menu, dialog and window abstraction instead of plain text. Kind of like IBM's SNA.
*11 - Shortly before Borland introduced 1.0, Apple finally gave us the Mac Plus with all the goodies of the 512ke plus SCSI plus a base RAM of 1 MiB, expandable to 4 MiB, finally a size enabling serious work. INHO the point from which the Mac can rally be classified as a PC. In 1987 the Mac II series was introduced and Borland upgraded Pascal to 1.1, supporting its 68020 as well.
