$s = 'Stack-Overflow Questions Tags Users';
echo preg_replace('/b(w)|./', '$1', $s);

the same as codaddict’s but shorter

• For unicode support, add the u modifier to regex: preg_replace('...../u',

Something like:

$s = 'Stack-Overflow Questions Tags Users';

if(preg_match_all('/b(w)/',strtoupper($s),$m)) {
    $v = implode('',$m[1]); // $v is now SOQTU
}

I’m using the regex b(w) to match the word-char immediately following the word boundary.

EDIT:
To ensure all your Acronym char are uppercase, you can use strtoupper as shown.

Just to be completely different:

$input = 'Stack-Overflow Questions Tags Users';

$acronym = implode('',array_diff_assoc(str_split(ucwords($input)),str_split(strtolower($input))));

echo $acronym;

$initialism = preg_replace('/b(w)w*W*/', '1', $string);

If they are separated by only space and not other things. This is how you can do it:

function acronym($longname)
{
    $letters=array();
    $words=explode(' ', $longname);
    foreach($words as $word)
    {
        $word = (substr($word, 0, 1));
        array_push($letters, $word);
    }
    $shortname = strtoupper(implode($letters));
    return $shortname;
}

Regular expression matching as codaddict says above, or str_word_count() with 1 as the second parameter, which returns an array of found words. See the examples in the manual. Then you can get the first letter of each word any way you like, including substr($word, 0, 1)

The str_word_count() function might do what you are looking for:

$words = str_word_count ('Stack-Overflow Questions Tags Users', 1);
$result = "";
for ($i = 0; $i < count($words); ++$i)
  $result .= $words[$i][0];

function initialism($str, $as_space = array('-'))
{
    $str = str_replace($as_space, ' ', trim($str));
    $ret = '';
    foreach (explode(' ', $str) as $word) {
        $ret .= strtoupper($word[0]);
    }
    return $ret;
}

$phrase = 'Stack-Overflow Questions IT Tags Users Meta Example';
echo initialism($phrase);
// SOQITTUME

$s = "Stack-Overflow Questions IT Tags Users Meta Example";    
$sArr = explode(' ', ucwords(strtolower($s)));
$sAcr = "";
foreach ($sArr as $key) {
   $firstAlphabet = substr($key, 0,1);
   $sAcr = $sAcr.$firstAlphabet ;
}

using answer from @codaddict.

i also thought in a case where you have an abbreviated word as the word to be abbreviated e.g DPR and not Development Petroleum Resources, so such word will be on D as the abbreviated version which doesn’t make much sense.

function AbbrWords($str,$amt){
        $pst = substr($str,0,$amt);
        $length = strlen($str);
        if($length > $amt){
            return $pst;
        }else{
            return $pst;
        }
    }
    
    function AbbrSent($str,$amt){
        if(preg_match_all('/b(w)/',strtoupper($str),$m)) {
            $v = implode('',$m[1]); // $v is now SOQTU
            if(strlen($v) < 2){
                if(strlen($str) < 5){
                    return $str;
                }else{
                    return AbbrWords($str,$amt);
                }
            }else{
                return AbbrWords($v,$amt);
            }
        }
    }

As an alternative to @user187291’s preg_replace() pattern, here is the same functionality without needing a reference in the replacement string.

It works by matching the first occurring word characters, then forgetting it with K, then it will match zero or more word characters, then it will match zero or more non-word characters. This will consume all of the unwanted characters and only leave the first occurring word characters. This is ideal because there is no need to implode an array of matches. The u modifier ensures that accented/multibyte characters are treated as whole characters by the regex engine.

Code: (Demo)

$tests = [
    'Stack-Overflow Questions Tags Users',
    'Stack Overflow Close Vote Reviewers',
    'Jean-Claude Vandàmme'
];
var_export(
    preg_replace('/wKw*W*/u', '', $tests)
);

Output:

array (
  0 => 'SOQTU',
  1 => 'SOCVR',
  2 => 'JCV',
)