It's actually quite interesting to take a look at how spellcasters have changed in StarCraft II. One of the most interesting aspects is spell research.
In StarCraft II, there are exactly 6 spells that require research. I don't mean per-race, I mean through the entire game. Some of them aren't really "spells" in the traditional sense. Psi Storm, Hallucination, and Seeker Missile are the traditional spells. Weapon Refit (gives BCs Yamato), Ghost Cloaking, and Banshee Cloaking are the others.
Every other spell, every one doesn't require research.
In SC1, every actual spellcaster only had one native spell, and some like the Arbiter didn't even have that. Beyond that, you had to research. Even spells on non-dedicated spellcasters like Corsairs had to be researched.
So let's explore what effect this has on the game.
What does spell research do? Effectively, it bumps a spell up a half-Tier. To the extent that a game of StarCraft plays out in a series of Tiers, a spell that requires research is much like a unit that requires a new tech building.
This means that Defilers (for example) require a Defiler Mound to build, but they become much more useful once you have produced the equivalent of another tech building. On the plus side, this means that all Defilers you produced while researching something useful immediately get better. On the downside, we have the SC1 Queen.
Queens just aren't worth it. 150/100 for the Nest itself, and 100/100 for each Queen. But a Queen can only cast Parasite (and infesting a Command Center isn't even a spell; it's simply an ability). The Queen is thus almost useless without spending extra money and precious time to research either Ensnare or Spawn Broodling.
If a spellcaster cannot quickly appear on the battlefield and make a difference, then it cannot be used as a defensive counter to some tactic or unit your enemy is using. If Queens are going to require 4 minutes from the time the Drone first starts the Nest to the time when they can actually do something, then they cannot be an effective short-term reactionary unit.
This means that typically, you need to see farther into the future to react to something with spellcasters. You have to know, before starting on your Templar Archives, that your opponent is doing something that a High Templar spell can stop. This effectively means that good researched spells must have general utility.
These spells can't screw around; they have to be seriously powerful. Looking at the popular researched spells in SC1, you see that this is very much the case. Psi Storm, Dark Swarm (admittedly doesn't require research, but Consume is basically required for Defilers, so it counts by proxy), Stasis and Recall: these spells are to be respected and feared. They change games.
But they are less than half of the spells in SC1. Those other spells don't change games, and they require too much effort/time to acquire. So nobody uses them.
And this is the purpose of having fewer researched spells in SC2. It allows you to quickly get something of value into play. It means that spells that have limited functionality or utility can still be used within those limitations. Force Field is not the kind of spell you would bother researching; it has fairly limited utility and is not always devastating to the enemy. But you don't have to research it; Sentries get it for free.
So again, look at the 6 researched spells in SC2. Psi Storm again, Seeker Missile, Hallucination, Weapon Refit, and the two Cloaks. Yamato can make BCs into devastating units against single targets. Psi Storm is obviously strong, even in its current not-as-powerful-as-SC1 state. Hallucination has lots of potential: it can confuse an opponent, create meat-shields for your Stalkers/Immortals, or simply act as scouts that allow you to see up cliffs. Seeker Missile is a semi-dodgeable Psi Storm. The cloaks force your opponent to get detection if they want to deal with the unit in question.
What this also does is allow spellcasters with semi-strong spells to immediately make a different on the battlefield. If you see Immortals coming, you can throw down a Ghost Academy and get Ghosts with EMP. You may not be able to stop the first Immortal strike, but because EMP is standard on all Ghosts, you are able to effectively react with a spellcaster.
Another thing you will note is that they got rid of those silly +50 max energy upgrades for spellcasters. These were replaced with upgrades for +25 starting energy for newly constructed spellcasters.
This shows the difference between how spellcasters were thought to be used by Blizzard and how they were actually used in SC1. It was intended that people would use casters sparingly, husbanding their energy and making due with regular units until a critical point, when actual spells would be brought to bear.
This also fits with the idea of having lots of spell research. If you are keeping your spellcasters around until they get full energy, or even the overfull +50 energy, then the time spent researching additional spells is not that bad. And if you are not building 10 High Templar, you have more gas available to do that research with.
Obviously, that's not how spellcasters were actually used. In the most degenerate case, High Templar are basically walking Psi Storm bombs. They are built, they drop Storm, and they are either killed in battle or become Archons. Arbiters are also semi-disposable, though this is as much due to their cloaking field and their priority targets for EMP as anything else.
Essentially, what this means is that the 50 starting energy for a caster is considered highly valuable. And this makes sense. The construction time for most casters is generally not as slow as the time it takes to gain 50 energy. Thus if you need another Psi Storm, it is generally faster to make a new High Templar than to wait for him to get 75 more energy.
This explains a lot. It explains why Arbiters are more valuable than High Templar: they take longer to produce than it does to regain that 50 energy. Similarly, Defilers are valuable because they take longer to produce than it does for them to Consume a Zergling.
Since the most efficient way to use spellcasters in the game is with this pattern, Blizzard changed the +50 max energy research into +25 starting energy research. For High Templar especially, this is a boon. It allows a newly Warped HT to instantly cast Psi Storm. A new Sentry can cast Guardian Aura immediately. A new Infestor can immediately cast Fungal Growth. And so on.
The game is still in Beta, so it is hard to say how this will work out in practice. One thing that would make the starting energy upgrades more valuable would be to globally reduce the starting energy of all casters and increase the starting energy upgrade respectively. The reasoning for this is as follows.
The time it takes to get 75 energy when starting at 50 is significant, but not huge. In some cases, this may be the travel time to the battle in progress. However, if you start with 25 instead of 50, now you have a pretty long time. The research is much more valuable then; not only can you cast your favorite spell, but it also allows you to cast the other spell from the start.
There is one other thing that is interesting when surveying the SC2 spellcasters: spell energy costs. Of the 3 main spellcasters in use in SC1, two of them were long-duration spellcasters. The mere presence of an Arbiter had an effect, so it didn't rely exclusively on its spells for its effectiveness. And Defilers, once Consume was researched, could quickly and effectively regenerate their energy. In both of these cases, spell cost is not a huge factor.
For all other spellcasters, it is. The SC1 Queen is our best example of this. Spawn Broodling is a spell that is worthy of research due to its power. It instantly kills one unit. It's 150 energy cost compared with the utility of this power is what makes it unworkable. Thinning out the number of Siege Tanks would be valuable, but 8 Mutalisks have more general utility than 4 Queens with Broodling. So as useful as it potentially is to kill even quite a few of a Terran's tank force or even just Medics for their M&M ball, it simply isn't worth the cost.
If Spawn Broodling had instead cost 75 energy, then you might have something of value.
Now, if you look at the spells in SC2, few of them go above 75 energy in cost. The only ones that do are Hallucination and the two Mothership spells. The latter is on a big-ticket item, and both of the spells are very powerful. The former is on an early-game caster that can be built en masse. Mass Hallucinations can be very dangerous, so keeping that to a specific strategy is important.