The primary benefit of a giant company is efficiency. They are able to provide their goods and services at a lower cost to the rest of the economy because of their scale. The downside is that any level of corruption/cheating within the company will become magnified by an order of magnitude, again because of their scale. And the thing is, the majority of companies are anti-competitive or otherwise have some tendency to "cheat" the market. That's human nature mixing with the incentives of the economy. Usually that's fine in the sense that we tolerate it as fellow, flawed humans. But when the company is giant that foul play becomes a big problem, and we see all the pain that tends to accompany giant corporations like those pointed out in the article.
So, is the increased market efficiency worth it? Doesn't the anti-competitiveness cancel out any benefit that that efficiency may have had? Or worse, perhaps the anti-competitiveness out-weighs the improve efficiency, and giants are actually hurting the overall efficiency of the market?
That would be my devil's advocate argument. If giant corporations tend toward anti-competitiveness by their nature, and if that anti-competitiveness is out-weighing any benefit from increased efficiency, then it makes economic sense to "ban" giant corporations altogether.
Another, perhaps playful, way to look at it is as an optimization problem. We can view market efficiency as the loss function, which we wish to maximize. Our current economy roughly applies Stochastic Gradient Descent (SGD) to this problem. Try a couple things at once, measure average loss, and then propagate that loss to the entities within the market roughly proportional to their contribution to the loss. In other words, companies that helped make the market more efficient are rewarded. Companies that did the opposite are punished. Those companies adjust one direction or the other, and we iterate again. The problem with naive SGD like this is that it _overfits_. Any data scientist will tell you that. Overfitting the economy results in a very optimal loss in the short-term (training set), but terrible loss (market efficiency) in the long-term (testing set). And that's exactly what we see. Giant corporations like Walmart give us really low prices, but in the long run they are a strain on local and global economies and stamp out competition that would have helped find a more optimal, global efficiency.
Digressing, this is of course idealistic. Making large companies outright illegal would be insane in the current political environment, would not be practical, and ultimately won't solve the issue. A more nuanced approach is appropriated, and perhaps that's exactly what Warren is attempting here. To that end, and if the above argument holds water, I'd have to say I agree with her approach and wish her luck. A little L2 regularization will go a long way to making out economy more robust.