TLDR: Statists measure success based on the state, so in order to convince them that anarchism us successful we would need to stop being anarchists.
"For anarchists who do know something about anthropology, the arguments are all too familiar. A typical exchange goes something like this:
Skeptic: Well, I might take this whole anarchism idea more seriously if you could give me some reason to think it would work. Can you name me a single viable example of a society which has existed without a government?
Anarchist: Sure. There have been thousands. I could name a dozen just off the top of my head: the Bororo, the Baining, the Onondaga, the Wintu, the Ema, the Tallensi, the Vezo…
Skeptic: But those are all a bunch of primitives! I’m talking about anarchism in a modern, technological society.
Anarchist: Okay, then. There have been all sorts of successful experiments: experiments with worker’s self-management, like Mondragon; economic projects based on the idea of the gift economy, like Linux; all sorts of political organizations based on consensus and direct democracy…
Skeptic: Sure, sure, but these are small, isolated examples. I’m talking about whole societies.
Anarchist: Well, it’s not like people haven’t tried. Look at the Paris Commune, the revolution in Republican Spain…
Skeptic: Yeah, and look what happened to those guys! They all got killed!
The dice are loaded. You can’t win. Because when the skeptic says “society,” what he really means is “state,” even “nation-state.” Since no one is going to produce an example of an anarchist state—that would be a contradiction in terms—what we’re really being asked for is an example of a modern nation-state with the government somehow plucked away: a situation in which the government of Canada, to take a random example, has been overthrown, or for some reason abolished itself, and no new one has taken its place but instead all former Canadian citizens begin to organize themselves into libertarian collectives. Obviously this would never be allowed to happen. In the past, whenever it even looked like it might—here, the Paris commune and Spanish civil war are excellent examples—the politicians running pretty much every state in the vicinity have been willing to put their differences on hold until those trying to bring such a situation about had been rounded up and shot."
“The only way you could disagree is if you didn’t read it” – come on, we’re really doing this? It shouldn’t be surprising to find disagreement on this exact point on Hexbear, a predominantly ML forum.
Graeber is saying that if everyone in a modern nation-state voluntarily dissolved their state and tried to live by anarchist principles, they would face violent reaction. He’s right! But he doesn’t follow this by arguing how anarchists could avoid, deter, or resist this violence, short of forming a state themselves. He suggests that discussing the violent reaction he foresees is basically an unfair line of reasoning: “The dice are loaded. You can’t win.” The ML response is that it’s not only fair, but crucial to discuss the entirely predictable problems your movement will encounter. If you don’t have an answer to those problems, you need a different plan.
The ML answer to the problem of violent reaction is a state that serves the interests of working people. That’s only “loading the dice” in the argument about the necessity of a state if anarchists can supply a better answer. Graeber has none, at least not in this excerpt.