That means the state, or associations of people who value privacy (who may not all be technically savvy!) who can pool their resources and offer something normal people can use. This, unlike the decentralised web, could engender the same kind of network effect that whatsapp and facebook benefit from.
I've written about this and heard more people make the same argument lately: it isn't technology that will bring about privacy on the internet, but a democratic organisation, founded by concerned citizens, governed by anyone who has an interest in promoting privacy and run on behalf of anyone who wants to benefit from it.
There are a lot of benefits to this approach: you lose the engineering challenges of VC backed companies (no data mining operation) and of cyberpunks setting out to let everyone be their own little island (protocols and amateur operators). Jurisdictions can offer different advantages, like company forms with regulated purposes written into their founding documents, or more favourable privacy laws.
Off the shelf software can scale all the way that is required for this sort of thing (think apache kafka as the basis for a messaging system like WhatsApp) and the organisation would only have to be financially sustainable like Mozilla, rather than profitable like Facebook.
It also puts privacy into the hands of normal people, who can take part in governance and protect their data that way, rather than having to learn techie things they don't care about.
Talking about the decentralised web is great, but we often forget what motivates people to want to decentralise in the first place: the fact that their desires are not something the market can cater for, and that what they believe should be their right is simply not available any other way. If we want privacy, we should think about all the means by which we can provide it in the general case, including by collectively owning the means of communication.