This makes me think that there is also a limit on the number of subnets. Can we even manage hundreds of subnets? Thousands? Tens of thousands?
That also means we currently have a one-size-fits-all infrastructure, with its pros and cons (including higher costs when they aren’t necessary for the dapp).
According to Dfinity, ICP is designed and engineered to support it. (eg millions of nodes )
I think that is a highly theoretical assertion, and by that timeline, we should have had many more nodes by now.
The main issue is that light can only travel so fast across the world, which imposes significant limits on a decentralized app on any blockchain, including the IC, and handling that amount of users and data across the world will likely create a ton of issues with the current infrastructure.
In a world of finite resources, where decentralizing an application typically costs 10 to 100 times more than keeping it centralized, certain capabilities will always remain either impractical or entirely out of reach for decentralized systems.
I don’t see why not. This is supposed to be the idea. Also note that the IC is not at all finished. It’s a work in progress that is continuing to do great things with each of its milestones. The IC is still in its infancy.
Flexibility and scalability relating to subnets is now coming under focus. I’m optimistic. Fantastic things have already been achieved that would previously have been seen as unrealistic.
This is yet another assertion without any supporting argument.
What do you believe is the fundamental issue with the current infrastructure?
On one hand, Dfinity is engineering ICP to replace the existing centralized cloud stack; on the other, we have an arbitrary claim dismissing its feasibility.
Please provide technical based arguments to support your stance.
I think it’s clear: you can’t entirely replace the existing centralized cloud infrastructure.
Why is anyone building on the IC at all, and why are users using the dapps, if the same things can be built on traditional IT stacks more cheaply?
I think the reason is that they can’t. The IC provides something that traditional IT does not. The need for trust, and the prevelance of institutionalised lying, corruption and/or fraud is a big problem.
I think users will vote with their feet.
There will always be off-chain stuff, much like how you have both closed source and open source software. The open source movement started small, but it’s now absolutely huge. It doesn’t have to consume the world to be successful.
In a world where token transactions and token interaction logic could run on an AWS database for a fraction of the cost, the market still values blockchain sovereignty and tamper-proof security for tokenized assets.
ICP just takes it a step further, applying it to the full stack, not just token records and interaction logic.
Ever wonder why so many Dfinity engineers came from IBM? It makes sense, considering IBM is a major player in distributed computing innovation, from cloud services to supercomputers.
Sure, but this is super far-fetched:
I’m pro ICP, but one thing is trying to fine some niche application that fits the IC capabilities and constraints, and another thing is claiming that everything can and should be run on top of the IC.
I never said that, and there will always be use cases that don’t fit well on ICP, either due to limitations or cost.
Right now, I’m mainly interested in hearing from OpenChat or Dfinity on whether the IC protocol, in its current form, could support onboarding 100M users to oc.app. If not, what changes would be needed to make that possible?
That is what I would like to know
You keep asking people this, but your own technical argument tends to be ‘Dfinity Says X’. I submit that just because Dfinity says X doesn’t mean it has to be so.
Consider that as you put it, “Dfinity is engineering ICP to replace the existing centralized cloud stack”. Well, how is it that, eight years into their research effort with 200 plus cryptographers, engineers and coders on payroll, they have not onboarded a single enterprise of any note?
Think of any startup in the non-crypto world that sets out with the stated aim of onboarding companies and does not succeed in onboarding any firms for nearly a decade. Would the founders of the startup retain the trsut of their investors? Dfinity has kept the community in thrall by continually promising new innovations. Instead of proving their USP as a decentralized cloud, they leapt to on-chain NFTs, socialfi and then Bitcoin integration. None of those worked either. So, now it is all AI stuff and the “self-writing internet”. The claims are always huge but the results in terms of community participation paltry.
Well, I have stopped drinking the Kool Aid. It appears to me increasingly like ICP is a pipe dream.
In its infancy?
What I do not understand is if the ICP is in fact the superior internet platform vs centralized Web 2.0, who really believes the right course of deployment should transpire over decades?
Ain’t nobody got time for that!
For all the problems that IC can solve that Web 2.0 will never solve by default, I would have thought that IC would be a child prodigy by now, blazing the way for the individuals intent on hosting their dApps on a truly worldwide decentralized platform that is environmentally friendly.
Because the energy usage of just online ads-the backbone of Web 2.0- is estimated to account for 2% of total worldwide energy consumption. see Digital Advertising May Be Causing Almost 2% of Global Carbon Emissions – Magic Lasso Adblock
Massive!
So, ‘green’ consumers would onboard to ICP in a heartbeat if doing so helped eliminate online ads and thus, save our planet.
Unfortunately, it appears that ICP has not only not matured enough to handle this massive migration, it is also not able to see the environmental impact of its prolonged infancy.
I say it’s time to grow up.
Subnets are currently limited to 120k canisters, this is because even if a canister is idle it still consumes a tiny amount of compute resources.
But once idle canisters are handled more efficiently, subnets will be able to support over 1M canisters (there is already an item in the IC roadmap for this).
Eventually the limit will be the number of messages into the subnet rather than the number of canisters.
A subnet can already handle over 1k messages per second (provided they are fairly lightweight).
In the future I think it is realistic that a subnet could support 10M canisters (assuming that their wasms are all the same and so can be deduplicated, and that on each round fewer than 1k are active at any one time).
This would allow OC to have 100M users, with up to 10k messages being sent per second, while only requiring 10 subnets.
So today in its current form, what number of people do you think can be successfully onboarded and regularly access any dApp hosted on ICP?
Thanks for the clear and detailed explanation. 
Thanks for your response, @hpeebles! Is the 1k messages/sec per subnet a strict limit? What is the size of these messages? Also, what is the current maximum number of subnets?
What people in ICP largely don’t understand is that “onchain” is just a framing of “stake mixology”. ICP is taking a different approach, skipping the entire “mixology process”. The rest of crypto is taking a more modular approach (market of specializing participants). ICP supporters like to tout that it does everything… But so do already existing web2 alternatives. People like to claim that Ethereum is decentralized, but it wasn’t always viewed that way. Same with Solana. The “token casino” as many ICPers like to call it is “stake mixology”. People like to call “meme coins” useless and focus on “real world assets”, but fail to understand that they’re putting the cart before the horse. Example, why would JPmorgan use Ethereum instead of their own fork of Ethereum (which they’ve done). Another example, The world wide web, just a collection of subnets communicating with each other. Over promising and chasing useless things, when real world adoption came from the “memes” and “hobbies”… “worthless blogs” as investors would call it… then they copy and start chasing blogs for SEO. “worthless videos” as investors would call it… then they copy and start chasing videos for relevancy. “worthless social media followers” as investors would call it… then they copy and start chasing likes and clicks for relevancy.
ICP claims to do everything on chain… But being onchain is just a framing of “stake mixology”.
From the social media wars to the mobile wars, from the browser wars to the search engine wars, and even the ongoing operating system wars, history has shown that technological advancement itself is rarely the primary challenge.
The real battle isn’t about building better technology—it’s about capturing market share. Technology is the easy part; adoption is the hard part. True success requires more than just innovation; it demands a deep understanding of social behavior, user psychology, and strategic distribution. The most advanced technology means nothing if people don’t use it.
I LOVE staking for 6 years and losing 50% in 3 months to get 23 ICP back in 80 days! This project is awesome!
