The Cloud Cartels: Web 2.0
The global flow of data is a story as deserving of a TV show as Narcos. It’s a story of servers, data vaults, hackers, Russian Submarines and the desert.
From the phones in our pockets, to optical fibre networks at the ocean’s floor — it is a vast network where the drug of choice is data known by its street name, bytes.
These are units of information that turn everything into 1s and 0s.
To understand the promise of a decentralised web, we first need to understand the history of the global data flow.
Web 1.0
For most of the 90s, internet architecture was on prem. That’s a fancy way of saying companies ran their websites on their own premises.
Thousands of internet servers were hosted all over the world with each node in the network responsible for maintaining itself.
Each node had to have physical web servers, data storage for user information and networking expertise. If too many people tried to access the site, it would crash because the severs couldn’t keep up.
Web 2.0
In 2006, AWS changed everything.
The internet company realised they could rent out compute power.
Cloud computing was born.
Virtualisation technology meant that one server could be divided into multiple virtual servers. Hardware was decoupled from the operating system.
Basically, Amazon could rent out computers stored in a big warehouse and people could access them via the internet.
The value proposition for companies was on par with fire for our primordial ancestors. They could have all their infrastructure managed by Amazon for cheaper with unlimited ability to scale as there was more traffic on their website.
The rise of cloud computing powered all the SaaS applications of Web 2.0.
And so began the era of centralisation.
Amazon set out to build massive data centres as the economies of scale flywheel started. More data centres lead to cheaper compute which lead to more data centres.
In 2010, Microsoft and Google entered the game with their own cloud centres.
Today, these data centres store most of the world’s information.
Over 90% of companies rely on public cloud infrastructure in some capacity with 60% of all workloads occurring on the cloud. Amazon alone operates 81 availability zones each with one or more data centres.
It’s pretty ironic that a company that started in the e-commerce business makes most of it’s revenue from data warehouses.
Remarkably, from 2008 to 2020, the public cloud market has compounded at an average of 32% per year.
But what’s more incredible is that it’s forecast to compound at 27% for the next 5 years.
To put a 30% CAGR in perspective, if you gave Jeff $10,000 in 2006, you would get a return of $1,900,496 by 2026.
But what does this actually mean for the internet?
It means that nearly a third of the world’s data on the cloud is controlled by one corporation…
The greatest sleight of hand the cloud companies have pulled is claiming that centralised internet architecture is on premise and that the cloud is ‘distributed’.
While each cloud provider’s infrastructure is distributed around the world, the power of the internet is far more concentrated and centralised than ever before.
The cloud cartels own Web 2.0.
Is the power slipping?
Nope. Everyone is going to the public cloud, even the CIA.
The public cloud providers have managed to build a utility on par with motorways, sewerage and energy infrastructure.
The great transition to cloud infrastructure is compounded by many flywheels.
But there is one less obvious flywheel, what happens when all the system admins forget how to manage their own infrastructure?
Imagine if there were only 383,000 people in the world who knew how to do open-heart surgery with one brand of scalpel.
Most system admins are transitioning away from legacy system administrator roles to cloud architects where they manage their organisation’s outsourced infrastructure.
It might be fine for now, but what happens when organisations realise that on prem might actually be better for them?
A16z calculated that the 50 software companies could save $4 billion by going back to on prem for some workflows.
It’s only going to get worse over the next decade.
High switching costs, economies of scale & network effects are all at play. This means the public cloud providers have serious pricing leverage.
If your heart could only be operated on by a few people, they could charge anything.
Web 3.0
A decentralised web means a lot more than just internet architecture, but let’s talk about it only in terms of that lens.
A decentralised internet would be a peer to peer network where resources are owned and shared by the network, not concentrated with one or two entities.
But there is a problem….
To power the internet, we need a lot of compute. And building mega data centres is actually a really good way of achieving that.
Even micro data centres are far more powerful than a peer to peer network of edge devices.
So much so that 60% of Ethereum nodes run in the cloud, mostly on AWS.
Note: These are not mining nodes, they are nodes primarily used for Daaps so the Ethereum network is not relying on the cloud nodes to verify the network.
While the future of Web 3.0 will always be somewhat centralised by large cloud providers, it can and will become more decentralised where it matters.
Here are some projects which give a glimmer of hope of what the internet could be like in the future…
Akash: The world’s first open source cloud provider.
- It is a cloud marketplace that matches anyone with excess compute with developers looking to deploy cloud applications
- You can deploy any containerised applications that can run on the cloud at a cheaper cost with faster speed
- Uses the blockchain to create a layer of trust to see the reliability of cloud vendors
Edge Network: A peer to peer network that leverages edge devices
- It allows developers to use the compute of edge devices such as phones, tablets and laptops
- Anyone can contribute computing resources to the network
- Acts like a content delivery network but allows logic at the edge
Salad: Share compute for epic loot
- It allows people to monetise their idle PCs for gift cards
- Uses network of devices to mine crypto
- Has paid back over $2 million dollars (wild, this market never existed previously)
Predictions for the internet
- I think that with the rise of networks like Akash, people will start investing more into private equity backed data centres. These types of marketplaces can allow a distributed network of data centres to compete with large cloud providers by automating elasticity across multiple availability zones. More cloud vendors = more decentralised internet.
- Edge computing, although a buzzword, will allow more players to develop edge infrastructure getting their foot in the door to IaaS market. We can already see this unfolding with players like Fastly and Cloudflare. While the edge computing market is a compliment to the public cloud, it could very well one day be a substitute for it.
- The next biggest cloud vendor will be a blockchain company not a tech giant
