This is the script of Sam Bankman-Fried (SBF)’s speech on World Blockchain Conference 2021:

Hey, my name is Sam Bankman-Fried. I’m the co-founder and CEO of FTX. Thanks for having me at the World Blockchain Conference 2021. Excited to talk to you guys today.

I’m here to talk about sort of the state of DeFi. We’ve seen a ton of growth in Decentralized Finance over the last year. It’s almost right to remember 15 months ago, DeFi was not talked about that much. And really the growth started with the launch of Compound in the spring of 2020. Since then, we’ve seen an explosion of interest of capital and attention and promise of the DeFi ecosystem.

The biggest kind of projects that have launched so far generally fall into sort of one of two different categories. One of them is a borrow-lending protocol. What that means is it’s a place where you can deposit one type of blockchain asset and use it as collateral to borrow another type. So if you had Bitcoins, but you wanted some dollars to fund your activities, you could use those as collateral for dollars. Compound and AAVE have been two of the largest borrow-lending protocols.

The other common type is DEX – Decentralized Exchange. It’s an onchain program that you can use to trade cryptocurrencies with other players. Uniswap and Sushiswap are two of the biggest DEXs today.

So far, those have been hallmarks of DeFi. We haven’t seen that many other types of programs really get that big yet, although one could imagine it. One example is social media. You can imagine onchain social media. We haven’t seen it yet that much, although BitClout has tried. But I think we might see more of that.

Pain Points of Building DeFi on Ethereum

A lot of what we’re gonna see though is, I think, a maturing of the types of products that already exist, because it’s really, really awesome that you can have an exchange that is built into blockchain. There’s a lot of improvement left to get in the existing ones. And all you have to do is compare a standard DEX to one of the larger centralized exchanges like FTX today. You can kind of feel the difference in the capabilities of the two. And part of this is working on the product. But part of it is also working on the blockchain technology.

I spent a lot of time poking around the DeFi ecosystem, helping to build projects, to guide them, invest in them. And it’s been a really memorable and really wacky, frankly, year. It’s grown faster than almost anything I’ve seen before. And some of it has been really exciting. Some of it has made no sense, but there’s a lot of potential in it.

What is that potential? And how can we get there?

The potential, as I see it, is could you get a billion people using a DeFi ecosystem? Could you get $1 trillion of assets in DeFi? That’s sort of the level of like having really reached the upper reaches, which is the potential. Could you get 5, 10% of the world’s activity happening on a blockchain? That’s what we should be aiming for in the upside cases of DeFi.

I’m not saying it will get there. It very well might not. But that is sort of like a huge success case. That has really strong requirements. The biggest one is scalability.

As everyone has become aware, the current DeFi ecosystem can’t meet the needs that people have for it. And the biggest problem is just the throughput.

On Ethereum, you can get about ten transactions per second. If you look at the biggest companies in the world, they tend to have millions of transactions a second. Whether you’re looking at posts on Weibo or Twitter, whether you’re looking at trades on one of the top exchanges, you’re generally looking at something like millions. So ten is not even close to being close to enough.

The ecosystem notes this. There are a lot of attempts to speed up DeFi. Some of these are entirely new layer 1 blockchains. Others are attempts to beef up Ethereum with a layer 2 or with ETH 2.0.

As everyone go through sort of like what those options are and what the pros and cons. The first is that ETH as we know today might not exist in a few years. And in fact the plan is to transition to a new Ethereum blockchain ETH 2.0. ETH 2.0 will be something like 100 times faster than Ethereum, which is good, but again, it has ten right now. So you can do 10 things a second for the whole world. You need to get to a million. Getting to 1000, which is where ETH 2.0 will be is not nearly enough. That alone isn’t gonna solve the problem, although it’ll help a little bit.

It’s also sharding, which means that you can have many copies of ETH 2.0 running at once. As you can get up to a large number of total transactions per second, but not in one composable ecosystem. They’re gonna be isolated from each other. And that’s the main drawback of ETH 2.0 is that because it’s gonna be sharded and that’s how it’s gonna get up to the throughput that it needs.

You’re gonna run into this issue, if you build a program on one of the copies of ETH2, one of the shard, a second program on a different shard, those can’t interact with each other. As you can imagine, trying to split a DEX between two different shards and if you buy Ethereum with USDC on one shard, you can’t then go use your USDC and sell it for TUSD on a second shard, because your Ethereum is on the wrong shard. And in general, it’s really nasty to have to split up, especially one industry between shards. And so I think it’s a big drawback of ETH 2.0 as a solution.

Another type of scaling solution that’s ready today is layer 2 and what layer 2 is basically is when you create a new blockchain, you run that a faster blockchain and then you periodically write down kind of the state from that new blockchain back to Ethereum. You still try and ultimately have the information end up on Ethereum, but most of the processing is happening on this faster blockchain. That is potentially a really powerful solution.

The problem that we’re seeing I think there’s a lot of layer 2 solutions is that they’re just still not fast enough. Even though they are a lot faster than Ethereum is, there’s still something like a few thousand transactions a second, and so that’ll help. But it just won’t help enough. If you put a 10,000 transaction per second layer 2 on top of Ethereum, even if you ignore sort of the awkwardness of transacting between the layer 2 and Ethereum and moving assets back and forth, you still have the fundamental problem that 10,000 is not enough. You need a million. Even if you solve that problem, it often takes like weeks to move assets literally weeks between Ethereum and layer 2 for sort of complicated reasons. And that might be able to cut down. That’s a second pain point.

I do think layer 2 could work, but in order for them to work, you’re gonna have to start with a layer 2 which is itself fast enough.

Solana: Building a Blockchain with 1 Million Transactions per Second

And so taking a step back, what I think you really need to reach the moments of DeFi is you need a single blockchain that has 1 million transactions a second. Once you have that, you can decide how it interfaces with the rest of the ecosystem. You can decide how you store data. Whether you write it back to other blockchains and other things. But you need to start with at least one very fast blockchain.

That is actually hard. It’s very difficult to design a decentralized, especially geographically decentralized blockchain that can handle, you know, tens of thousands, let alone millions of transactions per second, but it is doable of the existing blockchains.

I think that Solana is sort of the one which has the ambition to do this and which has a roadmap for it. Almost no other existing blockchains are even trying to get anywhere close to millions, which I think these are not being ambitious enough. And they’re not thinking of what DeFi could be.

That’s not to say that Solana will necessarily win or get there. Someone else could build a new blockchain that does or DeFi never grows that much, then this isn’t a problem.

But I do think that Solana is the most promising of the currently existing blockchains in terms of ability to handle one giant composable ecosystem. And that’s where Serum comes in.

About a year ago, I helped build out the start of the Serum ecosystem. And what Serum is, is basically a bunch of applications built on the Solana blockchain. The first of them is a DEX. What makes this different than other DEXs on DeFi?

I mean first of all, it’s on Solana, which means that it can be way faster. It takes less than 1 second to do a trade. Trades cost a tiny fraction of a penny to do rather than dollars which it costs on Ethereum. You can do tens of thousands of trades per second. So it’s massively more scalable.

Also, because you have the throughput, you can allow people to actually be placing and canceling orders like you do on a normal exchange or order book. The reason you’ll see these on automated market making DEX is that it just would be prohibitively expensive on Ethereum to be canceling an open order or place an open order. If it costs $30 each time you have to do that. But when it’s at a hundred of a penny, it’s way more practical.

So the Serum DEX is a fully onchain DEX. That is actually a full exchange experience with an order book, a matching engine, time price priority and the whole standard experience. And if you use it, you can actually do all of this today. That is sort of like that was the start of the Serum ecosystem.

And it really wasn’t attempt to answer the question if you’re trying to build an onchain ecosystem for a billion people, where did you start? The answer you start on blockchain that’s gonna be massively scaling. You build the most powerful primitive financial instruments you can. An order book is one of those.

Exchanges that have order books and matching engines are really at the heart of most finance. And they hadn’t existed on chain before this and in sort of a reasonable way.

So that’s where it started. But there’s been a lot more now. And people have built out AMMs on the Serum DEX. People built out borrow-lending protocols. And we’re starting to see more and more development in the ecosystem, which is really exciting. And it can all take advantage of all the rest of it.

And so one really cool example I want to use is Raydium. It is an automated market making DEX built on Solana, like Uniswap or Sushiswap, where market makers can just put their capital in a pool, earn interest rates on it, and that automatically gets used to provide liquidity to people want to trade. But one cool thing is that those trades that liquidity happens through the Serum DEX. And so anyone trading on Serum can automatically access all of the liquidity of Raydium. And Raydium can automatically get all the flow of anyone trading on Serum, which really shows the power of composability, which is one of the core principles of DeFi that you can take two decentralized programs on the same blockchain and glue them together. That sort of I think the start of what I hope will be a really promising, hopefully really scaling a DeFi ecosystem.

Share DeFi Ecosystem with Institutions

What else are we seeing? One big thing that we’re seeing not just in DeFi, but actually across all of crypto is institutions are starting to get involved.

Prior to this year, almost none of the flow in crypto was really coming from regulated financial institutions. That is starting to change. It hasn’t really changed yet, but the signs are there.

And as we talk to banks, mutual funds, pension funds and a whole host of financial institutions, we hear really consistent thing, which is that they were not previously involved in crypto, now they will be even if they don’t know exactly how.

So it’s sort of like to be determined exactly what happens there, but institutions are coming. You’ve seen a lot of interesting results about on the custody and exchange side, but you’re also starting to see something like that in DeFi, where people contend with the fact that the current DeFi ecosystem is awkward from regulated perspective.

There’s a lot of unanswered questions about will large banks be comfortable custody their funds on a DeFi protocol? Will they require a centralized custodian there? If so, how would that work? Is the permissionless nature gonna be an issue? Do you need to have sub structures of DeFi where only known counterparties can transact?

And we’ve seen a few different protocols including Compound and AAVE start to set up on the sort of walled gardens meant to sort of market themselves to institutions. And we’re going to see some of that on Serum as well. I think there’s gonna be some products coming out soon for that.

I don’t know what the future is to be honest, because fundamentally what we’re seeing across the entire crypto ecosystem is that a lot of people who want to get involved, but are used to a very different system.

In crypto, exchanges are full stack products and some stuff happens directly on blockchains. In institution finance, every step of the process is a different company, but all the companies have the same structure. And it’s all centralized custody in a revokable way. Those are just very different premises and they don’t interface very well with each other. Because in order for a centralized custodian to be able to promise what they want to to their customers, they have to know that their counterparties will be playing by the same rules. But that’s not sure if their counterparties are using crypto norms. This is a really awkward interface.

I don’t know what’s gonna happen. I don’t think anyone does. We’re gonna figure this out over the next few years as regulated financial institutions and cryptocurrency infrastructure providers negotiate and talk to each other and try and figure out what the future of this looks like and we’re gonna see this on DeFi as well.

I don’t think there’s any conclusions there yet, but I think there’s some people think they have conclusions. I think they’re over confident.

There are people building great things. They might matter, they might not. We’ll see.

DeFi BUIDLers Need to Have goals

What would I do if I were to be building DeFi right now? Obviously I have helped build somethings to DeFi. But one thing that actually I haven’t done that much is to work on actual layer 1 blockchains. I’ve helped advise a lot of protocols building on blockchains and I have dealt a lot with teams building blockchains, although I’m not myself on one.

I think my biggest take away from actually dealing with the teams building the blockchain themselves is, first of all, understand what your goal is as a company. Is your goal to build a system to help facilitate current DeFi transactions very soon? If so, you just optimize for how easy it is for people to get started and how similar to existing systems. If your goal instead is to try and build the long-term system that might have a gigantic DeFi ecosystem. I think you should think really hard about what that’s gonna require, because I think it requires a lot and in particular part is very high throughput, requires as fast of transactions as possible. Composability, paralleled ability all will still be fully decentralized. That’s tough.

It’s not undoable, but it’s hard, it’s a hard technological problem. You’re not going to get there if you don’t try and so from the very beginning, you have to be thinking about it. From the moment you start building out your protocol, you need to think about where you need to get in the end. And you need to make sure that you can and will get there. And you need to make sure you have a roadmap for this. I’m not saying you need to have a million transactions per second on day one. But you can’t ignore the possibility of ever getting there. I think a lot of blockchains are anything that’s sad. It’s gonna really curtail how big they can get.

That’s the biggest thing I would say, think hard about where you need to get in the end, what’s gonna be necessary to get there and make sure that your roadmap has a plausible technological path to get from here to there. Those are some of my thoughts.

As always, you can follow me on Twitter (@SBF_Alameda). You can follow FTX and project Serum on any number of social media sites.

I would be really excited to see more people building in the ecosystem. That’s a single most important thing from decentralization perspective, but also just because that’s what’s gonna build really cool projects.

So thanks for having me!