Sorry. I tried to keep this on the shorter side, but there’s so much to cover.
Algorand is a decentralized, fast-finality, moderate-throughput, forkless, monolithic, gasless, smart contract network with very low transaction fees.
It’s very cheap and easy to use. But it also has high inflation and no plans for economic sustainability past 2030. It’s the only popular PoS network that does not plan to pay for its network security in the long run. These issues are not due to its basic protocol and design, but due to how it’s managed by the Algorand Foundation. The Foundation has focused on marketing and short-term growth at all costs, sacrificing stability and sustainability.
##**Part 1 – Algorand Basics**
Algorand was launched in June 2019 with a pre-minted 10B max token supply that was originally planned to be completely-distributed by 2024. The goal was to design a blockchain that could solve the Trilemma. Instead of solving the Trilemma, it settled on an optimal point with very low fees, moderately-high speed, and moderate security.
* Algorand Consensus Protocol is a forkless variation of Byzantine Fault Tolerance (BFT), similar to what most Proof of Stake networks use.
* In Algorand’s PPoS algorithm, a Verifiable Random Function (VRF) will randomly (weighted by stake) select a validator for the block in secrecy. This provides more security since attackers won’t be able to attack the user in time.
* Algorand uses [~1400 **participation nodes**]( for consensus (block proposal and vote). Unfortunately, nothing is tracking the staking weight of the participation nodes, so we can’t estimate the security of the network. Algorand also uses ~120 **relay nodes** for routing connections between all other nodes and acting as archive nodes.
* Unlike nearly every other PoS network, Algorand does not expect to pay for its network security in the future. Instead, it expects both its participation and relay nodes to continue running altruistically without any rewards. I’m very skeptical on whether this is sustainable given long-term bandwidth and storage requirements. Even if it decided to use the transaction fee sink from its community rewards, its current yearly transactions fees are so low that they couldn’t even pay for a single engineer’s salary.
* **No slashing and the Nothing-at-Stake issue**: Algorand doesn’t slash for for misbehavior, doesn’t use lockup periods, and can’t lose delegators. Thus, it has a pretty big **Nothing-at-Stake** problem, which I consider this a large security flaw. Doubly-so since there are zero economic incentives for participation nodes to keep them honest. It requires trust that nodes are honest and altruistic, which goes against the principle of blockchains needing to be trustless.
####**Performance and Scaling**
* **Fast deterministic finality**: Algorand’s finality is deterministic and [settles in 4.5s]( of which 0.5s is spent on the block proposal. This is faster than most EVM-compatible blockchains, and fast enough to use it as a Medium-of-Exchange for point-of-sales systems.
* **Moderately-High TPS**: Algorand supports 1K TPS even with AVM smart contracts (though it’s currently only seeing 15 TPS of actual activity due to low demand).
* **Uncertain future scaling**: Throughput could theoretically scale to [50K TPS]( with higher block sizes and block pipelining. I would treat this with extreme skepticism. Their 2021 Performance report predicted 2.5s finality and 46K TPS by the end of 2021, and neither update is even scheduled as of mid 2022. It’s also very impractical because 46K TPS requires ~500TB of monthly bandwidth and ~100TB of storage. That’s an extremely high amount of data without sharding or multilayer chains.
* **No sharding**: Speaking of sharding, Algorand is not planning to use it. Sharding has limitations because it produces forked versions of the blockchain that need to be reconciled, which goes against Algorand’s design principles. It would also increase time to finality and introduces additional complexity.
* **Block Pipelining** is a future scaling solution that could increase throughput by 5x. It works by allowing the network to begin working on the next blocks (e.g. 4 blocks) before the current block is finalized. Current stats show that only 1% of blocks fail to stage successfully, so block pipelining should work 95% of the time.
* **Storage**: Unfortunately, as with every monolithic blockchain, there’s no good solution to long-term storage bloat. If Algorand ever reaches the 3000 TPS it needs to be economically-sustainable at its current fee structure, its ledger would grow 6TB in size monthly. You cannot expect archival nodes to run altruistically with that much storage. And if you’re running an Indexer with an Archive node, you need 3-5x that amount. State proofs won’t address this. Fortunately, participation nodes by default only [store about 1000 blocks]( Mainnet is currently [growing at 30GB/mo or 350GB/yr]( during a bear market. That’s still a lot of monthly bandwidth and could be very expensive in certain 3rd-world countries where low bandwidth caps are common (like America).
* **Low demand**: We have rarely seen the mainnet go above 50 TPS in the past several years due to lack of demand. This is one of the biggest concerns for Algorand insiders and why they have spent so much money on marketing lately. Without high demand, their transaction fees are insufficient to sustain the network security.
* **Monolithic**: Algorand is a monolithic blockchain and does not have plans for Layer 2 scaling. The downside is that its ecosystem can’t support layered application-specific blockchains.
* **No outages**: The Algorand mainnet [hasn’t suffered any outages or downtime]( since its launch in 2019. (The closest was in Oct 2021 when OVH cloud service went down and took half of Algorand’s network with it.) Its competitors Solana (multiple times, major), Avalanche (Jan 2022), Polygon (Mar 2022), Fantom (Feb 2021) have all had outages.
* Algorand has [2 classes of smart contracts]( that use TEAL, a Turing-complete language. Both types are atomic (all-or-nothing) and forkless.
* **Layer 1**: L1 contracts are for basic operations (e.g. token swaps) and can be directly computed by consensus in a single round. They have **cheap fees of 0.001-0.002 Algo**.
* **Layer 2**: L2 contracts are for customized complex contracts. These are computed out-of-band by a parallel “contract execution committee”, are stateful, and can take multiple blocks to run. The effects transaction are then batched into a series of L1 contracts and executed together atomically.
* In contrast, any dApp that uses EVM smart contracts has a very low TPS limit (~20 TPS), which is why many newer networks including Algorand have non-EVM optimizations (e.g. AVM).
* **Transfer fees are 0.001 Algo** (currently ~$0.0004), which are artificially-low because they are subsidized. Token (ASA) transfers and most L1 smart contracts also cost 0.001-0.002 Algo.
* If Algo changes sufficiently in price, a governance vote can always readjust the fee schedule..
* However, transaction fees do not pay the participation or relay nodes. They simply go into a [fee sink]( for future community rewards. This is masking the true cost of running the Algorand network.
* Algorand transaction fees only produce about [$150K annually]( many magnitudes smaller than the billions of dollars it has paid nodes so far.
* Staking was replaced by community governance voting in Sept 2021. Periodically, the Algorand Foundation will launch a governance vote, during which individual holders can “stake” their Algorand and vote for interest rewards. This provides a strong short-term incentive for investment. These rewards come from a pot of 2.5B Algo dedicated to Participation Rewards. This is a marketing gimmick considering that no other network needs to pay for governance participation.
* Due to how the governance lock-in is designed, there is a disincentive for centralized exchanges to participate. This creates more democracy in voting. For example, Binance dropped their Algo interest rewards after individual users withdrew their funds in mass before the governance voting, taking away all of Binance’s governance rewards.
##**Part 2 – Tokenomics and Long-term Sustainability**
The biggest issue with Algorand is that its tokenomics are designed for short-term network growth instead of long-term sustainability. The token dynamics have been changed before, and will eventually need to be changed again.
####**No Plans after 2030**
Algorand Foundation’s plans for [long-term economic sustainability]( have been [put off until 2030]( It originally designed for Algo’s 10B supply to be distributed over 6 years, with relay nodes being rewarded until 2022. That plan was [scrapped and remade in Dec 2020]( to extend the deadline to 2030 with rewards for relay nodes to last until 2024. There are no plans for sustainable rewards past 2030, and Algorand’s tokenomics is a ticking time bomb.
The rewards were pre-minted, but there is vesting schedule for those rewards that increases the circulating supply. [Circulating supply is expected to increase]( by 49% in 2021, 20% in 2022, 23% in 2023, before tapering off at 7.5% in 2024 to 2029. (Note that the Messari chart is actually underestimating past supply inflation because Algorand’s token issuance is algorithmic and increases when there is more network activity like in 2021.)
**Current fee structure is Not Sustainable**
Algorand only produces [~$150K annually from transaction fees]( which is barely enough to cover the annual salary of single engineer. If they want to support their current 120 relay nodes, they’ll likely need 100x the current fees unless everyone is super nice and working for free. Fat chance. Bandwidth costs aren’t cheap either, especially if the network grows and needs multiple 10Gbps sustained connections.
####**No economic incentives to run nodes**
There are 2 main types of nodes: Relay Nodes and Participation Nodes
**Relay Nodes** are maintained by a consortium of early investors, VCs, Universities, and other non-profits until 2024. These are being paid for through multiple rounds of massive grants totaling at least 2.5B Algo (worth billions of USD). Algorand is still the covering costs for future decentralized Relay Nodes through its [Community Relay Node Program](
It currently costs [$5-10K/year to run a cost-effective relay node on AWS]( which actually isn’t bad. But that’s assuming the engineer running it is doing that freely, and it doesn’t account for the much-needed network growth. The $150K in annual transaction fees will only cover the salary for 2 part-time engineers. In other words, if the Algorand network stops paying for ALL other rewards (community rewards, staking, governance, development, participation nodes), it still only makes enough to pay for 2 out of its 120 relay nodes.
What happens when all these groups used to getting paid in rewards gradually see their rewards disappear? Are they still going to stick around? Relay nodes are used to being paid billions of dollars that are eventually going to go to ZERO. Algorand Foundation believes that nodes can run altruistically. If they happen to be correct about that, then they made a colossal blunder by wasting billions of dollars on them in the first place.
Algorand’s marketing is absolutely massive. Algorand has paid for a FIFA sponsorship and exclusivity, Times Square ads, podcast ads, TD Garden ads, Napster, and much more.
I haven’t seen any other blockchain spend more on marketing than Algorand. It’s a lighter version of Crypto.com’s marketing strategy, and we know how that turned out.
**Incentive Bait-and-switch Tactics**
Algorand Foundation has a record of attracting groups with reward pools that eventually disappear. They attracted node runners (early relay nodes) with billions of dollars of rewards, set to last until 2024. They attracted stakers and participation nodes with rewards to last until 2022. They then attracted community participation with Governance rewards starting in late 2021 that is currently scheduled to run out in 2030. Most networks don’t provide any additional economic incentives for governance voting because voting is the incentive. It’s just another marketing tactic meant to grow its network.
The subsidized low transaction fees are also a marketing tactic.
The big question is whether a community so used to receiving economic incentives will stick around after those rewards go away. Some might, but will enough remain?
##**Solutions to Fix the Tokenomics**
**Algorand’s tokenomics is very fixable as long as the community is willing to make big changes.** First, the fee sink will need to be adjusted to pay for network security instead of community rewards.
One option is to hope the marketing strategy plays out and the network gets 100x more utility by 2030. There’s not single network among the top 50 smart contract blockchains that saw anywhere near that much real activity even at the 2021 peak. This is highly unlikely due to massive competition between all PoS blockchains, and without community rewards, every other network will be more attractive with staking rewards. The other problem with this is that bandwidth and storage requirements will also increase, leading to even higher costs of running nodes, leading to needing even more utility.
Another option is to use **tail-emissions** along with Ethereum’s **token burning** model. Algorand would need to increase its fee schedule enough to maintain minimum-viable security, which is roughly 100x its current fees. Fortunately fees are so low currently that even with a 100x increase, fees would still be in the pennies range. The problem is that increasing fees so drastically will likely decrease usage activity since Algorand has marketed itself as a low-fee network.
Both of these options still require an end to “max supply”. Instead, Algo would have a “total supply” target of 10B Algo. This is because crypto transaction activity is highly-volatile. For consistent security, you need a model that produces consistent fees, not one that is 1000x higher during bull cycles than bear cycles.
Most likely, Algorand will need to apply a mix of both options even though they partially-counteract each other.