How MicroAI plans to make $500m/year in revenue by 2030: 🧵
High Frequency Trading algorithms are one of the largest actors in traditional finance. Our biggest competitor made over $2bn in revenue this year.
But we're already two steps ahead.
1) Developing with Julia:
Julia is a new high-performance programming-language designed for technical computing. It was specifically developed to address the need for high-performance numerical and scientific computing without sacrificing ease of use.
HFT is a form of algorithmic trading where securities or tokens are bought and sold in extremely short time frames, often microseconds.
The sector DEMANDS high-performance computing resources, low-latency systems, and efficient algorithms to execute trades at the FASTEST possible speeds.
Julia’s high-performance capabilities make it well-suited for the rapid execution of trading algorithms. Its type declarations and multiple dispatch are being used to optimize Micro AIs algorithms for even higher performance. Julia has built-in support for parallel and distributed computing, which is essential for handling the massive data sets and computational needs of HFT.
Julia is PERFECT for High Frequency Trading.
Since it's so new basically no one is using it yet.
We are first movers.
---
2) Combining Deep Learning (AI) with HFT
We are NOT first-movers in this sector. While this is gate-kept to the public, most HFT houses are already using AI-powered HFT algos, which are profitable.
We take this on a new level though. Since we don't operate in TradFi but within the crypto market, we have unlimited access to data.
How you might ask... Etherscan. Every single order that has ever been placed on Ethereum can be used to train Micro AIs MLT-HFT algorithms.
We are already over 5,000,000,000 data-points deep in training.
With our Clustering Algorithm, which prevents HFT overfitting and at the same time allows for more data-points we expect the following amount of trained data-points:
2023 Q4: 11,000,000,000
2024 Q4: 50,000,000,000
2025 Q4: 180,000,000,000
2026 Q4: 2,000,000,000,000
This is possible through data-clustering.
2,000,000,000,000 data-points is EXPENSIVE.
Really expensive. While solutions like
@rendernetwork &
@ionet_official will ease the developmental costs, it will still be costly.
Which is why we spent over $50,000 this month on a new data-aggregation system, which aggregates the data-points and makes it cheaper to train models on.
We expect data-aggregation to be the centre of our MLT-HFT developments in 2023 Q4.
Thus while models continue to train in the coming months, they will continuously get cheaper.
This is not a sprint but a marathon. We're ready to dominate this $50bn market-sector by 2035.
Welcome to history.
Welcome on this journey.