The AI Coin Nobody Understands

With the Crypto x AI narrative running hot in recent weeks, a lot of the focus has been on three main tokens: VVV, GRASS and CHIP. These assets cover important areas of the AI trade, from private inference and pretraining data for LLMs to GPU financing.

But there is an AI token launched in 2017 that much of the market still does not fully understand. Part of the reason is that the team behind it operates a regulated hedge fund and is intentionally careful about what it shares publicly. Yet with firmwide AUM approaching $1B and the team now buying back the token, NMR deserves a closer look.

Numerai has curated clean and standardized stock and crypto market datasets that data scientists can use to train predictive models. These datasets contain over 1,000 different indicators and are built to the same standard as those used by quant firms to develop trading strategies.

Using AI, data scientists build models that aim to generate alpha uncorrelated to the market. Numerai then aggregates these predictions into a single Meta Model.

Numerai currently operates three funds using the same underlying alpha, with differences in leverage and structure to suit different investor preferences. Firmwide AUM now stands at around $770M, with backers including Paul Tudor Jones and JPMorgan.

Numerai had its only down year in 2023 before rebounding with its strongest year on record in 2024, delivering a 25.45% net return. Numerai is not trying to outperform the S&P 500 by taking more market risk. Instead, it hedges out broad market exposure to isolate its own alpha. That performance and the uncorrelated nature of its strategy help explain why JPMorgan committed up to $500M in capacity to Numerai in August 2025.

So how does the token fit into all of this, especially after Numerai raised $30M at a $500M valuation in November 2025?

Data scientists can submit models without staking, but only staked predictions are included in the Meta Model. Staking NMR also allows contributors to earn rewards based on the following formula:

Reward or loss = NMR staked × clip(3 × CORR60 + 15 × MMC60, −1, +1)

CORR measures prediction accuracy against the target, while MMC measures the model’s contribution to Numerai’s combined Meta Model. The formula can also slash poorly performing models, serving as a supply sink while encouraging contributors to stake only on models they believe can continue generating alpha.

Currently, 316 data scientists have staked 316K NMR, representing around 4.1% of the circulating supply. At the same time, new tokens are emitted to reward contributors. To date, $24.1M has been paid out, with monthly payouts averaging around $313K since April 2025.

We estimate that around 590K NMR was paid out over the past 18 months, averaging 32.8K tokens each month. Numerai has also bought back roughly $3.2M worth of NMR, or approximately 360K tokens based on their prices at the time of purchase between July 2025 and July 2026.

These tokens were then used to reward contributors instead of issuing additional NMR, meaning the protocol’s net inflation has remained minimal over the past year.

So where does the next growth unlock come from? What’s a newsletter without some alpha.

While much of the previous focus has been on the hedge fund and its performance, the team is now looking to scale its fundraising efforts and has made several strong hires in this area.

Here is some napkin math. If total AUM scales to $1B over the next year, which is less than 2x from current levels, a 2% management fee would imply $20M in annual fees before adding in performance fees.

NMR has no contractual claim on these fees, so the token thesis depends on Numerai continuing to use part of the economics from the business to buy back tokens. Even modest buybacks could be meaningful with NMR currently trading at a market cap of around $75M.

The flywheel would look something like this:

More buybacks → A higher NMR price → Greater incentives for contributors to earn NMR → Better models and more NMR staked → More NMR locked and stronger fund performance → More buybacks.

I recognize that more transparency from the team is needed. I also recognize that their hands may be tied because, at the end of the day, they operate a hedge fund and face restrictions on how much they can disclose.

Still, NMR is an interesting token that could benefit from the broader adoption of AI, and it is one I am watching closely.

— Kunal

Despite the market selloff, Etherfi’s weekly protocol revenue continues to grow, with card revenue now accounting for 63% of revenues.

All of finance is coming onchain and the institutional heavyweights are headed to DAS London this November this 10-11th.

Grab your tickets for unmatched access to the leaders bringing institutional finance onchain.