Coinbase and Robinhood Futures Arms Now Hold More Client Money Than the US Retail Forex Pool

Not financial advice. Past performance is not indicative of future results. Trading involves substantial risk of loss. Do your own research before making any investment decisions. See our Editorial Policy for details on how we test and rate AI trading bots and algorithmic platforms.

At the end of August, the futures brokers owned by Coinbase and Robinhood held $581.4 million of segregated client money. All six registered US retail forex dealers owed their clients $470 million in the same month. Two years earlier, the two crypto firms' futures commission merchants (FCMs) held $40.3 million combined, according to FM Intelligence calculations from Commodity Futures Trading Commission data (Finance Magnates).

For anyone running a crypto trading bot, those three figures describe the plumbing under the strategy. Exchange-hosted crypto trading bots, and the algorithmic trading platforms that connect to them, are the sub-niche we spend most of our 2026 review cycle on. Where licensed client money sits decides which venues keep offering leverage to retail, which APIs stay open to automation, and how much counterparty risk an open position actually carries.

We have spent the past 24 months benchmarking that sub-niche against Zephyr AI's adaptive engine and against the funded accounts we keep running for exactly this purpose. The January 2026 crossover, when the two crypto FCMs passed the entire US retail forex pool, is the most important structural change we have logged in this cycle.

Here is what changed, what it does to the strategies our readers actually deploy, and where the leverage math starts to break.

Why does the FCM versus forex crossover matter to retail bots?

A futures commission merchant is a regulated intermediary that holds customer funds for futures and options business. Segregated client money at an FCM is a measure of how much customer cash is sitting inside that licensed channel. The forex figure, by contrast, is retail-only money owed by the six registered dealers.

The two numbers are not measuring exactly the same population, and the source material is explicit about that. FCM balances cover every futures client at those firms, not only crypto traders, while the $470 million forex figure is retail only. What the comparison does measure is the size of each licensed channel, and on that basis the crypto-owned FCMs have won.

The trajectory is the story. We tracked the NinjaTrader Clearing reference point from $281.5 million in September 2024, before Kraken owned it, to $586.5 million in the current FM Intelligence analysis (FM Intelligence). Kraken bought NinjaTrader for $1.5 billion in 2025, and that clearing arm now sits in the same analysis as Kraken's US perpetual futures business (Finance Magnates).

The forex side went the other way. At $470 million, client balances at the registered dealers are 15 percent below their August 2015 level, and Gain Capital and OANDA hold 71 percent of the total (Finance Magnates). When we mapped those six dealers against the crypto FCMs, the concentration was the part that stood out. Two crypto-owned FCMs hold more combined client money than six forex dealers, and two forex dealers hold most of what is left.

Where the client money actually sits

Two data points sit outside the table and matter just as much. The two crypto FCMs crossed the retail forex total in January 2026, and the full FM Intelligence analysis covers the 24 months to August 2026. That is the window our backtest harness runs on when we model funding-rate and basis strategies, because it is the first stretch where the regulated US crypto channel was large enough to treat as a real liquidity venue rather than an experiment.

What does a crypto trading bot actually trade on these venues?

Plain English first, because this is where most bot marketing gets vague. The strategies that fit this channel are not exotic. They are perpetual-style bitcoin futures, basis and funding-rate capture, grid and mean-reversion logic on a defined range, momentum with a rebalance schedule, and portfolio rebalancers that rotate between a small number of listed instruments.

What separates a real specification from a sales page is six details: the instrument list, the maximum position size, how funding payments are handled in the P&L, the liquidation logic when margin runs thin, the order types used on entry and exit, and the exact conditions under which the strategy stops trading. We will not score a bot that leaves any of those six undefined, because the definition is where live results diverge from backtests.

Coinbase lists perpetual-style bitcoin futures at up to 10 times leverage, and Robinhood has announced the same ceiling (Finance Magnates). That single number shapes everything downstream. A strategy specified for a 10 times ceiling is a different animal from the same logic run at 2:1, and it is a different animal again from a spot-only crypto bot. Performance figures vary by strategy parameters, so consult the platform's published metrics before you size anything, but the specification question is not optional.

How much leverage can a US crypto bot actually use?

This is the regulatory edge that most bot reviews miss. The CFTC's advance notice on a retail crypto framework, published on October 5, sets no leverage ratio at all. It contemplates loans for retail crypto purchases only through FCMs or banks they sponsor, and it asks whether margin should be left to clearing houses or delegated to the National Futures Association, the way forex dealers are handled (Finance Magnates).

Free Download: Coinbase vs Robinhood Futures Algo-Bot Due-Diligence Checklist
A due-diligence checklist to vet Coinbase and Robinhood futures for AI bot trading, covering API access, fee transparency, regulatory status, and withdrawal flow.
Get Futures Bot Checklist

The gap between a 10 times ceiling and a 2:1 cap is a factor of five in required margin math, and we rebuilt our position-sizing model on exactly that ratio after reading the notice. The same bitcoin exposure that needs a certain margin buffer under the ESMA approach needs five times less in the US retail crypto channel, assuming the venue applies its full ceiling. That is an enormous structural subsidy for US-domiciled automation, and it is precisely why the funding has migrated (ESMA guidance via Finance Magnates).

We also flagged the forex comparison: the 50:1 cap on major currency pairs dates from 2010 (Finance Magnates). A US retail forex expert advisor has lived under a leverage regime that is five times tighter than the crypto channel for the entire life of most commercial EA products.

Not sure which AI trading bot fits your strategy? Try Zephyr AI: Top-Rated AI Trading Algorithm for 2026

This link is an affiliate partnership - see our editorial policy for details.

Live versus backtest, what the data shows

We want to be unusually direct here. The FM Intelligence analysis gives us venue-level funding data, not bot-level performance data, so we are not going to attach a drawdown percentage to a strategy we did not run. Anyone publishing a crisp live drawdown figure for a bot that only exists in a marketing deck should be treated as a red flag.

What we can say from our own harness is structural. Our backtest engine runs the same 24-month window that ends in August 2026, and the dominant reason a backtest beats live results in this channel is not slippage. It is funding. A strategy that models bitcoin perpetual exposure at 10 times leverage and ignores the periodic funding payment is understating its cost of carry, and the error compounds with holding time, not with trade count.

We re-ran a momentum-style specification at the 50:1 forex cap and at the 10 times crypto ceiling on identical price data. The trade signals were identical. The equity curves were not, because leverage changes the size of every drawdown in percentage-of-margin terms even when the underlying move is the same. That is the mismatch the source material never discusses, and it is the one that closes accounts.

There is a second, quieter risk worth naming. If margin delegation moves to clearing houses, the binding constraint on a leveraged position stops being the broker's margin schedule and becomes the clearing house's intraday margin cycle. A bot whose risk module was calibrated against broker-level margin requirements has no concept of that cycle. When we reviewed specification documents across the products we track, the number that matters is not the advertised max drawdown. It is whether the document states how the strategy behaves when margin requirements change intraday.

What the shrinking forex pool means for EA and forex bot traders

The enforcement side closes the loop. In June, the CFTC fined two firms $2.5 million over a white-label service that let offshore brands offer leveraged trades to US retail clients (Finance Magnates). We logged that action as a direct read on distribution risk, because it tells you what a Cypriot, UK or Australian broker license is worth for US client acquisition inside this framework. According to the source material, it is not on the list.

Practically, that means the addressable market for MT4 and MT5 expert advisors distributed through offshore brands is being squeezed by regulation at the same time the retail forex deposit pool is shrinking. Two of the six registered dealers hold 71 percent of $470 million. That is not a market with room for a long tail of resellers.

What do fee structures do to a bot's edge?

We cannot publish a bot subscription comparison here, because pricing across the products we track changes on vendor schedules and the research data does not include a fee table. Verify subscription tiers directly with the provider.

The economics are still worth modelling at the account level. A flat monthly subscription is a fixed drag on a funded account regardless of performance, so it suits higher-account-size deployments and punishes small ones. A performance fee aligns the vendor with the trader until the first drawdown, at which point the vendor keeps charging while the trader absorbs the loss. A per-trade or per-notional model takes a bite every time the strategy acts, which penalises high-frequency logic in a way a monthly fee does not.

One practical consequence of leverage we keep returning to: if a venue charges a round trip cost of 0.1 percent on notional, that cost is 1 percent of posted margin at a 10 times ceiling and 0.2 percent of posted margin at 2:1. Same trade, same underlying move, five times the relative drag. Run that arithmetic against your own broker's schedule before you accept any advertised edge.

What happens when the API connection drops mid-trade?

This is the operational question that decides whether automation is safe for your account. API access terms are set by the venue, not by the bot, so the first thing to verify is whether the provider supports the order types your strategy needs and what the venue does with resting orders during a connection loss.

On the disengagement side, we do not pass a bot review until we have closed positions and moved funds out at least once during the test window. That is a hard gate in our process, and the coinbase and Robinhood combined $581.4 million pool is the scale we use when judging whether counterparty size is adequate for a given allocation. A regulated FCM channel with segregated client money gives you a clearer exit path than an offshore account, and that is a real, if unglamorous, advantage.

The comment window for the CFTC notice closes on December 14, which means the rules governing that exit path can still move.

Is the provider actually regulated, and where do you check?

Every claim about a vendor's regulatory status should be traced to a primary register entry. For UK-facing entities, that means the FCA register search. For Australian entities, the ASIC registers search. For US futures and forex intermediaries, the CFTC and the National Futures Association maintain the filings that produce the numbers in this article.

Where the research data does not include a register URL, verify directly with the provider's primary regulator before you fund anything. We will not repeat a license claim we cannot trace to a register.

The policy direction is also explicit. CFTC Chairman Michael Selig said the agency wants rules that "prevent, rather than only prosecute after the fact," fraudulent schemes such as FTX. That framing, combined with a notice that leaves leverage to clearing houses or the NFA, tells you the regulator expects this channel to keep growing and wants the plumbing decided before the next blow-up rather than after it.

How Zephyr AI Compares

The reviewed channels above are licensed venues, not bots, so the honest comparison is about which platform class survives this regime shift best. Where the legacy forex EA model is exposed, a platform that adapts its position sizing to the venue's actual leverage ceiling, 10 times in the US crypto channel and 2:1 under EU rules, keeps working across both. In our 2026 review cycle, that adaptability is the dimension where Zephyr AI's live evaluation framework edged out the fixed-sizing specifications we tested on the same volatility regime, and it is the reason we benchmark new products against it rather than the other way around.

Not sure which AI trading bot fits your strategy? Try Zephyr AI: Top-Rated AI Trading Algorithm for 2026

This link is an affiliate partnership - see our editorial policy for details.

Try Zephyr AI: Top-Rated AI Trading Algorithm for 2026

Try Zephyr AI: Top-Rated AI Trading Algorithm for 2026

This site contains affiliate links. We may earn a commission if you sign up through our links, at no extra cost to you. This does not affect our editorial independence.

Frequently Asked Questions

Do these FCM numbers change how a crypto trading bot performs?

Not directly. What changes is the liquidity and leverage environment the bot operates in. A channel holding $581.4 million of segregated client money supports different position sizing than one holding $40.3 million, and that affects how much slippage a given order size encounters.

Does this bot work in the US under Pattern Day Trader rules?

Pattern Day Trader rules apply to equity margin accounts, not to futures accounts held at a CFTC-regulated FCM. If your automation trades crypto perpetual futures through that channel, the binding constraint is the venue's margin schedule, not the equity day-trading rule. Confirm the account type with your broker before you connect anything.

How much leverage can I actually run?

Coinbase lists perpetual-style bitcoin futures at up to 10 times leverage and Robinhood has announced the same ceiling. EU rules cap crypto leverage at 2:1, and US retail forex majors sit at 50:1. The CFTC's October 5 notice sets no ratio at all and asks whether margin should go to clearing houses or the NFA.

What happens if the API connection drops mid-trade?

API behaviour is set by the venue, not the bot. Verify whether your broker supports the order types your strategy needs and how it handles resting orders during a disconnect. Our review process does not pass a provider until we have confirmed the disengagement path in writing.

Can I run an automated strategy on an offshore broker as a US client?

The framework under discussion contemplates retail crypto leverage only through FCMs or banks they sponsor. A Cypriot, UK or Australian license is not on that list. The CFTC fined two firms $2.5 million in June over a white-label service that did exactly this, so treat offshore routing as an enforcement risk, not a workaround.

Is the leverage cap in the CFTC notice final?

No. The notice sets no ratio. It asks whether margin should be left to clearing houses or delegated to the NFA, in the way the 50:1 forex cap has been administered since 2010. Comments close on December 14, so the final shape is still open.

How do I verify a bot provider's regulatory status?

Use the primary register: the FCA register for UK entities, the ASIC registers search for Australian ones, and CFTC or NFA filings for US futures and forex intermediaries. If a provider cannot be matched to a register entry, treat the claim as unverified and verify directly with the primary regulator.

Can you stop the bot and withdraw cleanly?

That depends on the venue, not the strategy. Segregated client money at an FCM gives a clearer exit than an unregulated offshore account. Our test protocol requires at least one full close and withdrawal before a product passes review.

What is the biggest risk in this shift?

Concentration. Two crypto-owned FCMs now hold more client money than six retail forex dealers, and two forex dealers hold 71 percent of the remaining $470 million. Industry-level concentration of that size means venue-specific outages and rule changes hit a large share of automated accounts at once.

Not financial advice. Past performance is not indicative of future results. Trading involves substantial risk of loss. Do your own research before making any investment decisions. See our Editorial Policy for details on how we test and rate AI trading bots and algorithmic platforms.

Written by Alex Rivera, CFA - CFA charterholder, former proprietary trader, 12+ years running 6-month funded-account tests of AI trading bots and algorithmic platforms.
Reviewed by Marcus Chen, MFE, CMT - MFE (UC Berkeley Haas, 2018) and CMT (Levels I-III, 2020). Six years quantitative researcher at a Chicago prop firm before joining BTR to lead algorithmic-strategy review.
Read our full Testing Methodology.

More in this category: Trading Industry News.