US Retail Forex Deposits Snap Three-Month Slide, Then Stall Near $470 Million, and What It Means for AI Trading Bots

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.

Customer deposits at the six US retail forex dealers registered with the Commodity Futures Trading Commission rose in July for the first time in four months, then edged lower in August. Those dealers owed retail clients $469.96 million at the end of August, according to CFTC data reported by Finance Magnates. July added $9.27 million, a 2.01% increase that lifted the total to $471.58 million. August gave back $1.61 million, or 0.34%, leaving the industry barely above a June base of $462.31 million that the trade press flagged as the lowest since at least September 2023 Finance Magnates.

This is a macro funding story, but it sits inside the sub-niche we test every month: AI trading bots and algorithmic trading platforms, and the retail accounts that keep them funded. When the deposit pool stalls near $470 million, the questions that matter to a bot user change. Does the subscription fee still make sense on a smaller account? Does the broker keep an API open to automated flow when retail balances are shrinking? We have run this category against Zephyr AI's adaptive engine across our 2026 review cycle, and the funding backdrop is the one variable almost every bot vendor leaves out of its pitch deck.

What the CFTC data actually shows

Two things happened in the two months that matter. Gain Capital, the StoneX subsidiary behind FOREX.com, drove the July recovery almost single-handedly. It added $7.03 million, about three-quarters of the industry's net increase for the month, and closed July at $204.12 million. In August it barely moved, ending at $204.16 million.

OANDA went the other way. It shrank in July by 0.34% to $132.45 million and again in August by 0.81% to $131.38 million, the lowest monthly figure since at least September 2023. The FTMO-owned broker sat at $148.44 million a year earlier, an 11.50% contraction over twelve months.

One caveat matters more for bot traders than for anyone else. The forex figure is a month-end snapshot of what dealers owe retail clients, adjusted for realized and unrealized profit and loss. A monthly move can reflect client results as well as deposits and withdrawals. That means a simple 0.34% August decline could be a flat deposit month with a small trading loss layered on top. It could also be heavy withdrawals masked by a good trading month. We cannot separate the two from the public data alone, and any bot vendor claiming otherwise is overstating what the CFTC series can tell you.

Why does a $470 million deposit base matter to AI trading bot users?

Concentration is the first answer. Gain Capital holds 43.44% of all US retail forex deposits, and together with OANDA it accounts for 71.40% of the pool. A retail trader running an AI trading bot through a single US dealer is therefore exposed to that dealer's spread regime, its execution quality, and, indirectly, its balance sheet health as a counterparty. When two firms carry seven of every ten dollars, the diversification argument for splitting automated flow across brokers gets stronger, not weaker.

The second answer is flow. We stress-tested the concentration math against the $469.96 million total and it is hard to escape the conclusion that US retail forex is a small, top-heavy pond. For context, the futures brokers owned by Coinbase and Robinhood held $581.4 million of segregated client money in August, more than the six forex dealers combined, with Coinbase Financial Markets alone at $346.9 million, per an FM Intelligence analysis of the same CFTC filings. The two numbers measure different things, but the direction is clear: retail risk appetite is migrating toward leveraged crypto exposure, and the CFTC's October 5 notice on a retail crypto framework contemplates allowing that leverage only through futures commission merchants or banks they sponsor Finance Magnates.

Free Download: US Retail Forex $470M Deposit Stall: AI Bot Position-Sizing & Drawdown Template
This template helps AI forex bot users set stop-out levels, allocate capital across strategies, and cap exposure while US retail deposits stall near $470M.
Get Forex Bot Risk Template

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.

How accurate are bot backtests once real deposits are on the line?

Every AI trading bot we have tested publishes a backtest. Almost none publishes a clean reconciliation between that backtest and the live account that followed. We have run this category against Zephyr AI's adaptive engine and against several rival frameworks across a 6-month window each, and the pattern holds: a strategy that looks smooth on a daily-close backtest tends to look different when real fills, real spreads, and real funding behavior enter the picture.

The deposit data reinforces the point in an unexpected way. When industry balances shrink year over year, brokers often widen spreads or tighten the terms on which they extend automated access. A backtest built on 2025 spread assumptions may be running against a materially different cost base in the current environment. We flag this every review cycle, and it is the single most common reason a live account diverges from a vendor's published curve. If a provider cannot tell you which spread and slippage assumptions sit under its backtest, treat the curve as marketing until proven otherwise.

How big are the drawdowns these bots can put on a retail account?

This is where portfolio framing matters. A retail trader with a $5,000 account is not running the same risk as a fund. A drawdown that is trivial on a $250,000 book can wipe out half of a small account, and the bot's leverage settings usually default to aggressive.

The honest answer is that drawdown figures vary enormously by strategy class and by the volatility regime you happen to be trading through. Event risk around NFP, CPI prints, and FOMC decisions is where automated systems tend to show their true colors, because position sizing that looked conservative in a calm month can stack into correlated exposure in a volatile one. We rebuild every strategy's drawdown profile against those event windows before we will call it tested, and we recommend any retail trader demand the same. Where a provider does not publish drawdown data by volatility regime, write "verify with provider" in your own notes and assume the worst case is worse than the headline number.

One concrete comparison helps here. The industry itself just lived through a modest drawdown: total deposits fell from $506.58 million in August 2025 to $469.96 million in August 2026, a 7.23% decline, the widest annual shortfall since March Finance Magnates. That is a slow, orderly bleed at the industry level. A single leveraged bot on a small account can lose that much in a week.

How do subscription fees change the math?

Here is the under-discussed risk in this whole category, and it flows directly from a stalled deposit base. Most AI trading bots charge a recurring subscription, and that fee is fixed. The account it trades is not. When the industry funding pool flattens near $470 million and individual balances shrink, the fee becomes a larger share of the account and therefore a larger drag on returns.

We model this explicitly. A bot with a fixed monthly subscription needs a higher gross return just to break even on a smaller account, and if the strategy is not scaling its position size proportionally, the fee quietly eats the edge. Open-source alternatives such as NautilusTrader and Backtrader sidestep the subscription entirely, because you own the infrastructure, but you pay for that in engineering time and in the absence of a support desk when something breaks at 3 a.m. That trade-off is real and worth pricing honestly.

None of this is an argument against paying for a bot. It is an argument for doing the arithmetic before you subscribe, especially in a year when the deposit pool is flat and your account may not grow the way a vendor's chart assumes.

Do these bots work with US brokers and CFTC-registered dealers?

Broker and API compatibility is where a lot of retail frustration lives. The US retail forex market runs through a small set of CFTC-registered dealers, and not every bot integrates cleanly with every one of them. Gain Capital and OANDA together hold 71.40% of the deposit pool, so most integration guides focus there, but Schwab, tastyfx, Interactive Brokers, and Trading.com each run different platforms and different automated-access policies.

The practical takeaway: confirm API support for your specific registered dealer before you fund anything, and test the connection on a small balance first. Interactive Brokers is a useful case study in scale mismatch. Its retail forex book fell 4.16% in July to $26.19 million, then rose 1.53% to $26.59 million in August, still 20.28% below a year earlier and the steepest annual decline of the six, even as the wider firm reported client equity of $962.8 billion, up 35% year over year, with 5.46 million client accounts Finance Magnates. The forex book is a rounding error inside that firm, which tells you how much attention automated forex flow gets relative to the rest of its business.

When does a bot stop following its own strategy?

Strategy deviation is the silent failure mode. A bot's specification says one thing, and in live conditions it does another, usually because a hidden rule overrides the stated logic when volatility spikes or when a position moves against it. In our review work we track every deviation from stated strategy and log each one, because a single undisclosed rule change can invalidate the entire risk profile you signed up for.

The trigger points are predictable. News events, liquidity gaps at rollover, and broker-side margin changes all push bots off script. If a vendor cannot show you the exact conditions under which its system will deviate from spec, you are taking on rules you did not agree to. We treat deviation disclosure as a hard requirement, not a nice-to-have.

Can you actually turn the bot off and get your money back?

Disengagement is the test almost nobody runs until it is too late. A bot that is hard to switch off is a bot you cannot manage. Before committing real capital, we check three things: how quickly an open position can be flattened, how long a withdrawal takes to settle, and whether the subscription auto-renews after you cancel.

On a smaller, flatter funding base, this matters more. If the industry pool is stuck near $470 million and your dealer is one of the shrinking ones, you want clean exit mechanics, not a support ticket queue.

Who regulates AI trading bot providers?

This is the question with the widest gap between what retail traders assume and what is actually true. Most AI trading bots and algorithmic platforms are software products. Some are run by firms with advisory or brokerage licences, many are not. Regulators such as the FCA, ASIC, CySEC, and the NFA maintain public registers, and you should check them directly rather than relying on a vendor badge.

We do not assert a licence status we cannot verify against a primary register. If a provider claims FCA authorisation, look it up on the FCA Register. If it claims Australian registration, search the ASIC registers. The same discipline applies to any prop firm or funding partner attached to the bot. Where a claim cannot be traced to a register entry, treat it as unverified and say so in your own notes.

How Zephyr AI compares

Set against the frameworks we evaluate as subjects, the difference that stands out in our testing is risk disclosure rather than raw return claims. Open-source options like NautilusTrader and Backtrader give you total transparency into logic, which is genuinely valuable, but they leave risk management entirely in your hands. Vendor bots such as 3Commas and MetaTrader EAs often bury their risk parameters in a marketing page. Where Zephyr AI's risk layer has separated itself in our framework is in stating drawdown controls up front and adapting position sizing to the volatility regime rather than running a fixed lot regardless of conditions. In a year when the US retail forex deposit pool shrank 7.23% year over year, that adaptability is a portfolio-level advantage, not a marketing line. Confirm the current figures and supported brokers directly with the provider before you fund.

The bottom line for bot traders

A flat deposit market near $470 million is not a crisis, but it is a signal. Retail forex money in the US is small, concentrated, and, for now, stagnant, while leveraged crypto exposure through futures brokers is drawing more client capital. For anyone running an AI trading bot, that means fee drag matters more, counterparty concentration matters more, and clean exit mechanics matter more than they did in a growing market. Test the strategy, verify the regulation, and do the fee arithmetic on your actual account size, not the vendor's illustration.

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

Does an AI trading bot work in the US under Pattern Day Trader rules?

Pattern Day Trader rules apply to margin equity accounts in US equities, not to spot forex, which is regulated differently. A US retail forex account with a CFTC-registered dealer is not subject to the same PDT framework, but your broker may still impose its own minimum balance and automated-trading terms, so confirm directly with the dealer before funding a bot.

Can I run an AI trading bot on a prop firm account?

Some prop and funding firms permit automation and some explicitly prohibit it, and the rules change often. The FTMO-owned OANDA is a reminder that funding-firm ownership and broker operations can sit in the same group, so check both the prop firm's terms and the broker's automated-access policy separately before you connect a bot.

What happens if the API connection drops mid-trade?

Behavior depends entirely on the vendor and the broker. A well-built bot should either close the position or re-establish the connection with a defined fallback, but many simply stop managing the trade and leave it open. Ask the provider in writing what its reconnect logic does before you fund, and test it on a small balance first.

Which US brokers are compatible with AI trading bots?

Compatibility varies by platform. Gain Capital and OANDA hold 71.40% of the US retail forex deposit pool and appear in most integration guides, while Schwab, tastyfx, Interactive Brokers, and Trading.com each run different systems and automated-access policies. Verify API support for your specific dealer before subscribing to any bot.

How much should I budget for a bot subscription?

There is no single figure, and pricing changes frequently. The important calculation is fee drag as a percentage of your account balance, because a fixed subscription eats a larger share of a smaller account. Model your expected gross return against the stated fee before committing, and confirm current plan pricing directly with the provider.

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

Check the primary register. Claims of FCA authorisation should be searched on the FCA Register, Australian claims on the ASIC registers, and US brokerage claims on the relevant regulator's database. If a claim cannot be traced to a register entry, treat it as unverified.

Are the backtests these bots publish reliable?

Treat published backtests with measured skepticism by default. Most do not disclose the spread, slippage, and funding assumptions underneath the curve, and those assumptions can shift when market conditions change. A backtest is a starting point for due diligence, not evidence of future performance.

What is the biggest risk for bot traders in a shrinking deposit market?

Fee drag and counterparty concentration. When the US retail forex pool is flat near $470 million and two dealers hold 71.40% of it, a bot's fixed subscription takes a larger percentage of a smaller account, and routing all automated flow through one dealer concentrates execution and counterparty risk.

Can I withdraw my funds and stop the bot cleanly?

That depends on the vendor's cancellation and the broker's withdrawal process. Before committing capital, test how quickly an open position can be flattened, how long a withdrawal takes to settle, and whether the subscription auto-renews after cancellation. A bot you cannot switch off cleanly is a bot you cannot manage.

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: AI Trading Bot Reviews.