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.

The Retail CFD Industry Is Running Out of FX Traders

Currency pairs made up just 13.7% of the volume retail brokers reported in Q2 2026, down from 26.8% a year earlier, according to FM Intelligence data published by FinanceMagnates.com. Monthly FX volume across the tracked brokers fell to $4.2 trillion from $6.4 trillion, while index, commodity, equity and crypto CFD volume rose to $26.3 trillion from $17.4 trillion over the same period (FinanceMagnates.com, 2026). For anyone running an AI trading bot or algorithmic trading platform across a retail CFD book, that is not a headline about broker marketing. It is a structural shift in what the underlying instruments actually do, and it changes the economics of every strategy that was calibrated on EUR/USD mean reversion between 2019 and 2023.

We have spent our 2026 review cycle re-running strategy classes that were designed for FX liquidity conditions against the instrument mix brokers are now reporting. In our 2026 review cycle, we benchmarked a multi-strategy automation stack against the Ellington AI trading platform to see how a portfolio-level risk engine handled the same rotation. The gap between what most retail bots are specified to trade and what the volume data says retail traders are actually trading is now wide enough that it deserves its own review.

What does the volume data actually say?

The numbers are unambiguous, and they are worse for FX specialists than the headline share decline implies.

Source: FM Intelligence, via FinanceMagnates.com.

The five largest brokers — EC Markets, TMGM, IC Markets, IG Group and JustMarkets — now hold 41.8% of the volume FM Intelligence attributes to named firms. That concentration is not new; the same analysis showed the top-five share barely moving between the end of 2021 and the end of 2025, at 38.4% and 38.2% respectively. What is new is that the share has risen in each of the three quarters since, on the same set of 51 to 52 named brokers.

Three of those five route all of their reported volume through MetaTrader. That matters for our readers because MT4 and MT5 expert advisors are the most common retail automation format, and the platforms that dominate volume are the platforms where EA-based FX strategies are most easily deployed. If the volume is leaving FX, the EA strategies calibrated for FX are being run against a shrinking share of the book.

Why the top-five share rose while everyone else shrank

The arithmetic here is worth walking through, because it is easy to misread the concentration story as growth.

FM Intelligence's July ranking put the industry's decline at 9.3% between Q1 and Q2 2026. The median named broker lost 12.6%, and 47 of the 51 brokers present in both quarters reported lower monthly volume. Two firms explain the gap between the 9.3% industry figure and the 12.6% median: EC Markets added 23.6% and TMGM held its volume, while the other 49 named brokers fell 12.3% together. Within the top five, IC Markets, IG Group and JustMarkets each lost between 11% and 13%.

So the top-five share did not rise because the top five grew. It rose because the rest of the field shrank faster. That is a materially different signal for anyone evaluating a broker's liquidity for algorithmic execution. A shrinking median broker is a broker whose order flow is thinning, and thinner order flow in a market-making or hybrid model has a direct effect on the fill quality your bot actually receives versus what your backtest assumed.

The August study of volume per active account found trading activity per account falling at 45 of 51 brokers in the same quarter. A broker that reported a double-digit decline in Q2 sat in the middle of its peer group on both measures. In plain English: this is not one or two brokers having a bad quarter. It is the median retail CFD account trading less, and trading less in FX.

What does the bot actually trade?

This is where the FX decline stops being a broker story and becomes a strategy specification problem.

Most retail AI trading bots and algorithmic platforms we evaluate in our 2026 algorithmic testing program are specified around a narrow instrument universe. A typical FX-focused bot might list eight to twelve major and minor pairs, with parameters tuned to the volatility regime of 2021-2023. When we ran a similar mean-reversion specification through our live-trading evaluation framework on a funded brokerage account across our most recent six-month window, the strategy's opportunity set had visibly contracted relative to its own backtest assumptions — the instrument mix it was designed for no longer represented the volume share it was calibrated against.

Arkadiusz Jóźwiak, Editor-in-Chief at Comparic.pl, told FinanceMagnates.com that "the moves that pull retail money in over the past year have been in metals, equities and crypto, not in currency pairs." That is the trade itself changing, not just the marketing. A bot that only trades FX is a bot that only trades 13.7% of the reported retail book.

For contrast, the Ellington AI trading platform we benchmarked in the same cycle runs multi-asset coverage by default, which means the same portfolio-level risk engine can allocate across the instruments where volume actually moved — indices, metals, crypto CFDs — without a separate deployment. That is the concrete dimension where multi-strategy automation beats a single-asset-class bot in this regime.

Is the FX decline real, or an artifact of the data?

This is the most important caveat in the source material, and any honest review has to lead with it.

FM Intelligence estimates the FX split for many smaller brokers, and it changed those estimates in the fourth quarter of 2025. Eighteen brokers show their published FX share stepping from 65-70% to 27-30% in a single quarter. That is a methodological break, not a trading event.

The decline holds when those firms are removed. Among the 32 brokers whose shares moved gradually, the FX share fell to 11.7% from 20.5%, and 30 of the 32 reported a lower share. IC Markets went from 19% to 10% and Saxo Bank from 22% to 16%. Those are gradual moves, not reclassifications, and they point the same direction.

The FX figures come from published quarterly tables that mix broker submissions with FM Intelligence estimates. FM Intelligence expanded its coverage to 265 brokers in June 2026, with 20% of the Q1 2026 figures classed as verified and 78% as estimated (FinanceMagnates.com, 2026).

For our purposes, the honest read is this: the direction of the FX decline is robust across both the full sample and the cleaned sub-sample, but the precise magnitude of the share shift is estimate-dependent. We would not build a bot's asset allocation on the 13.7% figure alone. We would build it on the fact that 30 of 32 gradually-moving brokers reported a lower FX share, which is a 93.75% hit rate on direction.

Backtest versus live: the gap that keeps widening

Here is the under-discussed risk in this data set, and it is the one most bot reviews miss entirely.

When a bot's backtest is calibrated on historical FX volume and volatility, and the live instrument mix has rotated toward indices, metals and crypto CFDs, the backtest is not just stale — it is measuring a market that no longer exists at the same weight. The gap between backtest and live performance is not only about slippage, latency, or overfitting. It is about the opportunity set itself shrinking underneath the strategy.

We flag this as a specification mismatch rather than a performance failure, because the two require different fixes. A performance failure is a parameter problem. A specification mismatch is an instrument-universe problem, and no amount of parameter tuning fixes a strategy that is fishing in a pond that is being drained.

Free Download: Retail FX/CFD Broker Due-Diligence Checklist for AI Bot Deployment
A step-by-step checklist to verify a CFD broker's regulatory status, fee transparency, withdrawal flow, and API compatibility before connecting your AI trading bot in a shrinking FX-trader market.
Vet Your Broker Now

Backtest and live performance figures vary by strategy parameters and provider — verify directly with each bot provider before drawing conclusions. The table above reflects specification differences, not measured returns.

Not sure which AI trading bot fits your strategy? Try Ellington — The AI Trading Platform for 2026

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

How do fees interact with a shrinking opportunity set?

A subscription fee is a fixed cost against a variable opportunity set, and that ratio has gotten worse for FX-only bots.

If a bot charges a flat monthly subscription and its addressable market has contracted from roughly a quarter of retail CFD volume to under 14%, the fee-to-opportunity ratio has roughly doubled without the vendor changing its price. We have not seen a single FX-focused bot provider adjust its pricing to reflect that. In our 2026 review cycle we logged this as a structural cost problem rather than a vendor failure, because the provider is not doing anything wrong — the market moved.

This is where fee transparency becomes a genuine differentiator. A platform that charges on a flat, published schedule across a multi-asset universe gives the retail trader a stable cost base against a broader opportunity set. A platform that charges per-strategy or per-instrument on a shrinking FX book gives the trader a rising effective cost per unit of opportunity. Neither model is inherently wrong, but the second one is now structurally disadvantaged, and most comparison tables still rank bots on headline subscription price rather than fee-to-opportunity ratio.

Which brokers and platforms are affected most?

The concentration data points to a specific answer.

Three of the top five brokers route all of their reported volume through MetaTrader. That means the MT4/MT5 expert advisor ecosystem is disproportionately exposed to the FX decline, because the platforms carrying the most volume are the platforms where EA-based FX strategies are most naturally deployed. An expert advisor written for a 12-pair FX universe is running on a book where FX is 4.0% of the top five's combined volume, down from 10.9% a year earlier.

For broker compatibility and API integration, the practical question a retail trader should ask is not "does this bot connect to my broker" but "does this bot's instrument universe match where my broker's volume actually is." A bot with a clean API integration to a MetaTrader-only broker is well-connected to a shrinking segment.

On the regulatory side, we do not assert license status for any provider in this article. If you are evaluating a bot provider or a prop funding partner, verify registration directly with the primary regulator — the FCA Register for UK entities, the ASIC registers for Australian entities, or the equivalent CySEC, ESMA, NFA BASIC or MAS directory for other jurisdictions. We never assert a license number we cannot cite to a primary register, and neither should any bot review you read.

How Ellington compares

The reviewed category here is the single-asset-class retail bot — the FX-focused AI signal provider or expert advisor whose specification was written for a market that is now a minority of reported volume. On the concrete dimension of instrument-universe coverage, the contrast is direct: an FX-only specification addresses roughly 13.7% of reported retail CFD volume, while the Ellington platform we benchmarked in our 2026 review cycle runs multi-asset coverage through a single portfolio-level risk engine, which means the same automation addresses the segment that grew 51.1% year over year rather than the segment that fell 34.4%.

That is not a claim about returns. It is a claim about specification fit, and specification fit is the first thing we test and the last thing most reviews mention.

Try Ellington — The AI Trading Platform for 2026

Try Ellington — The AI Trading Platform 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 a multi-asset AI trading bot work in the US under Pattern Day Trader rules?

Pattern Day Trader rules apply to margin equity accounts trading US-listed securities and are enforced at the broker level, not the bot level. A bot that trades CFDs, spot FX or crypto may fall outside the PDT framework entirely depending on the instrument and the venue. Confirm with your broker which instruments trigger PDT counting before deploying any automated strategy on a US account.

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

It depends entirely on the prop firm's terms, and most prohibit fully automated execution or require manual oversight. Verify the funding partner's rules directly — we do not assert prop-firm permissions without a primary source. If automation is permitted, expect instrument-universe restrictions that may exclude the multi-asset coverage that makes a bot useful in the current volume regime.

What happens if the API connection drops mid-trade?

This is the single most common failure mode we log in live testing. Behavior depends on the bot's order-handling logic: some platforms leave resting orders at the broker, others cancel on disconnect, and others hold positions open with no stop management. Ask the provider explicitly what happens to open positions on connection loss, and test it on a demo account before funding.

Is the FX decline a reason to stop trading currency pairs entirely?

No. It is a reason to stop assuming FX is the default retail instrument. FX still represents 13.7% of reported volume and $4.2 trillion in monthly turnover across tracked brokers. The point is allocation weight, not elimination.

How accurate are FX-focused bot backtests given this data?

Treat them with skepticism by default. A backtest calibrated on 2021-2023 FX volume and volatility is measuring a market with roughly double the FX volume share of the current one. Backtest data should be verified directly with the bot provider, and you should ask specifically what instrument universe and date range the backtest covers.

Do subscription fees make sense for a single-asset-class bot?

Increasingly, no — at least not at flat pricing. A fixed monthly fee against a contracting opportunity set raises the effective cost per unit of opportunity without the vendor changing anything. Compare fee-to-opportunity ratio, not headline price, when evaluating any AI trading bot in 2026.

Which brokers are most exposed to the FX volume decline?

The five largest — EC Markets, TMGM, IC Markets, IG Group and JustMarkets — hold 41.8% of named volume and generate just 4.0% of it in currencies, down from 10.9% a year earlier. Three of the five route all reported volume through MetaTrader, which concentrates EA-based FX exposure.

Can I stop an AI trading bot cleanly?

Withdrawal and disengagement experience varies widely. Some platforms close all positions and revoke API keys immediately; others leave open trades requiring manual closure. Test the disengagement path on a small funded account before scaling, and confirm in writing whether the provider retains any order-routing access after cancellation.

Is the FM Intelligence FX share figure reliable?

Directionally yes, precisely no. The full-sample figure is 13.7%, but 18 brokers had their FX share estimates reclassified in Q4 2025. Among the 32 brokers whose shares moved gradually, the FX share fell to 11.7% from 20.5%, with 30 of 32 reporting a lower share. Use the direction, not the decimal.

The takeaway for retail bot buyers

The retail CFD industry is not running out of FX traders because FX stopped working. It is running out of FX traders because the moves moved — into metals, equities and crypto — and the median broker's book followed. A bot specification that does not follow is a specification that is now fishing in a smaller pond with the same fixed costs.

The practical checklist for our readers: confirm the bot's instrument universe, confirm whether that universe matches where your broker's volume actually is, confirm the fee-to-opportunity ratio rather than the headline subscription price, and confirm the disengagement path before you fund. And verify every regulatory claim against the primary register — FCA, ASIC, CySEC, ESMA, NFA BASIC or MAS — because a bot review that asserts license status without a register link is not a review.

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.