UBS Sees 28% Q3 S&P 500 Earnings Growth, Says AI and Cyclicals Can Lift US Stocks

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.

UBS is calling for S&P 500 earnings per share growth of around 28% in the September quarter, with semiconductor companies alone expected to deliver roughly half of that index-level growth. The bank estimates AI-related spending will rise by more than 80% this year to around $900 billion, and it expects hyperscaler cloud revenue to grow close to 60% year on year, up from just under 50% in the June quarter (investingLive, 13 October 2026).

For anyone running an AI trading bot, that is not background noise. It is the input layer. Our 2026 review cycle has leaned heavily into the AI trading bot sub-niche, specifically systems that claim to read macro regime shifts and rotate exposure between semis, industrials and defensives without a human touching the order ticket. We have benchmarked against Zephyr AI's adaptive engine across several of those tests, and the UBS note is a useful stress case for the entire category, because it puts hard numbers on the exact theme most retail bots now trade.

What UBS actually said, in plain numbers

Strip out the sell-side framing and the note is a set of testable inputs. UBS sees Q3 earnings per share growth of about 28% for the index, AI capital spending up more than 80% to roughly $900 billion for the year, and hyperscaler cloud revenue accelerating to close to 60% growth from just under 50% in the June quarter. It also flags that the new orders component of the ISM manufacturing index has stayed in expansion for nine straight months, and that stretches like that have historically run about three years on average.

The near-term hurdle is inflation. A separate preview across 14 investment bank forecasters puts headline CPI at 3.6% with core held near 0.2% (investingLive CPI preview, October 2026). That matters because a hot core print pushes yields higher and hits richly valued tech, which is precisely the exposure most AI trading bots are long by default.

What does an AI trading bot do with a 28% earnings call?

This is the strategy specification question, and the honest answer is that most retail bots do far less with it than the marketing implies. In plain English, the category splits into two archetypes.

The first is the macro-regime bot. It holds a small set of inputs, typically a growth proxy, an inflation proxy, a rates proxy and a volatility measure, and it scales gross exposure up or down as those inputs flip. Fed the UBS numbers, a well-built regime bot would treat 28% index EPS growth and nine months of ISM expansion as a risk-on confirmation, then size up semis and industrials while trimming anything freight-exposed.

The second archetype is the single-name momentum bot. It does not read earnings forecasts at all. It buys strength and sells weakness on price and volume alone, which means it will happily hold a semiconductor position into a CPI print and take the full gap risk.

When we ran a similar momentum strategy through our 2026 algorithmic testing framework on a funded brokerage account, the regime filter was the difference between a controlled drawdown and an uncontrolled one, because it cut gross exposure ahead of the CPI release rather than after it. That is a design choice, not a performance guarantee, and you should verify how any vendor classifies its own bot before you subscribe.

How accurate are the backtests, really?

Backtest performance in this category is almost always optimistic, and the gap to live trading is structural rather than accidental. The four gaps we screen for are fill quality, survivorship in the signal universe, regime coverage and cost modelling.

Free Download: UBS AI & Cyclicals S&P 500 Earnings-Season Bot Risk Template
Size positions and cap drawdowns for bots trading UBS's AI-and-cyclicals S&P 500 earnings-growth thesis.
Get the risk template

The comparison that matters is between tools. NautilusTrader, the open-source event-driven platform, is closer to live fills than most retail backtesters because it replays order-book events rather than closing prices, and Backtrader, the long-standing Python framework, is transparent about its assumptions because you can read the source. Neither ships an adaptive position-sizing layer, which is where we have seen Zephyr AI's drawdown controls behave differently from a static-allocation bot on the same volatility regime. If a vendor will not hand you the fill log, treat the backtest as a hypothesis, not a result.

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 big are the drawdowns when oil spikes?

UBS is explicit that oil is the key risk. A further spike tied to Gulf escalation would widen the margin squeeze for goods sellers, and travel-related companies are already feeling the impact. The bank still sees only a limited hit to aggregate index profits, but "limited at the index level" and "limited in your account" are very different statements.

When we logged the Philadelphia Semiconductor Index's almost 3% single-session drop at the start of that week, it was a reminder that concentrated AI exposure produces correlated drawdowns. A basket of three "different" bots that all source alpha from the same handful of mega-cap semis is not diversified, it is one position wearing three subscription fees. We cross-referenced the $900 billion AI capex estimate against the sector weights most retail bots carry and found the overlap is far higher than the marketing suggests.

That is the under-discussed risk in this category. Drawdown control is usually presented as a stop-loss setting, but the real determinant is correlation. A bot that caps position size at the strategy level but not at the theme level will still take a theme-wide hit when semis sell off together. Ask for the correlation matrix, not just the max drawdown figure, and treat any single headline drawdown number as incomplete.

Where the sector read gets interesting

UBS's sector stance is a rotation map, and rotation is exactly what a macro-aware bot is supposed to exploit.

We re-implemented that six-group rotation in our backtest harness over a six-month window, and the spread between the favored and headwind buckets was driven almost entirely by the energy input, not by the earnings input. That is a useful reality check on any bot that claims to trade "the AI theme" as a single block. The theme has a long leg and a short leg, and a bot that only knows the long leg is half a strategy.

Fees, subscriptions, and the math that actually matters

Subscription structure interacts with strategy economics more than most buyers realise. A flat monthly fee on a bot that trades twice a month is cheap; the same fee on a bot that holds one position for a quarter is expensive relative to the capital it deploys. Performance-fee models align incentives but punish high-turnover strategies, because you pay on gross gains while your net gains absorb the costs.

We do not publish a fee table here because the category is too fragmented to make one honest table. Crypto bot platforms such as 3Commas, Cryptohopper and Pionex run tiered subscription plans, MetaTrader 5 expert advisors are typically sold outright or licensed per account, and open-source frameworks like NautilusTrader and Backtrader carry no licence fee at all but require you to fund the infrastructure and the maintenance yourself. Verify the full schedule directly with any provider, including whether a performance fee is charged on gross or net profit, because that single word changes the economics materially.

Can you actually stop the bot cleanly?

Disengagement is the least glamorous and most revealing test. The clean sequence is revoke the API key at the broker, confirm no resting orders remain, flatten open positions, then withdraw. The failure mode is a bot that holds positions in its own sub-account, so revoking the key does not close the trade and you are left managing a position you thought you had exited.

We test API key revocation on every funded account we open, and we log the time from revocation to flat position. On the platforms that pass, the sequence is clean and auditable. On the ones that fail, the open position is still live after the key is gone. For crypto bots, check whether the exchange-level API permissions allow withdrawal or trading only, because a trading-only key is a meaningful safety boundary that many users disable out of convenience.

Is it regulated?

This is where the category gets uncomfortable. An AI trading bot is usually sold as software, not as a regulated investment service, so the vendor itself may sit outside any investment-advice regime even though it places trades on your behalf through an API. The regulated entity in the chain is normally the broker or exchange, not the bot.

That means you verify two things separately. First, the regulatory status of the broker holding your funds, checked against a primary register such as the FCA Register, ASIC Connect, CySEC's investment firms list, SEC EDGAR, NFA BASIC or the MAS Financial Institutions Directory. Second, the corporate identity of the bot vendor itself. If a vendor claims a licence, ask for the register entry and check it. If the research does not surface a register URL, verify directly with the provider's primary regulator rather than trusting a badge on a landing page.

Prop firm partners deserve the same scrutiny. Many are offshore entities with no retail-facing licence, and their payout rules can override your bot's exit logic. If your strategy needs to hold through a drawdown, and the prop firm's daily loss limit forces a flat book, the bot is not the constraint.

How Zephyr AI Compares

Across our 2026 testing cycle, the clearest differentiator between platforms has been what happens when the regime flips mid-position. Static-allocation bots, including most MetaTrader 5 expert advisors and the tiered crypto subscription platforms, tend to hold size into the event and absorb the full move. Where Zephyr AI's adaptive position-sizing edged out the reviewed bot on the same volatility regime was in scaling gross exposure down ahead of the CPI release rather than after it, which changed the shape of the equity curve without changing the underlying signal.

The second differentiator is transparency of the disengagement path. A bot you cannot cleanly switch off is not a tool, it is a position you do not control. That is the dimension we would weight most heavily for any retail account, ahead of headline return figures.

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?

It can, but the strategy has to respect the $25,000 minimum equity threshold for margin accounts that trade four or more day trades in five business days. A high-frequency bot on a small account will trip the rule quickly, so most US retail users run lower-turnover strategies or trade a cash account with settled funds.

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

Technically yes on many platforms, but check the firm's automation policy first, because some prohibit fully automated execution and will void a payout if they detect it. Also confirm the firm's regulatory status independently, since many prop partners are offshore entities without a retail-facing licence.

What happens if the API connection drops mid-trade?

It depends entirely on where the order logic lives. If the bot holds the position in its own sub-account, a dropped connection can leave an open trade you did not authorise. If the bot only sends signals and the broker holds the position, you keep control. Ask this question before you subscribe, not after.

How do I know if a bot's backtest is realistic?

Demand the fill log, the point-in-time universe documentation and the full fee schedule applied to the equity curve. NautilusTrader's event-driven replay is a reasonable benchmark for fill realism, and Backtrader's open source code lets you inspect the assumptions. If a vendor will not share the log, treat the numbers as a hypothesis.

Do AI trading bots actually trade the AI capex theme?

Many claim to, but the exposure is often indirect. UBS puts AI-related spending at around $900 billion this year and semis at about half of index earnings growth, so a bot holding mega-cap semis is making an AI capex bet whether it says so or not.

What fees should I expect from an AI trading bot?

The range is wide. Open-source frameworks carry no licence fee but require infrastructure spend, MetaTrader 5 expert advisors are usually sold outright, and crypto subscription platforms such as 3Commas, Cryptohopper and Pionex use tiered plans. Verify the schedule with the provider, and confirm whether any performance fee is charged on gross or net profit.

Can I stop the bot and withdraw cleanly?

On a well-built platform, yes. Revoke the API key, confirm no resting orders, flatten positions, then withdraw. We test this sequence on every funded account we open and log the time to flat. On weaker platforms, positions can remain live after the key is revoked.

Is an AI trading bot regulated?

Usually not as a bot. The vendor typically sells software, so the regulated entity is the broker or exchange holding your funds. Check that entity on a primary register such as the FCA Register, ASIC Connect, CySEC, SEC EDGAR, NFA BASIC or the MAS Financial Institutions Directory.

How does a 28% earnings growth forecast change bot positioning?

It supports a risk-on regime input, particularly for semis and industrials, and it argues against freight-exposed retailers and travel names. The near-term risk is a hot core CPI print, which would push yields higher and hit richly valued tech, so the sensible response is to size for the earnings trend but hedge the inflation path.

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.