AI vs ASX Runners: Spot Hype, Catch Real Signals

A 39,650% peak gain isn’t “a good year.” It’s a stress test for every part of your investment process: your data, your rules, your risk limits, and your ability to tell signal from structure-driven chaos.

That’s the most useful lesson from the ASX “Runners of the Year” list. Yes, 2025 rewarded gold producers and US-address critical minerals plays. But it also produced a headline-making outlier: a microcap that went vertical in roughly 48 hours on a thin float, concentrated ownership, and aggressive trading dynamics.

This post is part of our AI in Finance and FinTech series, and I’m going to take a stance: most investors don’t need better stock tips—they need better detection systems. AI-driven analytics can help you spot when a move is fundamentally supported (cash flows, commodity prices, permits, resource updates) versus when it’s mostly a market microstructure event (liquidity squeeze, order-book vacuum, bot-driven momentum).

What 2025’s ASX “runners” really tell us

Answer first: 2025’s biggest ASX price runs were driven by a mix of macro tailwinds (gold), geopolitics (critical minerals supply chains), and market structure (tiny floats + concentrated registers)—and that mix is exactly where AI models can outperform gut feel.

The source list highlights a few clear patterns:

• Gold was a dominant theme. Gold ran from below A$4,000/oz to above A$6,700/oz at one point (a rise of more than 65% by October), then hovered around the mid A$6,000s/oz. When the commodity moves that far, producers and near-producers tend to move harder.
• Critical minerals got repriced through a geopolitical lens. US focus on non-Chinese supply chains flowed into ASX narratives—especially projects with US assets or “right place, right time” adjacency.
• Biotech took a beating. Several ASX biotechs were described as being cut by half or more, and even established names saw large drawdowns.
• The most extreme runners weren’t “discovered,” they were “manufactured” by liquidity conditions. That’s not an insult—it's just how markets behave when scarcity meets demand.

The practical question for investors and financial professionals is simple: how do you build a repeatable method to classify these regimes early?

The microcap reality: when price is a liquidity event

Answer first: Some ASX surges are primarily liquidity events, not valuation events—and AI can help you detect the difference by tracking float, order-book fragility, concentration, and abnormal volume/price patterns.

The poster child here was Kaili Resources (KLR), which reportedly peaked up 39,650% (from 0.8c to $3.18) with much of the spike compressed into a short trading window. The narrative includes classic ingredients:

• Extremely tight register (a high share concentration among top holders)
• Very low starting liquidity (small dollar value traded early on)
• A catalyst that, on its own, didn’t “explain” the valuation implied by the peak price

This is the sort of event where traditional fundamentals analysis often arrives late, because the event is not waiting for your discounted cash flow model.

What AI should be monitoring in real time

If you’re building (or buying) fintech tooling for market monitoring, these are high-value features that are measurable and predictive of “blow-off” dynamics:

1. Free-float estimates and concentration proxies
   • Top-20 holder concentration (where available)
   • Estimated free float vs. market cap
2. Liquidity stress indicators
   • Dollar volume vs. market cap (and sudden step-changes)
   • Bid-ask spread expansion
   • Order-book depth thinning across levels
3. Abnormal momentum signatures
   • Gap frequency (open-to-close discontinuities)
   • Multi-sigma returns relative to the stock’s own history
4. Halt and volatility regime shifts
   • Trading halts, rapid reversals, and intraday drawdowns

A useful one-liner I’ve found for teams building alerts:

If the move requires “perfect liquidity” to exit, it isn’t a move—it’s a trap.

Gold winners: where AI helps without pretending to predict the future

Answer first: In commodity-led rallies, AI adds the most value by mapping exposure (who benefits, how much, and when) and by enforcing risk controls, not by claiming it can forecast the gold price.

Gold’s run mattered because it was broad enough to lift:

• Major producers (the source mentions 150%+ gains for several)
• Developers and restart stories where operating leverage kicks in
• Explorers that gain attention when the tape is strong

In a gold-led year, many investors make a basic mistake: they treat “gold exposure” as a single bucket. It isn’t.

A simple AI-assisted “gold exposure map”

A practical approach used in institutional research workflows is to classify companies into exposure tiers and let a model keep the classification updated:

• Tier 1: Producers (revenue sensitivity to spot price)
• Tier 2: Near-term developers (financing + construction risk dominates)
• Tier 3: Explorers (discovery probability + sentiment dominates)

From there, machine learning can help with:

• Earnings sensitivity estimation (how margins change as spot changes)
• News classification (permits, grade, metallurgy, capex revisions)
• Portfolio-level risk (crowding, factor exposure, drawdown control)

The goal isn’t to “beat the market with AI.” It’s to avoid the very human habit of letting a strong chart turn into an oversized position.

Critical minerals and the “US address premium”: separate narrative from economics

Answer first: The 2025 critical minerals theme rewarded projects aligned with US supply-chain priorities, but AI is needed to keep you honest about what’s narrative versus what’s economically bankable.

The list highlighted strong runs in names tied to rare earths, antimony, scandium, and other strategic materials—often with US-based assets or adjacency to known deposits.

Two things can be true at once:

• Policy shifts can create real capital flows.
• Markets can overpay for proximity, symbolism, or “strategic” headlines.

Where AI-driven due diligence shines

For advisers, analysts, and fintech teams supporting clients, AI is valuable for converting messy public info into structured, comparable fields:

• Project stage (concept → drilling → resource → feasibility → financing)
• Jurisdiction and permitting milestones
• Capex/opex ranges and revision frequency
• Commodity-specific pricing realism (spot vs. contract dynamics)

This is also where risk assessment becomes a product feature, not a PDF at the end:

• A model can flag when a company’s valuation implies unrealistic timelines.
• A model can compare peer-stage valuations and highlight outliers.

If you work in wealth or brokerage, this is exactly the kind of “assistive intelligence” clients will pay for in 2026: not predictions, but fewer blind spots.

AI for retail investors and advisers: a practical workflow

Answer first: The best AI in investing is a workflow that combines screening, alerts, and position rules—so you’re not making the biggest decisions at the most emotional moments.

Here’s a workflow that fits how real people operate, especially around year-end when portfolios get tidied up and “hot lists” circulate.

Step 1: Build an “explainable” runner screen

Use a screener that answers one question: why is this stock moving?

Minimum fields:

• Price change over 1D/1W/1M
• Volume vs. 30-day average
• Market cap and estimated float
• Catalyst tag (earnings, permit, resource update, takeover, macro)

Step 2: Add a “hype risk score”

Create a composite score that increases when:

• Price acceleration outpaces liquidity growth
• Concentration is high
• Order-book depth is fragile
• Social/news velocity spikes without matching fundamentals

This is where many fintech platforms can differentiate: make the risk visible when excitement is highest.

Step 3: Codify exit rules before entry

I’m opinionated here: if you can’t explain your exit in one sentence, you don’t have one.

Examples of rules that are easy to automate:

• Hard stop-loss and trailing stop
• Max position size (especially for microcaps)
• “Two-day rule” for parabolic moves (reduce exposure after extreme gaps)

Step 4: Use AI for post-trade learning

Most investors never audit their decisions. AI can.

• Did you buy after a volatility regime shift?
• Did you average down into structural illiquidity?
• Did you ignore insider selling signals?

Even a simple dashboard that shows “what you tend to do wrong” pays for itself.

What financial institutions should learn from these runners

Answer first: Extreme ASX runners are a product problem for financial services—clients need guardrails, not just access—and AI is the only scalable way to deliver that responsibly.

Banks, brokers, and wealth platforms already use AI in finance for fraud detection and operational efficiency. The next wave is client-facing:

• Personalized risk warnings based on trading behavior
• Suitability checks that adapt to volatility regimes
• Portfolio stress testing that includes liquidity and concentration risk

This also ties into adjacent fintech areas like AI credit scoring and broader risk models: the best systems don’t just predict default; they predict behavior under stress. And speculative trading frenzies are behavior under stress.

The biggest gap in most investing apps isn’t research. It’s risk context.

A better way to approach the 2026 “runner hunt”

The temptation after a list like this is to chase the next 40-bagger. That’s usually the wrong goal. A better goal is to build a process that can survive both outcomes: a genuine multi-year compounding story and a two-day liquidity bonfire.

If you’re an investor, start with a simple upgrade: use AI tools to classify moves (fundamental, macro, structure) and force position sizing discipline in microcaps.

If you’re a financial professional or fintech builder, the opportunity is bigger: package these capabilities into your platform—alerts, explainability, and guardrails—so clients can participate in markets without getting chewed up by them.

Where do you want your edge in 2026: in finding the next runner, or in being the person who can tell which runner is real before the crowd figures it out?