AI selloff: Screening for quality tech stocks after the reset

Key points:

• The AI selloff has raised the bar for stock selection. The opportunity is not simply in buying what fell the most, but in finding companies where the price has reset while business quality, AI monetisation and valuation support remain intact.
• The screen favours quality-growth tech with real AI relevance. It filters for strong revenue and EPS growth, healthy margins, positive free cash flow, low leverage, valuation discipline and a meaningful pullback from highs.
• Micron, Nvidia and Microsoft rank highest in the model. Micron benefits from the HBM bottleneck, Nvidia remains the clearest AI revenue engine, and Microsoft offers diversified AI exposure through cloud, Copilot and enterprise software.

The AI selloff has made stock selection more important. The opportunity is not simply in buying what fell the most, but in identifying companies where the share price has corrected while business quality, AI monetisation and valuation support remain intact.

The recent selloff in AI-linked stocks has raised a familiar question for investors: is this a buying opportunity, or a warning sign?

The answer is unlikely to be the same for every stock.

A broad “buy the dip” approach can be risky when valuations are still elevated, earnings expectations are high, and many investors are crowded into the same AI winners. The better approach is to screen for companies where the price has reset, but the business case has not broken.

This screener is designed to identify US-listed technology stocks with strong growth, healthy margins, positive free cash flow, manageable leverage and a meaningful pullback from recent highs.

Screening criteria used

The first step is to create a quality-growth universe. The screen uses the following criteria:

• US-listed technology stocks: Keeps the focus on the part of the market most exposed to the AI-driven selloff.
• Market cap above USD 10 billion: Filters for larger, more liquid companies and avoids highly speculative smaller names.
• 5-year revenue CAGR above 10%: Looks for companies with a track record of sustained top-line growth.
• 5-year diluted EPS CAGR above 12%: Ensures growth has translated into earnings, not just sales.
• Next-year sales growth above 15%: Helps identify companies where forward growth momentum remains visible.
• Forward gross margin above 40%: Screens for higher-quality technology businesses with pricing power, differentiated products or scalable models.
• Net debt/EBITDA below 1x: Avoids companies with high balance-sheet risk, which matters more when volatility rises.
• Positive trailing 12-month free cash flow: Important because investors are increasingly questioning whether AI investment can be funded sustainably.
• Share price below 90% of its 52-week high: Ensures the stock has actually corrected and is not still sitting near peak optimism.
• Forward P/E below 55x: Adds valuation discipline and avoids names still priced for perfection.

This is not a deep-value screen. It is a quality-growth pullback screen. The goal is to find companies where the market has reset expectations, but the fundamentals may still be intact.

The 12 stocks that passed the screen

• Nvidia (NVDA): The leading AI chip company. Nvidia is central to AI compute demand through GPUs, data-centre platforms and AI infrastructure.
• Microsoft (MSFT): A cloud and software leader. Its AI exposure comes through Azure, Copilot, enterprise software and productivity tools.
• Micron (MU): A memory semiconductor company. Micron is increasingly important to the AI supply chain because high-bandwidth memory is a key bottleneck for AI workloads.
• KLA Corp (KLAC): A semiconductor equipment company focused on process control, inspection and metrology for advanced chip manufacturing.
• Marvell Technology (MRVL): A semiconductor company exposed to custom silicon, data-centre connectivity and networking chips.
• Arista Networks (ANET): A cloud networking company providing high-performance networking equipment for hyperscalers and data centres.
• ServiceNow (NOW): An enterprise software company helping businesses automate workflows, with AI increasingly embedded into its platform.
• Lumentum (LITE): An optical networking company providing components used in cloud and AI data-centre connectivity.
• Teradyne (TER): A semiconductor test equipment company exposed to testing demand for advanced chips, AI compute and memory.
• Nova (NVMI): A semiconductor metrology company providing process-control tools used in advanced chip manufacturing.
• Onto Innovation (ONTO): A semiconductor equipment company exposed to inspection, metrology and advanced packaging, including AI-related chip manufacturing.
• Guidewire Software (GWRE): A cloud software company serving the insurance industry, with AI optionality but less direct AI monetisation than infrastructure names.

Post-selloff AI quality score

After identifying the 12-stock universe, the next step is to rank the names.

The model uses four factors and a total score of 100 points.

1. Business quality — 30%

This captures whether the company is fundamentally strong enough to deserve a second look after the selloff.

The sub-factors include:

• 5-year revenue CAGR
• 5-year diluted EPS CAGR
• Next-year sales growth
• Gross margin
• Net debt/EBITDA
• Free cash flow yield

2. AI monetisation engine — 30%

This is the manual overlay.

It asks: is AI already showing up in revenue, backlog, orders, cloud usage, pricing power or customer adoption?

3. Valuation improvement — 20%

This factor measures whether the selloff has created better valuation support.

To make the valuation score more robust, I would combine:

• Absolute valuation: forward P/E
• Lower forward P/E scores higher.
• This captures whether the stock is still priced for perfection.
• Relative valuation: P/E Z-score versus 5-year history
• A negative Z-score means the stock is trading below its own 5-year average valuation.
• A positive Z-score means the stock is still trading above its own 5-year average valuation.
• This helps identify where the valuation reset is actually meaningful.

Suggested valuation scoring

A. Forward P/E score — 50% of valuation score

• P/E below 15x = 5
• 15–25x = 4.5
• 25–35x = 4
• 35–45x = 3.5
• Above 45x = 3

B. Relative P/E Z-score — 50% of valuation score

• Z-score below -1.5 = 5
• -1.5 to -0.5 = 4.5
• -0.5 to +0.5 = 4
• +0.5 to +1.5 = 3.5
• Above +1.5 = 3

Combined valuation formula

Valuation score = Forward P/E score × 50% + Relative P/E Z-score × 50%

Then this combined valuation score feeds into the overall model at a 20% weight.

4. Correction from high — 20%

This uses the field: price as a percentage of 52-week high.

The lower the number, the bigger the correction.

Suggested scoring:

• Down more than 40% = 5
• Down 25–40% = 4.5
• Down 20–25% = 4
• Down 15–20% = 3.5
• Down 10–15% = 3

This gives credit to stocks where the valuation reset has been more meaningful.

Overall Scoring formula

The final score is a weighted average, not a simple average.

Total score =

• Business quality × 30%
• AI monetisation engine × 30%
• Valuation improvement × 20%
• Correction from high × 20%

This matters because not all factors should carry the same importance. After an AI selloff, business quality and actual AI monetisation should matter more than just how far a stock has fallen.

Indicative ranking from the 12-stock list

*Guidewire has no relative P/E Z-score available in the screen, so the valuation score uses only its forward P/E for now. A neutral relative valuation score could also be used.

What the ranking suggests

The model still puts Micron, Nvidia and Microsoft at the top because they combine strong business quality with clear AI monetisation and reasonable valuation support after the selloff.

Micron screens especially well because AI is not just about compute. It is also about memory bandwidth, and HBM remains one of the clearest bottlenecks in the AI infrastructure chain. That gives Micron a stronger AI monetisation profile than a traditional memory-cycle label would suggest.

Nvidia remains one of the strongest AI revenue engines in the market. The correction from highs is less dramatic than some other names, but its business quality and direct AI monetisation visibility remain exceptional. Its negative relative valuation Z-score also suggests the stock has become more reasonable versus its own history. The key risk is that expectations are still high, so earnings will need to keep supporting the valuation.

Microsoft offers a more diversified AI exposure through cloud, enterprise software and productivity tools. It is less directly cyclical than semiconductors, while still offering exposure to AI adoption through Azure, Copilot and enterprise software demand. Its relative valuation also looks more supportive after the reset.

The next group — Arista, Marvell, Teradyne, KLA, Nova, Onto Innovation and Lumentum — represents the broader AI infrastructure chain. These companies are exposed to networking, custom silicon, semiconductor testing, metrology, optical connectivity and advanced chip manufacturing.

This is important because the AI trade is not only about GPUs. AI infrastructure also needs memory, networking, testing, inspection, packaging and data-centre connectivity.

ServiceNow and Guidewire are more software-led stories. ServiceNow has a credible enterprise AI adoption angle and a meaningful correction from highs, but investors may want more evidence that AI is driving revenue, upsell or margin expansion. Guidewire is more of a cloud software compounder with AI optionality, rather than a direct AI monetisation engine.

Risks to watch

There are still important risks.

• Earnings expectations may still be too high: Some AI leaders may need to deliver very strong results simply to justify current valuations.
• AI capex could slow or shift: If hyperscalers change spending plans, parts of the AI infrastructure chain could reprice quickly.
• Correlations can rise again: Even diversified AI infrastructure names may sell off together if the AI trade unwinds further.
• Valuation reset does not always mean cheap: A stock can be down from its highs and still be expensive.
• Software AI monetisation needs proof: Companies with strong AI narratives but limited financial evidence may remain vulnerable.
• Macro and geopolitical risks remain relevant: Higher yields, Fed repricing, energy volatility or geopolitical shocks can keep pressure on long-duration growth stocks.

Bottom line

The selloff has created a more interesting entry point in parts of technology, but it has also raised the bar for stock selection.

The best post-selloff candidates are not necessarily the stocks that have fallen the most. They are the ones where: price has corrected, but earnings power, business quality and AI monetisation remain intact.

This screener highlights a useful watchlist across AI compute, cloud, memory, networking, optical infrastructure, semiconductor equipment and selected software names.

For investors, the next phase of the AI trade may be less about owning everything with an AI label — and more about identifying the companies that are actually monetising the buildout.