Investor Q&A: Is the AI selloff a reset or a warning shot?

Key points:

1. AI selloff looks like a reset, not a regime break: The selloff is painful but still concentrated in crowded AI and tech winners, rather than showing broad recession stress, disorderly yields or an earnings collapse. The bull-market structure is not broken, but the easy rerating phase of AI is likely over.

2. AI has moved from storytelling to proof: Valuations have risen sharply, positioning is crowded, and “good” earnings may no longer be enough. The next phase needs evidence of real AI monetisation, capex discipline, funding sustainability, margin resilience and returns on infrastructure spend.

3. Stay invested, but diversify and be more selective: This is not a market to abandon risk, but investors should reduce concentration in crowded AI winners, prefer quality companies with visible earnings and free cash flow, and keep diversification through defensives, real assets and inflation-sensitive exposures as Fed, oil and geopolitics remain key risks.

1. What really caused the selloff?

It was not just one data point or one headline.

The jobs report may have been the spark, but the market was already vulnerable. It had become overbought, top-heavy and crowded around one dominant theme: AI.

Several things came together at once.

AI crowding: Semis and AI-linked names had become the default long trade. When everyone owns the same winners, even a small disappointment can trigger a much bigger unwind.

Top-heavy leadership: A small group of AI winners had been carrying the broader index. That makes the market look stronger than it really is on the way up, but more fragile on the way down.

Valuations priced for perfection: The SOX index was around 18x forward earnings at the end of March and is now closer to 30x. That is a huge rerating in a short period. The easy part of the rally was multiple expansion. From here, earnings need to do more of the heavy lifting.

Expectations are harder to beat: The Broadcom example was important because it showed that “good” is no longer good enough for AI-linked names. Investors want upside surprises, stronger guidance, clear monetisation and evidence that AI demand is still accelerating.

AI funding questions: The market is also starting to ask who funds the next leg of AI infrastructure. Alphabet’s funding moves, and now Meta’s, are reminders that AI is not just a growth story. It is also a very capital-intensive one. Investors are becoming more focused on financing needs, capex discipline, dilution risk and return on investment.

Fed repricing: Stronger US data makes the Fed put look further away. That matters when valuations are stretched. Higher-for-longer rates put pressure on long-duration growth stocks, and AI has become one of the market’s biggest long-duration trades.

2. Is this tech selloff the start of a broader bear market?

This looks more like a correction within a bull market than the start of a broad bear-market shift. The selloff has been painful, but it does not yet have the classic ingredients of a deeper breakdown: recession stress, disorderly yields, oil spiralling towards extreme levels, or a broad earnings collapse.

The pain is also not uniform. If this were full-blown risk-off, everything would be getting sold. Instead, the pressure is still concentrated in the crowded AI and tech winners, while defensives, value areas and laggards are still finding buyers. Read here on how boring sectors are coming to the rescue as AI questions get harder.

That matters. It suggests this is still a positioning and valuation reset, not yet a regime break.

3. Is the AI bubble bursting?

A bubble bursts when the core demand story collapses. That is not what we are seeing. What we are seeing is a valuation and expectations reset.

AI demand, capex and earnings momentum are still strong in many areas. But the technical backdrop has become far less supportive. The trade is crowded, valuations are higher, and investors now need more proof.

The bar has gone up. We have moved from the rerating phase to the proof phase.

That means AI can still be a long-term structural opportunity, but it is unlikely to be a straight line up from here.

4. Where does the AI theme go from here, and what risks should investors watch?

AI is likely to remain a long-term structural theme, but the next phase will probably look very different from the first one.

The first phase was about excitement, scarcity and rerating. Investors rewarded almost anything linked to chips, data centres, cloud, memory, networking and AI infrastructure. That phase was powerful, but it also pulled forward a lot of future returns.

The next phase needs more proof.

The market will increasingly separate companies with real AI monetisation from those simply carrying an AI label. Investors will want to see AI showing up in cloud usage, software pricing, chip orders, backlog, margins, customer adoption and guidance.

The key question from here is not whether AI is real. It is whether the earnings, cash flows and returns on capex can justify the valuation base.

There are several risks investors need to track.

Token pricing pressure: This is becoming one of the most important risks to the AI boom. The debate is no longer just about how powerful models are, but how cheaply they can deliver answers. If more software calls are routed to cheaper models, especially Chinese models, the pricing power of US hyperscalers and AI labs could come under pressure.

That matters because large AI capex plans are being justified by the assumption that token demand will keep rising and that pricing will support strong returns. But if token prices hit a ceiling, or if enterprise customers push back on AI costs, the revenue model becomes more complicated.

Energy cost pressure: AI is not just a compute story. It is also an energy story. Training and inference require large amounts of power, while data centres also need cooling, grid access and reliable electricity supply. If energy prices rise because of Middle East tensions, or if power availability becomes constrained, the cost base for AI infrastructure could move higher.

That matters because investors are already questioning whether AI capex can generate strong enough returns. Higher energy costs would make that hurdle harder to clear, especially for hyperscalers, data-centre operators and AI infrastructure providers that cannot pass those costs on.

Capex discipline: AI infrastructure spending is enormous. Investors will increasingly ask whether hyperscalers and AI companies can show returns on that spending. If capex keeps rising faster than revenue, margins and free cash flow could come under pressure.

Overcapacity risk: If too much compute, data-centre or memory capacity is built ahead of real demand, the market could shift from scarcity pricing to excess supply. That would be a problem for parts of the AI hardware chain.

Funding and dilution risk: AI is capital-intensive. If companies need to issue debt or equity to fund infrastructure, investors may become more focused on balance-sheet risk, dilution and the cost of capital.

Earnings concentration risk: If only a handful of companies capture most of the AI profits, the broader AI basket may struggle even if the theme remains structurally strong.

So the risk is not “AI is over.” The risk is that AI remains real, but earnings expectations are too high, margins are less clean, and the return on capex takes longer to prove.

That is why we have moved from the rerating phase to the proof phase. AI can still work, but leadership is likely to become more selective and more volatile.

5. Is this a buy-the-dip moment?

This is not the kind of selloff where investors should blindly chase every AI-linked stock that has fallen. The market is becoming more selective, and that is healthy.

The better opportunity may be in quality AI names that have corrected but still have strong earnings visibility, balance-sheet strength, pricing power, free cash flow and a clear monetisation path.

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 is also the reason for using a more structured framework for corrected AI stocks — looking at business quality, AI monetisation, valuation improvement and the scale of the correction. That is the focus of the “corrected quality AI stocks screener” article.

6. What should investors watch next?

The most important signals are:

US data: A hot inflation print would make it harder for the Fed to sound relaxed. After the strong jobs report, markets are more sensitive to any upside surprise.

Bond yields: Yields are higher, but not yet breaking. If they become disorderly, the pressure on long-duration growth stocks will increase.

Oil: Oil around current elevated levels is uncomfortable, but not yet a full macro shock. The bigger risk is if markets start pricing a sustained supply disruption.

AI earnings guidance: Good may not be good enough. Investors want proof that AI demand is accelerating and monetisation is real.

Market breadth: The key test is whether rotation can keep working. If money keeps moving into equal-weight, staples, healthcare, utilities, real estate and other laggards, the market can absorb the AI reset. If AI selling starts dragging everything else down, correction risk becomes more serious.

7. Is the market fully pricing in geopolitical risk?

Geopolitics is getting worse, not better, and markets are still underpricing the tail risk.

For now, markets are treating the Middle East as a volatility event, not a macro shock. Oil holding below extreme levels despite fresh Iran headlines suggests investors are not yet pricing a sustained supply disruption.

That can change quickly.

The transmission channel is through oil, inflation expectations, shipping routes and broader risk sentiment. If energy infrastructure, shipping routes or the Strait of Hormuz come back into focus, markets may need to reprice much more aggressively.

The key point: geopolitics is not the main driver of the AI selloff, but it reduces the market’s ability to absorb other shocks.

8. Is the Fed likely to hike rates this year?

The Fed hike debate is back because the US economy is still resilient, the labour market surprised on the upside, oil is higher, and inflation risks have not disappeared.

Markets are now pricing the risk of one Fed rate hike this year, but that still looks unlikely in our view unless inflation broadens beyond energy.

The Fed can cool demand, but it cannot produce more oil, reopen shipping routes or reduce geopolitical risk. So hiking aggressively into a pure supply shock can be dangerous.

That said, the Fed cannot ignore second-round inflation effects. If higher oil starts feeding into broader prices, inflation expectations and wage-setting, then the policy debate becomes more complicated.

So the Fed message is nuanced: the market may be right to reintroduce hike risk, but a hike is still a high bar. The Fed would probably need evidence that the oil shock is becoming a broader inflation problem, not just a temporary supply-side shock.

9. Why is gold falling in a geopolitical shock?

Gold has stopped behaving like a clean safe haven, and that is an important signal.

The latest correction has pushed gold below its 200-day moving average for the first time since October 2023. That is a meaningful technical setback for a market that has spent much of the past four years in a powerful uptrend.

The issue is that gold is currently being hurt by the wrong kind of geopolitical risk.

Normally, gold performs best when geopolitical stress comes with falling growth expectations, lower real yields, a weaker dollar and expectations of central bank easing. This time, the risk is more energy-driven. Higher oil prices are feeding inflation concerns, supporting bond yields and keeping the dollar firm. That raises the opportunity cost of holding a non-yielding asset like gold.

There are several near-term risks to watch.

Higher real yields and a stronger dollar: This remains the biggest headwind. If inflation stays sticky and markets keep pricing Fed hike risk, gold may struggle to attract fresh investment demand.

Technical selling below the 200-day moving average: The break below the 200-day average matters because many systematic funds, momentum traders and risk-managed strategies use it as a trend filter. A sustained break can trigger position reductions and discourage fresh buying.

ETF outflows and weaker investment demand: If gold ETFs continue to see outflows, it suggests investors are reducing portfolio hedges rather than adding them. Without renewed investment demand, rallies may struggle to hold.

EM central bank selling risk: Some emerging-market central banks may choose to sell part of their gold reserves to support currencies, provide dollar liquidity or smooth market stress. This would not necessarily change the long-term central bank diversification story, but it could add tactical supply into a weak market.

Margin and liquidity pressure: When volatility rises and investors need to raise cash, even strong long-term assets can be sold. Gold can become a source of liquidity, especially after a large multi-year rally.

For the downside pressure to ease, gold needs two things.

First, inflation concerns need to cool. That could come through a Middle East peace deal, softer US activity data, lower energy prices or clearer evidence that the Fed does not need to lean more hawkish.

Second, gold needs to regain momentum. A move back above USD 4,500 would be the first important sign of stabilisation, followed by a recovery above the 50-day moving average near USD 4,600.

Until then, traders may stay focused on downside risks, while longer-term investors may wait for a catalyst that shifts attention back to gold’s structural supports: central bank reserve diversification, fiscal debt concerns, currency debasement risks and a more fragmented geopolitical order.

So the answer is: gold can still fall further in the short term, especially if yields and the dollar stay supported. But the bigger bull-market structure is not necessarily broken.

10. How should investors position in this environment?

This is not a market to abandon risk. But it is a market to be more selective.

Stay invested, but reduce concentration risk: AI can still work, but portfolios should not depend on one crowded theme.

Prefer quality AI over expensive AI: Look for strong balance sheets, real revenue, free cash flow, pricing power and visible AI monetisation.

Diversify beyond AI: Staples, healthcare, utilities, banks, real estate and dividend compounders can matter when high-growth leadership stumbles. This is exactly the point of the “Boring Saved the Day” article: diversification still matters most when the market’s favourite trade starts to wobble.

Keep some inflation hedges: Energy exposure and selective real assets may still have a role if geopolitical risks remain elevated. Gold can still be part of the toolkit, but it is not currently acting as a perfect hedge.

Use volatility more deliberately: For long-term investors, staggered buying may make more sense than chasing one-day rebounds. For active investors, rotation and breadth matter. For traders, risk management matters more than conviction because CPI, oil headlines, Fed repricing and AI earnings can all move markets quickly.