UBS: Top Signals Have Emerged, but It's Too Early to Exit China AI Tech Stocks
Claire Weston
UBS flags four near-peak signals for China's AI tech hardware stocks, yet keeps an overweight rating — earnings growth is still running at ~80%, and history shows the final three months before a top average 48% more upside.
What are the four warning signals?
Mutual-fund holdings in A-share telecom and electronics sectors are near all-time highs, with heavy crowding in the same names.
Forward P/E ratios are also near historical peaks, and the valuation discount versus global peers has essentially vanished.
New-share supply is accelerating and capex keeps climbing — though capacity expansion remains relatively restrained, partly because high-end manufacturing equipment is supply-constrained.
This means → all four gauges are flashing at once — the market is hot, but not yet at the "last leg" stage.
If the signals are flashing, why stay?
The core reason is earnings momentum: expected profit growth sits at roughly 80%. History shows that when growth exceeds 30%, tech stocks outperform the broad market by 10%–20% per year.
Over the past three months, earnings estimates for these stocks were revised up 15% on average, while the broader market saw estimates cut by 2% — money is voting with conviction.
In plain terms = these companies are making money faster, not slower. Walking away now means leaving the fattest stretch on the table.
Risk note: historical data show that once revenue growth peaks, AI tech stocks underperform the market by ~2% over the following six months. Watch for the inflection.
How far out does order visibility stretch?
Contract liabilities — signed orders not yet delivered — and inventory-to-revenue ratios keep rising. Order visibility now extends to the end of 2027.
Orders accelerated markedly after DeepSeek's release in early 2025, with demand spreading from AI servers into power, cooling, data transmission, packaging, PCBs, and upstream materials.
This reflects an AI data-center build cycle still trending upward, with the supply chain stretching longer and the pool of beneficiaries widening.
Where does this rally sit versus history?
The 4G, 5G, and cloud-computing upcycles each lasted about two years, delivered ~100% excess returns, and saw valuations expand past 19× P/E.
The current AI rally, which kicked off with the DeepSeek moment in early 2025, has already produced 215% excess returns and a 16× P/E re-rating — above historical averages in absolute terms.
But UBS highlights a key pattern: in the final three months before past tech peaks, the relevant sectors still gained an average of 48%.
In plain terms = the numbers look stretched, but by the historical script, "near the top" is not the same as "at the top" — the last leg is often the most explosive.
Are China AI tech stocks still cheap versus global peers?
Since 2021, both Chinese and global AI hardware companies have roughly doubled revenue, but global peers have pulled clearly ahead on earnings growth, free cash flow, and margins.
Chinese stocks still trade at lower price-to-book and price-to-sales multiples, but on a P/E basis they are no longer cheap — the discount advantage is essentially gone.
This means → global peers are growing earnings faster in 2026, but Chinese peers edge ahead in 2027 — that is China's relative-advantage window.
What is the real test in 2027?
UBS stays positive for the rest of 2026 but warns that 2027 could be choppier as capacity expands and the growth base rises.
The report states: "2027 will be the year that tests which companies — and sectors — can truly deliver on AI promises."
Preferred sub-sectors include optical modules, memory, GPUs, copper-clad laminates, and semiconductor equipment leaders.
In plain terms = the bar for the next phase is whether a company can break into global supply chains and hold pricing power and a tech moat — the era of "anything AI goes up" is ending.
Content is for reference only, not financial advice.