UBS: China AI Tech Forward Earnings Growth ~80%, Valuations Stretched but Trend Intact
N.R. Finch
UBS's latest strategy report calls China's AI hardware stocks "stretched, but not breaking" — excess returns have hit 215% since DeepSeek, with a 16× P/E re-rating, yet forward earnings growth still runs at ~80% and fundamentals remain intact.
How much has this AI rally delivered, and how does it compare to history?
Since DeepSeek triggered a re-pricing in early 2025, China AI hardware stocks have delivered ~215% excess returns with a ~16× P/E re-rating.
Historical comparison: the 4G, 5G, and cloud computing cycles each lasted about two years, delivering ~100% excess returns and 19+ points of valuation expansion.
This means → the current rally has more than doubled prior cycles' price gains, yet its valuation expansion has actually not caught up with them.
Valuations are this high — why isn't UBS calling for an exit?
The core argument is earnings momentum: over the past three months, AI tech earnings forecasts were revised up ~15%, while the broader China market saw forecasts cut ~2%.
UBS back-testing shows that when forward earnings growth exceeds 30%, tech hardware stocks typically outperform the market by 10%–20% per year.
Forward earnings growth currently sits at ~80%; even after a pullback, it could stay above 30% into 2027.
In plain terms = as long as earnings are growing fast, the market will pay up for expensive stocks. UBS is betting that growth rate has not peaked yet.
Crowding is high — what are the risk signals?
Mutual-fund positioning in electronics and telecom is near all-time highs.
A-share turnover concentration in top names is near record levels; the valuation discount of China AI hardware stocks versus global peers has largely vanished.
New share supply is rising and capex is turning up — classic late-cycle signals.
This means → crowding alone is not a reason to exit immediately, but the margin for error has become very thin.
Do macro factors matter? What actually drives these stocks?
UBS ran a correlation analysis: the U.S. 10-year Treasury yield shows a correlation of just 0.13 with China AI hardware stocks; the renminbi exchange rate, just 0.11.
The real drivers are industry-cycle indicators — contract liabilities-to-revenue at 0.75, forward revenue growth at 0.74, inventory-to-revenue at 0.64.
In plain terms = whether U.S. bond yields rise or the yuan weakens barely matters for these stocks. What matters is whether orders are coming in and products are shipping.
What does UBS recommend doing now?
After revenue growth peaks, AI hardware stocks underperform the broad market by an average of ~2% — an acceptable drag during a strong earnings phase.
This means → UBS is not calling for a full exit, but for a shift from "buy the sector" to "pick the winners" — moving from beta to alpha.
This reflects a judgment: the phase of broad-based sector gains is fading; the next leg of returns will come from choosing the right companies, not the right theme.
Why is 2027 the make-or-break year?
UBS characterizes 2026 as "the pro-cyclical trend continuing" — some smaller names may still ride sector beta and demand spillover.
But 2027 is when the market will rigorously test which companies can actually deliver on AI orders, defend technology moats, and sustain margins.
Across the supply chain, bottlenecks are spreading from GPUs alone to the entire infrastructure layer — HBM (high-bandwidth memory, the storage sitting next to AI chips that shuttles data at high speed) has become a core constraint, and advanced packaging is now a key supply bottleneck as well.
In plain terms = 2026 can still be a "rising tide lifts all boats" year. By 2027, it becomes a winnowing — companies that cannot show real orders and profits will be left behind.
Content is for reference only, not financial advice.