Broadcom's miss sparked fears the AI trade is over. Is it?

When Broadcom reported earnings that fell short of expectations for its custom AI chip business, the reaction was swift and indiscriminate — the stock fell more than 12%, dragging Micron Technology down more than 7% and ARM Holdings down more than 4% as investors rotated out of anything touching the AI hardware trade.

The selling spread beyond US markets. In South Korea — where the KOSPI has become the clearest proxy for the global AI investment cycle — the index fell more than 6% at its intraday low. Samsung Electronics and SK Hynix, the two giants that produce the vast majority of the high-bandwidth memory chips (HBM) that power AI data centres, bore the brunt.

The narrative hardened quickly: the AI trade is over, valuations were stretched, the cycle is turning. Is it?

AI demand is not slowing — it is not even close to being met.

Broadcom's Q2 earnings were exceptional by any historical standard: AI semiconductor revenue rose 143% year-on-year to USD 10.8 billion, and its Q3 AI guidance of USD 16 billion implies more than 200% year-on-year growth. The stock fell because the market had expected USD 17.2 billion — a shortfall of USD 1.2 billion on guidance that, if achieved, would already represent one of the fastest growth trajectories ever recorded by a company of Broadcom's scale. The reason is company-specific: Google, Broadcom's largest AI customer, has begun diversifying toward MediaTek-designed custom silicon, shifting part of its chip spend away from Broadcom. This is a Broadcom problem. It is not a semiconductor demand problem.

While US markets processed the guidance miss, TSMC's Chairman CC Wei was at his annual shareholders' meeting in Hsinchu saying the opposite. His message was unambiguous: global chip supply will fall short of AI-fuelled demand for years to come, sustaining revenue growth for the firm. Even with new manufacturing capacity being built in the US, TSMC cannot fulfil the demand led by its US customers — "it will be a long time before we can meet customer demand," Wei said, reiterating full-year sales growth guidance of more than 30% and signalling he would "like to" raise prices while avoiding the abrupt hikes that memory producers have imposed.

The demand signal from the equipment side is just as strong. Gary Dickerson, CEO of Applied Materials — one of the world's largest semiconductor equipment makers — describes his pipeline as offering "unprecedented visibility," with customer conversations already centred on 2027 and 2028 requirements, even as the company has doubled its operational capacity and is still falling short. Equipment orders are placed one to two years before chips are produced. The fact that Applied Materials is fielding customer conversations about 2027 and 2028 means semiconductor manufacturers are planning for years of sustained demand.

On the same day its stock sold off, SK Hynix was telling investors that its planned US listing — a raise of up to USD 14 billion targeting completion within 2026 — had received 'tremendously positive' feedback from stockholders. It also said favourable HBM pricing conditions are expected to continue into next year, and that demand for LPDDR memory — a low-power chip used in phones and tablets — from Nvidia's Vera Rubin AI platform may tighten supply across the broader memory market from 2027.

This is not the picture of a demand cycle that is turning.

The memory shortage is severe enough to produce going-concern warnings

If the demand signals from TSMC, Applied Materials and SK Hynix were not enough, the losers of the current chip shortage tell the same story.

Nintendo raised the price of its Switch 2 console by USD 50, citing component costs. Camera maker GoPro filed a going-concern warning with the SEC, citing memory cost increases of 80% to 115% in a single week in March — the first named corporate casualty of the shortage. Counterpoint Research forecasts global smartphone shipments will fall 13.9% in 2026, the steepest annual contraction on record and "exacerbated by the Iran war," with Xiaomi projected to lose 28% of its volumes as budget-tier manufacturers absorb cost increases that price-sensitive consumers will not tolerate.

The shortage is now large enough to require a government response. Nine US trade associations — including the Alliance for Automotive Innovation, the National Retail Federation and the Medical Device Manufacturers Association — wrote jointly to the US Treasury and Commerce on 3 June 2026, warning of "an urgent imbalance in the market for memory chips" threatening American households and supply chains.

None of this is accidental. HBM — the specialised memory stack that every Nvidia AI server requires to function — generates three to five times the revenue per wafer of the standard DRAM that goes into phones and gaming consoles. Samsung and SK Hynix are directing capacity toward it, leaving consumer electronics manufacturers to compete for what remains.

A shortage that produces going-concern warnings, triggers a nine-industry coalition letter to the US Treasury and drives a 14% increase in smartphone wholesale prices is not a symptom of slowing demand. It is evidence that HBM demand remains exceptionally strong.

If valuations worry you, buy Asia semiconductors

If valuations worry you, the answer is not to avoid semiconductors — it is to buy the cheaper ones. Asia semiconductor companies are not riding the AI wave. They are the manufacturing backbone of the global AI buildout: the memory, the foundry, the packaging and the equipment that every AI chip in the world requires before it can function. They are producing the same earnings cycle as their US peers and trading at a discount to them.

The rally has been strong, but so have the earnings. Samsung's operating profit multiplied more than eightfold year-on-year in Q1 2026, SK Hynix posted an operating margin of 72% — an all-time record — and TSMC raised its full-year guidance to above 30% growth. When the stocks fell on 5 June 2026 — Samsung by 6.4% and SK Hynix by 9.9% — the earnings did not fall with them. Investors who have been waiting for a better entry point now have one.

Our preferred vehicle for this thesis remains the Global X Asia Semiconductor ETF (HKEX:3119), which captures the full Asian semiconductor supply chain without the concentration risk of a single-market ETF. Based on a fair PE ratio of 18x, our target price for the ETF is HKD 318. 3119 last traded at HKD 182.1, implying upside of more than 74%.

Here is why we prefer it over the single-market alternatives. Single-market ETFs on Korea and Taiwan carry a concentration problem that 3119 is specifically designed to avoid. Samsung and SK Hynix together represent approximately 50% of the KOSPI benchmark, making the Franklin FTSE South Korea ETF (NYSE:FLKR) effectively a concentrated bet on two names. TSMC and MediaTek together account for approximately 27% of the Franklin FTSE Taiwan ETF (NYSE:FLTW) — better diversified, but still a two-name concentration in a single market. Neither vehicle gives you the full supply chain.

3119 holds both Korean and Taiwanese names alongside Japanese semiconductor equipment makers. Its top holdings: SK Hynix (14.6%), Samsung (11.1%), MediaTek (10.6%), TSMC (7.7%), Tokyo Electron (5.0%), Advantest (3.5%) and Kioxia (3.1%) — with the remaining weight spread across the broader Asian semiconductor supply chain. Same thesis. Better diversification.

So: is the AI trade over, valuations stretched, the cycle turning? The companies that see every chip order in the world — TSMC, Applied Materials, SK Hynix — are all saying the same thing: supply cannot keep up, and it may not be able to for years. The selloff did not change that. Rallies do not move in straight lines — two steps forward, one step back is how every durable cycle has always worked. The selloff was the one step back. The entry point is better than it was 48 hours ago. The thesis is unchanged. Buy Asia semiconductors.

Declaration:

This research report was prepared with the assistance of artificial intelligence (AI) tools. iFAST Financial Pte Ltd does not rely exclusively on AI for content generation; the content of this report – including all investment theses, ratings, price targets and conclusions – has been independently reviewed and verified by the research analyst(s) to ensure accuracy and professional integrity.

For specific disclosure, at the time of publication of this report, IFPL (via its connected and associated entities) holds a NIL position in the abovementioned securities. The analyst who produced this report holds positions in Global X Asia Semiconductor ETF.