• The GF CSI All-Share Information Technology ETF (SZSE: 159939) offers the clearest earnings visibility, with global AI capex and domestic substitution already translating into demand for the chips and hardware required to build data centres.

• The iShares Hang Seng TECH ETF (HKEX: 3067) offers an attractive re-rating opportunity, trading well below its long-term valuation average as the food-delivery subsidy war eases and AI monetisation accelerates through Alibaba Cloud and Tencent advertising.

• We favour an equal-weight core allocation to both — one already delivering confirmed earnings, the other a deep value play with an earnings recovery story still to play out — with a smaller, complementary allocation to the T. Rowe Price China Evolution Equity Fund for broader exposure.

Global attention has focused overwhelmingly on the US AI boom, and understandably so. Private AI funding in the US reached roughly US$286 billion in 2025 — more than 20 times the amount invested in China — according to Stanford HAI's 2026 AI Index Report. Yet less talked about is China's own, parallel opportunity across two distinct areas: the hardware suppliers already benefiting from AI infrastructure spending, and the internet platforms whose AI investments are only beginning to translate into revenue and earnings. This creates two complementary trades:

• The GF CSI All-Share Information Technology ETF (SZSE: 159939) covers chips and the hardware required for data centres, including servers, optical modules and other computing, networking and electronic components. It represents the “selling equipment and computing infrastructure” side of the value chain.
• The iShares Hang Seng TECH ETF (HKEX: 3067) covers cloud computing, foundation models and internet applications. It represents the “selling services and models, while improving platform monetisation” side of the value chain.

159939: AI capital expenditure has already translated into earnings

Hardware is usually the first segment in the AI value chain to show earnings, because it sits closest to the spending. Before models can be trained or deployed at scale, companies first buy chips, servers, optical modules, and other data-centre equipment — so as global tech capex rises, orders reach infrastructure suppliers before they reach the model and application layers. That's the segment 159939 covers: chips and the hardware supporting data-centre buildout.

The evidence is already in the financials, driven by two forces moving in the same direction. First, the scale of global AI capex continues to direct real orders to Chinese hardware suppliers. Second, export restrictions are pushing Chinese companies toward domestic alternatives, adding a parallel source of demand.

Table 1: Q1 2026 Performance of Selected Portfolio Holdings and Segment Leaders

Company	Sub-segment	Sales Growth YoY (%)	EPS Growth YoY (%)
Cambricon Technologies	Chips – Design	159.13%	182.35%
Hua Hong Semiconductor	Chips – Foundry	18.22%	513.10%
GigaDevice Semiconductor	Chips – Memory	119.38%	525.71%
Foxconn Industrial Internet	Infrastructure – Servers	56.52%	103.85%
Yuanjie Semiconductor	Infrastructure – Connectivity	320.90%	1129.41%

This shows up at the company level, and across both demand channels. Cambricon, a domestic AI chip designer gaining share as export restrictions push Chinese buyers toward homegrown alternatives, posted sharp growth: revenue up 159.13%, earnings up 182.35% — a clear domestic-substitution story. Foxconn Industrial Internet, which assembles AI servers and networking equipment for both global hyperscalers and China's own cloud providers, saw revenue climb 56.52% and earnings 103.85%, capturing demand from both drivers at once.

Together, these examples show up directly in the index-level numbers: the underlying index posted 20% revenue growth and 74% earnings growth in Q1 2026, year-on-year — a thesis already tested against a full reporting season, not a forward assumption. Global AI capex remains firmly intact: China's June exports rose 27% year-on-year, the fastest pace in four months, driven chiefly by AI-related chip and computing-power orders.

The domestic-substitution driver — Chinese buyers turning to homegrown chips because they can no longer easily buy foreign ones — holds as long as US export controls stay in place, backed by policy and government funding, with no sign of near-term reversal. Only a durable easing of trade tensions would remove that driver, and nothing currently on the table points to that outcome. Until then, 159939 remains a high-conviction core position.

3067: Traditional businesses are recovering, while AI commercialisation is beginning to emerge

Compared with 159939, 3067 sits further downstream in the AI value chain.

Its major holdings are consumer internet platforms rather than infrastructure suppliers. Their earnings this year have been distorted by two temporary factors: a food-delivery subsidy war that significantly weakened the profitability of Alibaba, Meituan, and JD.com; and heavy AI investment, which has weighed on margins in the near term even as its benefits have yet to show up in revenue.

Traditional businesses: The drag from platform price competition is easing

Regulators have repeatedly criticised unsustainable low-price competition, while platform subsidy activity has started to moderate. The platforms themselves have confirmed the shift directly: Meituan's CEO told investors the past six months had “proven this competition creates no real value for the industry,” Alibaba management said Q4 spending on Taobao Flash Sale would “contract significantly,” and JD confirmed its 2026 delivery investment would be lower than 2025's. The market therefore increasingly expects the industry to move away from disorderly price competition and towards a more stable competitive environment.

Selling and marketing expenses provide the clearest evidence of this shift: at Meituan, JD.com, and Alibaba, spending has declined noticeably from each company's respective cyclical peak, while operating profit has begun to recover. This is the strongest data point supporting the view that the worst phase may already have passed — for 3067, it removes one of the two drags on this year's earnings.

The other, AI investment, is where the more interesting story is.

Chart 1: Instant delivery price war: marketing spend retreating from peak

New businesses: Alibaba and Tencent demonstrate two different paths to AI monetisation

Alibaba has committed RMB380 billion of capital expenditure over three years to cloud computing and AI infrastructure. This investment is now beginning to translate into revenue. Alibaba Cloud Intelligence Group's revenue growth accelerated to 38% year-on-year in the quarter ended March 2026, while cloud revenue also increased as a share of group revenue.

More importantly, a clear inflection point has emerged. AI-related product revenue as a proportion of Alibaba Cloud's external commercial revenue — that is, revenue generated from customers outside the Alibaba Group — rose from above 20% in the quarter ended June 2025 to 30% in the quarter ended March 2026. According to Alibaba's own management guidance, the company expects this proportion to exceed 50% within the year.

The story is therefore shifting from “large capital expenditure with returns still unproven” to “investment beginning to translate into revenue at scale.”

Chart 2: Alibaba Cloud Intelligence’s revenue has been accelerating since the AI capex ramp-up

Tencent shows a different route to monetisation: AI is acting as an efficiency catalyst on top of an existing business, rather than a new revenue line of its own. This is most visible in advertising.

In the first quarter of 2026, Tencent's Marketing Services revenue grew 20% year-on-year, accelerating from 17% in the fourth quarter of 2025 — more than twice the pace of 9% group revenue growth.

The company explicitly attributed this acceleration to AI-driven upgrades to its recommendation models — the automated system that decides, for each user, which ad to show and how to price it. Tencent's AIM+ intelligent advertising system currently handles around 30% of advertiser spending on its platforms, and early case data already show measurable gains in advertising pricing and conversion rates for advertisers using it. As advertisers see a better return on ad spend, they have more reason to keep — or even increase — their budgets with Tencent, including by reallocating spend from other platforms. That combination of pricing power and share gain is what pushes Tencent's advertising revenue higher.

The market has already rewarded this shift: Alibaba's Hong Kong-listed shares jumped 12% on 8 July — their biggest one-day gain in ten months — after an analyst pre-earnings update reported narrowing losses in the company's instant-commerce business, with the Hang Seng TECH Index rising 5% alongside it.

Independent labs: 3067's exposure is arriving through IPOs, not through picking a competitive winner

Many leading Chinese model companies, including DeepSeek and Moonshot AI, remain privately held — which means the Hang Seng TECH Index still provides relatively limited direct exposure to pure-play foundation-model companies, regardless of how those companies rank against each other. That began to change when Zhipu AI and MiniMax listed in Hong Kong earlier this year and were added to the index in June — not because of their competitive standing, but because the index's own inclusion methodology picked them up once they went public. For the first time, 3067 gained direct exposure to independent Chinese foundation-model companies.

Both companies are growing rapidly — Zhipu's revenue increased 131.9% in 2025, while MiniMax's revenue rose 158.9% — but both remain significantly loss-making, and their combined index weight is still relatively small, which limits their near-term impact on the ETF's overall performance. If — and only if — their commercialisation continues to progress and revenue growth remains strong, their index weights could gradually increase; only under that condition would 3067 gain more direct participation in China's foundation-model industry itself, on top of the AI monetisation Alibaba and Tencent already generate through cloud, advertising and platform ecosystems. As more model makers reach IPO scale, 3067's exposure here should broaden the same way — one listing at a time.

Key risks

Macro risk: a weaker-than-expected China macro recovery could dampen consumer demand and enterprise IT spending across both exposures.

iShares Hang Seng TECH ETF (HKEX: 3067)

Re-escalation risk: if delivery-war competition among Alibaba, Meituan and JD.com resumes rather than continuing to fade, the 3067 recovery thesis would take longer to play out than currently expected.

AI monetisation risk: AI-related revenue at Alibaba and Tencent is still a minority of the total and depends on continued adoption; a slowdown in this trend would weaken the core argument for 3067's re-rating.

GF CSI All-Share Information Technology ETF (SZSE: 159939)

Capex-cycle risk: if hyperscalers delay spending or domestic AI infrastructure orders weaken, highly cyclical holdings across semiconductors, servers, PCBs and optical components could see earnings downgrades.

Policy-reversal risk: the domestic-substitution driver behind this year's earnings growth depends on US export controls on chips remaining in place. A durable easing of trade tensions — for instance, at the September US-China summit — would remove one of the two demand engines behind 159939's Q1 print, even as the global-capex driver remains intact.

Recommendation

We continue to favour an equal-weight core allocation to 159939 and 3067. One is already delivering confirmed earnings; the other is a deep value play with an earnings recovery story still to play out. Pairing them means you get exposure to two different stories in China's tech trade.

159939's earnings are supported by global infrastructure spending and a domestic-substitution driver we hold as high-conviction for as long as US export controls remain in place. 3067's opportunity rests on valuation: it has fallen to close to one standard deviation below its long-term average, leaving expectations sufficiently depressed that even modest positive catalysts — a further easing of delivery competition, continued acceleration in Alibaba Cloud, or stronger AI-driven advertising monetisation at Tencent — could trigger a meaningful rebound.

For investors seeking broader exposure beyond the two core ETF trades, the T. Rowe Price China Evolution Equity Fund offers actively managed, complementary exposure to less widely owned companies benefiting from AI investment, technological upgrading, and the green transition — capturing opportunities in industrials, specialised components, energy and other upstream businesses that are underrepresented in 159939 and 3067.

Chart 3: Forward P/E of HSTECH Index has fallen to near 1 standard deviation below its long-term average

Table 2: CSI All-Share IT Index earnings forecast

CSI All Share Info Tech Index	2025	2026E	2027E	2028E
PE Ratio (X)	68.0	51.3	41.7	34.7
Earnings Growth	45%	70%	23%	20%
Earnings Per Share	0.69	1.16	1.43	1.72
Target Price (CNY) Based on fair PE ratio of 55x				17224
Upside Potential				58.0%
Source: Bloomberg Finance L.P., iFAST Estimates Data as of 12 August 2026