The era of Federal Reserve Chair Wash has formally begun, with last week's hawkish policy meeting statements sending a chill through the market sentiment.
Hong Kong stocks continued their decline, breaching the 24,000-point level.
On the evening of June 20th Beijing time, US-Iran negotiations encountered further complications. The Iranian military announced the closure of the Strait of Hormuz, citing actions by the US and Israel. Former US President Trump suggested that if a deal with Iran cannot be reached, the US might impose tolls for passage through the strait. Until a formal, permanent ceasefire is established, various frictions are likely to recur.
Deutsche Bank, in its latest research report, comprehensively raised its inflation forecasts and completely reversed its previous monetary policy predictions. It now expects the Federal Reserve to implement two rate hikes (totaling 50 basis points) by 2026, pushing the federal funds rate to 4.1%, and does not rule out the possibility of an earlier hike in July.
This week's release of the US May core PCE price index will be scrutinized to see if it reinforces these expectations.
The primary issue for Hong Kong stocks currently is the collective weakness of established blue-chip stocks. If the major giants cannot halt their decline, it will be difficult for the broader market to stabilize.
Technology stocks have become the focus of market capital flows. Whether the tech sector can maintain its strength will depend on upcoming catalysts, such as the earnings report from memory chip giant Micron Technology and whether NVIDIA's annual shareholder meeting presents new expectations.
The Lujiazui Financial Forum released relevant policy signals. While detailed implementation rules for artificial intelligence have not been formally issued, large AI models are already aligned with the fifth set of listing standards. Elon Musk predicted that domestic large models would catch up to Fable by Q1 2027, with ZhiPu's Tang Jie suggesting the catch-up speed could be even faster.
Other positive catalysts include: leading domestic internet companies are making bulk purchases of domestic GPUs, and Day0 hardware-software co-verification indicates the maturing domestic computing ecosystem. The leasing price for NVIDIA's B200 GPU is set to rise from $2.63 to $5.10 per hour upon contract renewal in October, an increase of approximately 94%. Domestic GPU alternatives are expected to continue gaining strength.
Intel CEO Chen Lifu stated he is shifting investment focus towards advanced packaging technologies like EMIB, glass substrates, and new materials such as gallium nitride (GaN), silicon carbide (SiC), indium phosphide (InP), and synthetic diamonds. These related areas are worth watching.
Featured Stock of the Week
Morgan Stanley predicts the global AI optical module PCB market will grow from $620 million in 2025 to $3.77 billion by 2028, a more than fivefold increase in three years.
Shenzhen has issued a document promoting the upgrade of optical modules to 1.6T and 3.2T, focusing on developing packaging technologies like silicon photonics and CPO.
Higher optical module speeds correlate with more expensive circuit boards: a 400G module board costs about $55, an 800G board rises to $70-$80, and a 1.6T board adds another $20-$30.
The company's profitability continues to improve, with its gross margin on an upward trajectory. Q1 optical module gross margin was 32.8%, and institutions forecast the full-year 2026 comprehensive gross margin will exceed 30%. Every 10% increase in the proportion of 1.6T products boosts overall gross margin by 1 percentage point.
The company is one of the few domestic firms capable of mass-producing and supplying 800G/1.6T silicon photonics products, with self-developed, end-to-end mass production capabilities.
Core business highlights: 800G optical modules account for over half of its optical module revenue, with a gross margin of 28%–32%, significantly higher than 400G and lower-speed products. They are compatible with NVIDIA GB200/H200 servers, with a full-year 2026 shipment target exceeding 3 million units, maintaining a premium price range of $380–$400 per unit.
Mass batch shipments of 1.6T optical modules are scheduled to begin in Q1 2026, featuring a 3nm DSP silicon photonics solution. Per-unit gross margin is over 35%, with unit bandwidth cost reduced by 50%. With adoption by NVIDIA, Meta, and Microsoft, these modules accounted for 45% of optical module revenue in Q1, with penetration expected to exceed 60% in the second half.
Forward-looking technology includes pre-research prototypes for 3.2T/6.4T NPO/CPO, targeting ultra-large-scale AI clusters post-2027. Its long-distance 1.6T solution aims to capture the incremental market for data center interconnects, a rare strategic move among peers.
The company benefits from a triple resonance of increasing share of high-end 800G/1.6T shipments, silicon photonics cost reduction, and the ramp-up of high-margin overseas orders, with scale effects continuously releasing profit potential.
It has ample orders in hand, being a core supplier for Cisco's 800G modules with approximately 60% share. Orders are being fulfilled steadily, supporting full-year revenue in the range of 7 to 8 billion yuan. Meta represents the most significant new variable for H2 2026 performance, with production capacity already reserved in advance.
Full-year 800G shipments are projected at 600,000-700,000 units. Looking further ahead to 2027, a framework agreement for 1 million units is in place, with 1.6T modules ordered simultaneously. There is also a stable order for 300,000 sets of DCOM optical modems, generating revenue of about 900 million yuan.
The company supplies Microsoft with 800G silicon photonics modules in bulk, with 2026 additional orders exceeding 300 million yuan. A 1.6T framework agreement is in place, with follow-on orders expected as AI server production expands.
For NVIDIA, it supplies 800G modules for GB200/GB300 servers with ongoing replenishment orders. Its 1.6T silicon photonics modules have passed system-level certification, with small-batch deliveries starting in Q2.
Total high-speed optical module deliveries are projected at 2.8-3.3 million units, with an annual production capacity cap of 5 million units. Orders are robust, and capacity is ramping up as needed.
The company's Malaysia factory is dedicated to fulfilling overseas orders for Meta and Cisco. Its Mexico factory is scheduled for production in Q2 2026, reaching full capacity of 1 million units by year-end to handle new overseas orders for 2027.
Substantial framework orders in hand support high full-year earnings growth. Profit elasticity is expected to further amplify in the second half as 1.6T volumes increase. Medium to long term, the company stands to benefit strongly from global AI computing upgrades and the iteration of next-generation optoelectronic packaging technologies, indicating solid growth certainty.
Industry Perspective
How much further can the AI rally go? Xingzheng Strategy points out that a month ago, it proposed that the core indicator for the AI rally's sustainability lies in hyperscalers' capital expenditures (Capex). Since 2013, the year-on-year change in Capex has shown a high positive correlation with SOX earnings growth expectations. Furthermore, in Q3 1999, the peak in US information processing equipment and software investment growth led the peak in tech sector earnings growth by 1-2 quarters. This view has now gained widespread market acceptance.
As of the latest data, market expectations for the Capex of the top five cloud service providers (CSP5) this year are 7.5 trillion yuan (+74%), with expectations for 2027 at 9 trillion yuan (+22%).
It's worth noting that over the past two years, the initial Capex growth expectations at the start of the year have ended up being about three times lower than the final figures (with significant upward revisions occurring particularly across the four earnings seasons). Therefore, it is not advisable to linearly extrapolate based on current expectations.
In essence, the amount and pace of Capex are primarily determined by the return on investment (ROI), which currently depends on the Annual Recurring Revenue (ARR) of large AI models. For example, concerns about an "AI bubble" in Q4 last year began with the slowdown in OpenAI's ARR growth from June to August and ended with the "sudden rise" of Anthropic in December. The expected Capex growth for 2026 in Q4 last year was less than 40%, but the actual Q1 figure was 84%.
Looking ahead, Xingzheng Strategy believes that before the two companies go public in the second half of the year, the "positive" information they disclose is likely to continue supporting a virtuous cycle of rising ROI and Capex, as companies tend to emphasize their strong technology and commercialization capabilities during fundraising periods.
For instance, in February and May of this year, the announcements of successful fundraising rounds by Anthropic and OpenAI largely coincided with the disclosure of their ARR figures.
Data also provides support: Since 2023, for OpenAI and Anthropic's market-based fundraising rounds (excluding single-company directed investments), the probability of the SOX index rising in the month of successful fundraising is 72.7% (90% if excluding March this year and April last year), with a median increase of 9.6%, significantly higher than the average of the three months prior to fundraising (1-2%).
Regarding specific ARR numbers, two reference points are offered. The first comes from Altimeter Capital, which led Anthropic's funding round in May, predicting total ARR for major AI labs to reach $300 billion by the end of 2027.
The second comes from overseas buy-side expectations for the combined ARR of Anthropic and OpenAI by the end of 2026 (currently ~$80 billion): 31% expect $140-160 billion, 25% expect $160-170 billion, 25% expect $180-200 billion, and 19% expect over $200 billion.
Market Data Analysis
Hong Kong Exchanges data shows the total number of open contracts for the Hang Seng Index Futures (June) is 137,670, with a net open interest of 52,816 contracts. The settlement date for the June Hang Seng Index Futures is June 29, 2026.
With the Hang Seng Index at 23,925 points, the concentration zone for bull contracts is near the central axis, while bear contracts show deviation, indicating potential short-selling pressure in the Hong Kong market.
New Federal Reserve Chair Wash declined to provide forecasts for the latest dot plot, leaving the decision for the next meeting to be judged based on the latest data and circumstances. The market currently expects the Fed to raise its policy rate by 25 basis points in October this year.
Hong Kong stocks are following the volatility of US markets, with the Hang Seng Index viewed as bearish for the week.
Closing Commentary
On the surface, the indices are weak, but the core issue is a systemic drying up of liquidity, not merely a fundamental collapse.
Five major "liquidity drains" have precisely hit the vulnerability of Hong Kong as an offshore market: fluctuating Sino-US relations leading to continued reduction in Western long-term capital allocation; the Federal Reserve's balance sheet reduction over nearly four years draining global US dollar liquidity; surging Japanese and Korean stock markets creating a strong siphoning effect; coupled with Hong Kong's own massive IPO and secondary financing activities aggressively drawing capital. The remaining capital is heavily concentrated in AI hardware technology, causing traditional blue-chip stocks to fall into a vicious cycle of "valuation pressure - capital flight."
Hong Kong now resembles a "large ship docked in port" – its valuation is at a global low, and asset quality has not deteriorated systematically. What it lacks is a "tailwind." Expectations for Federal Reserve rate cuts in the second half of the year and a recovery in mainland corporate earnings could become key catalysts to move the market.
Historical patterns also show that Fed rate-cutting cycles often serve as windows for high-beta rebounds in Hong Kong stocks.
However, investors must remain clear-eyed: the long-term trend of Sino-US decoupling has not changed, and the inherent liquidity fragility of Hong Kong as a front line of this contest is difficult to fundamentally alter.
Rather than constantly asking "when will the tailwind come," the current approach should be to patiently screen for truly high-quality assets with long-term value, and then make deliberate allocations once a liquidity inflection point is confirmed.
Comments