Sinolink Securities released a research report stating that AI demand appears robust in both the short and medium term, as evidenced by multiple companies in the AI supply chain exceeding earnings expectations, the rapid iteration of Google's TPU, the explosive revenue growth of Anthropic, TSMC's accelerated expansion of advanced process capacity for AI chips, major CSPs securing long-term agreements to lock in memory capacity, and rising GPU rental prices. The firm forecasts that the number of ASICs from Google, Amazon, Meta, OpenAI, and Microsoft will experience explosive growth between 2026 and 2027. Currently, many AI-PCB companies are seeing strong orders, operating at full capacity with full sales, actively expanding production, and are expected to sustain high earnings growth. Demand for AI copper clad laminates is also strong. Due to slow capacity expansion among overseas laminate manufacturers, leading mainland Chinese laminate producers are poised to benefit significantly. Key sector sentiment indicators are as follows: consumer electronics (steady improvement), PCB (accelerated improvement), semiconductor chips (steady improvement), semiconductor foundry/equipment/materials/components (steady improvement), display panels (bottoming out), passive components (steady improvement), and packaging & testing (steady improvement).
The main points of Sinolink Securities' view are as follows: Surging demand for AI computing power is driving earnings beats across the supply chain, with capacity locking for critical materials potentially becoming the norm. Recently, several AI supply chain companies have issued better-than-expected earnings guidance or reported strong orders. Cisco has secured $5.3 billion in AI infrastructure orders from hyperscale cloud providers for its 2026 fiscal year to date. Based on this momentum, Cisco has significantly raised its full-year AI order expectation from a previous $5 billion to $9 billion, an increase of 80%. SMIC's revenue guidance for Q2 2026 is for sequential growth of 14% to 16%, with a gross margin guidance of 20% to 22%, representing a 2 percentage point increase from the previous quarter's guidance. Based on customer demand and order backlog, management expressed greater optimism about this year's overall operations compared to the previous quarter. Applied Materials reported its Q2 FY2026 (ending April 26, 2026) results, with several performance metrics reaching record highs. Revenue for the next quarter is projected to grow nearly 23% year-over-year, exceeding market expectations. The company forecasts its semiconductor equipment business to grow over 30% in 2026. Taiyo Yuden also reported better-than-expected results for FY2025, with profits surging 535% year-over-year, and predicts its AI server MLCC revenue will jump 80% in FY2026.
Amid robust AI demand, on May 13th, Tower Semiconductor announced it had signed a long-term silicon photonics technology contract worth approximately 8.821 billion yuan with a key customer. This order will translate into revenue starting in 2027, laying a solid foundation for future earnings growth. Tower Semiconductor has already received a $290 million capacity reservation prepayment from the customer to secure dedicated silicon photonics foundry capacity. Concurrently, the parties have entered into a deeper, long-term partnership, signing a larger-scale wafer foundry agreement for 2028, with related additional prepayments due by January 2027, further locking in future capacity supply. The firm believes that with the explosive demand for global 800G and 1.6T optical modules, silicon photonics technology—leveraging its advantages of high bandwidth, low power consumption, and high integration—is well-suited for high-speed optical interconnect needs in the AI computing era and represents a core technological direction for data centers and high-speed optical modules. Leading industry players securing capacity through prepayments and long-term agreements is becoming a mainstream trend. Other critical materials for AI computing hardware, such as EML optical chips, Faraday rotators, and memory chips, are also seeing capacity locked in via long-term agreements.
On May 14th, at its "2026 Technology Symposium," TSMC stated it will increase capacity for its most advanced 2nm process chips. The first-year output of 2nm wafers is expected to be 45% higher than the output for 3nm at the same stage. The compound annual growth rate (CAGR) from 2026 to 2028 is projected to reach 70%. Meanwhile, CoWoS advanced packaging capacity is also expected to achieve a CAGR exceeding 80% between 2022 and 2027. Driven by the explosive global growth of AI and high-performance computing applications, TSMC is advancing fab expansion at twice its historical pace while deploying new capacity globally and increasing investment in advanced processes and packaging. TSMC plans to build and renovate a total of 18 semiconductor fabs worldwide, with capacity deployment covering two core areas: wafer manufacturing and advanced packaging, including five advanced packaging plants. TSMC also revised its forecast for the global semiconductor market upward, now expecting it to exceed $1.5 trillion by 2030, up from a previous estimate of $1 trillion. Within this, AI and high-performance computing are projected to account for 55%, smartphones for 20%, and automotive applications for 10%.
Risk warnings: Risks associated with demand recovery falling short of expectations; risks of slower-than-expected progress in AIGC; risks of further escalation in external sanctions.
Comments