High Bandwidth Memory (HBM) is emerging as the preferred solution for overcoming memory transfer speed restrictions due to the bandwidth limitations of DDR SDRAM in high-speed computation. HBM is recognized for its revolutionary transmission efficiency and plays a pivotal role in allowing core computational components to operate at their maximum capacity. Top-tier AI server GPUs have set a new industry standard by primarily using HBM. TrendForce forecasts that global demand for HBM will experience almost 60% growth annually in 2023, reaching 290 million GB, with a further 30% growth in 2024.

TrendForce’s forecast for 2025, taking into account five large-scale AIGC products equivalent to ChatGPT, 25 mid-size AIGC products from Midjourney, and 80 small AIGC products, the minimum computing resources required globally could range from 145,600 to 233,700 Nvidia A100 GPUs. Emerging technologies such as supercomputers, 8K video streaming, and AR/VR, among others, are expected to simultaneously increase the workload on cloud computing systems due to escalating demands for high-speed computing.

HBM is unequivocally a superior solution for building high-speed computing platforms, thanks to its higher bandwidth and lower energy consumption compared to DDR SDRAM. This distinction is clear when comparing DDR4 SDRAM and DDR5 SDRAM, released in 2014 and 2020 respectively, whose bandwidths only differed by a factor of two. Regardless of whether DDR5 or the future DDR6 is used, the quest for higher transmission performance will inevitably lead to an increase in power consumption, which could potentially affect system performance adversely. Taking HBM3 and DDR5 as examples, the former’s bandwidth is 15 times that of the latter and can be further enhanced by adding more stacked chips. Furthermore, HBM can replace a portion of GDDR SDRAM or DDR SDRAM, thus managing power consumption more effectively.

TrendForce concludes that the current driving force behind the increasing demand is AI servers equipped with Nvidia A100, H100, AMD MI300, and large CSPs such as Google and AWS, which are developing their own ASICs. It is estimated that the shipment volume of AI servers, including those equipped with GPUs, FPGAs, and ASICs, will reach nearly 1.2 million units in 2023, marking an annual growth rate of almost 38%. TrendForce also anticipates a concurrent surge in the shipment volume of AI chips, with growth potentially exceeding 50%.

NVIDIA’s robust financial report reveals the true impact of AI on the technology industry, particularly in the AI server supply chain. …

ChatGPT’s debut has sparked a thrilling spec upgrade in the server market, which has breathed new life into the supply chain and unlocked unparalleled business opportunities. Amidst all this, the big winners look set to be the suppliers of ABF (Ajinomoto Build-up Film) substrates, who are poised to reap enormous benefits.

In the previous article, “AI Sparks a Revolution Up In the Cloud,” we explored how the surge in data volumes is driving the spec of AI servers as well as the cost issue that comes with it. This time around, we’ll take a closer look at the crucial GPU and CPU platforms, focusing on how they can transform the ABF substrate market.

NVIDIA’s Dual-Track AI Server Chip Strategy Fuels ABF Consumption

In response to the vast data demands of fast-evolving AI servers, NVIDIA is leading the pack in defining the industry-standard specs.

This contrasts with standard GPU servers, where one CPU backs 2 to 6 GPUs. Instead, NVIDIA’s AI servers, geared towards DL(Deep Learning) and ML(Machine Learning), typically support 2 CPUs and 4 to 8 GPUs, thus doubling the ABF substrate usage compared to conventional GPU servers.

NVIDIA has devised a dual-track chip strategy, tailoring their offerings for international and Chinese markets. The primary chip for ChatGPT is NVIDIA’s A100. However, for China, in line with U.S. export regulations, they’ve introduced the A800 chip, reducing interconnect speeds from 600GBps (as on the A100) to 400GBps.

Their latest H100 GPU chip, manufactured at TSMC’s 4nm process, boasts an AI training performance 9 times greater than its A100 predecessor and inferencing power that’s 30 times higher. To match the new H100, H800 was also released with an interconnect speed capped at 300GBps. Notably, Baidu’s pioneering AI model, Wenxin, employs the A800 chip.

To stay competitive globally in AI, Chinese manufacturers are expected to aim for the computational prowess on par with the H100 and A100 by integrating more A800 and H800 chips. This move will boost the overall ABF substrate consumption.

With the ChatBot boom, it is predicted a 38.4% YoY increase in 2023’s AI server shipments and a robust CAGR of 22% from 2022 to 2026 – significantly outpacing the typical single-digit server growth, according to TrendForce’s prediction.

AMD, Intel Server Platforms Drive ABF Substrate Demand

Meanwhile, examining AMD and Intel’s high-end server platforms, we can observe how spec upgrades are propelling ABF substrate consumption forward.

• AMD Zen 4:

Since 2019, AMD’s EPYC Zen 2 server processors have used Chiplet multi-chip packaging, which due to its higher conductivity and cooling demands, has consistently bolstered ABF substrate demand.

• Intel Eagle Stream:

Intel’s advanced Eagle Stream Sapphire Rapids platform boasts 40-50% higher computation speed than its predecessor, the Whitley, and supports PCIe5, which triggers a 20% uptick in substrate layers. This platform employs Intel’s 2.5D EMIB tech and Silicon Bridge, integrating various chips to minimize signal transmission time.

The Sapphire Rapids lineup includes SPR XCC and the more advanced SPR HBM, with the latter’s ABF substrate area being 30% larger than the previous generation’s. The incorporation of EMIB’s Silicon Bridge within the ABF substrate increases lamination complexity and reduces overall yield. Simply put, for every 1% increase in Eagle Stream’s server market penetration, ABF substrate demand is projected to rise by 2%.

As the upgrades for server-grade ABF substrates continue to advance, production complexity, layer count, and area all increase correspondingly. This implies that the average yield rate might decrease from 60-70% to 40-50%. Therefore, the actual ABF substrate capacity required for future server CPU platforms will likely be more than double that of previous generations.

ABF Substrate Suppliers Riding the Tide

By our estimates, the global ABF substrate market size is set to grow from $9.3 billion in 2023 to $15 billion in 2026 – a CAGR of 17%, underscoring the tremendous growth and ongoing investment potential in the ABF supply chain.

Currently, Taiwanese and Japanese manufacturers cover about 80% of the global ABF substrate capacity. Major players like Japan’s Ibiden, Shinko and AT&S, along with Taiwan’s Unimicron, Nan Ya, and Kinsus all consider expanding their ABF substrate production capabilities as a long-term strategy.

As we analyzed in another piece, “Chiplet Design: A Real Game-Changer for Substrates,” despite the recent economic headwinds, capacity expansion of ABF substrate can still be seen as a solid trend, which is secured by the robust growth of high-end servers. Hence, the ability to precisely forecast capacity needs and simultaneously improve production yields will be the key to competitiveness for all substrate suppliers.

Read more:

• HBM Supply Leader SK Hynix’s Market Share to Exceed 50% in 2023 Due to Demand for AI Servers, Says TrendForce
• Shipments of AI Servers Will Climb at CAGR of 10.8% from 2022 to 2026, Says TrendForce

(Photo Credit: Google)

As the struggle between China and the United States continues, in order to avoid upcoming geopolitical risks, not only have Taiwanese ODM manufacturers begun to shift some production locations, but market research firm TrendForce has also observed that American OEM companies have started to take action, discussing with partners how to reduce the proportion of Chinese supply chains and components.

TrendForce points out that, at present, American cloud service providers (CSPs) and OEM manufacturers have not yet been able to completely cut ties with Chinese-produced components. Among these, passive components and mechanical assemblies are more difficult to relocate due to factors such as cost and yield. However, other components (such as PCBs and power management control ICs) have plans to move out of China.

But where will these component manufacturers go if they want to move out of China? According to TrendForce’s analysis, PCB manufacturers are currently eyeing shifts to Thailand, Malaysia, Vietnam, and India; power management ICs and control ICs have already moved out of China and relocated to Taiwanese factories; mechanical assemblies and MLCC capacities still mainly come from China, with the former being requested to move but facing challenges due to cost and yield considerations.

TrendForce notes that the aforementioned production line and material shifts are primarily led by American CSPs. The overall server supply chain’s subsequent changes still need to be observed. For example, major players like Google, AWS, and Meta have not only moved most of their L6 production lines to Taiwan but also plan to establish bases in Southeast Asia after 2024 to handle cases within the United States, and reserve flexible production lines along the US-Mexico border, which will significantly increase utilization within this year.

According to TrendForce, the consumer electronics market will feel the brunt of the weakening stay-at-home economy, the pandemic in China, international tensions, and rising inflation in 1H22. Coupled with the traditional off-season, demand for relevant applications such as PCs, laptops, TVs, and smartphones has cooled significantly and downstream customers have successively downgraded their shipment targets for the year, while demand for automotive, Internet of Things, communications, and servers products remain good. At the same time, the supply chain will build higher inventories in general to mitigate the risk of material shortages due to transportation impediments induced by the spread of the pandemic and the ongoing war between Russia and Ukraine.

1. Foundries

Due to the prolonged lead-time of semiconductor equipment and limited new capacity in 1Q22, the overall foundry capacity utilization rate remains fully loaded, in particular, component mismatch issues continue for parts produced at mature nodes (1Xnm~180nm). Looking forward to the second quarter, although growth in global wafer production capacity remains limited, due to weak demand for end products, continuing international tension, and China’s forced lockdowns and supervision due to the recent spread of the pandemic, there is an opportunity for the supply chain to obtain a more adequate supply of wafers that were previously squeezed by production capacity.

2. Servers

The overall supply of key server materials improved slightly in 1Q22. In addition, due to increasing orders from ultra-large data centers, the general supply cycle of NetCom chips such as LAN IC/chip remains as long as approximately 40 weeks but the demand gap can be bridged by instituting urgent order fees, mitigating actual impact. As the aforementioned situation eases, additional orders for ODM motherboard production are moving briskly, prompting continued stocking of FPGAs and PMICs materials. NetCom chips are also overstocked and the overall market has a reached a “rich get richer” mindset. Material shortages at second-tier ODMs still stifle the production of motherboards for a small number of customers but does not affect the overall server market supply. With improvements in material supply, server shipments will increase significantly in 2Q22, growing an estimated 15.8% QoQ to 3.6 million units.

3. Smartphones

Affected by sluggish seasonal demand, the Russian-Ukrainian war, and rising inflation, market demand has cooled. Thus, material delivery issues in the supply chain have eased compared to 2H21. Although there is still a shortage of certain components, most of these shortages are concentrated in mid/low-end smartphone products. The lead time for 4G and low-end 5G SoCs is approximately 30 to 40 weeks, which is limited by production capacity planning. Since last year, the demand of the mid/low-end mobile phone market has not been met. This is followed by A+G sensors with a lead time of approximately 32~36 weeks and OLED DDIC and Touch IC with a lead time of 20~22 weeks. The production volume of smartphones in 2Q22 will be affected by the interaction of the aforementioned factors with a forecast production volume of 323 million units, or only 6% QoQ, which is lower than the performance of previous years.

4. Notebooks

Also affected by weakening end market demand, discounting client SSDs that are no longer oversupplied, Type C IC, WiFi, and PMIC all currently boast long lead times, with Type C IC the lengthiest at 20~25 weeks. However, compared with TrendForce’s assessment at the beginning of this year, the delivery cycle has not grown longer, so the lead time of these three types of products is expected to improve by the end of 2Q22. As supply chain backlog continues to improve, shipments of notebook computers (including Chromebooks) is expected to reach approximately 55.1 million units in 2Q22, down 0.7% QoQ.

5. MLCC Passive Components

From the perspective of other key components, taking MLCC as an example, demand for major consumer electronic products such as mobile phones, laptops, tablets, and TVs declined significantly in 1Q22, resulting in high consumer product specification MLCC inventory levels held by original suppliers and channel agents and this situation may continue into 2Q22. At present, the stocking momentum for automotive and industrial MLCCs has steadily increased, while consumer specification products have yet to escape the pattern of oversupply. In 2Q22, the MLCC market has the opportunity to alleviate its component mismatch issues through gradually increased production capacity and automotive and server ICs supplied by semiconductor IDM companies, driving stocking momentum at automotive power, server, fast charging, and charging/energy storage equipment OEMs. Vehicle and industrial MLCCs have the opportunity to become primary growth drivers in 2Q22 with Murata, TDK, Taiyu and Yageo as the primary beneficiaries. Consumer specification products, which account for the bulk of MLCC production from suppliers in Taiwan, South Korea, and China, may face continued market demand weakness in 2Q22 due to a slowdown in demand for mobile phones and laptops and continuing inventory adjustment by branded companies and ODMs.

Looking forward to 2Q22, not including servers, demand for end products related to the consumer category remains weak. Components that were originally oversupplied will face more severe price tests due to the imbalance between supply and demand. In terms of materials in serious short supply, more output will be transferred to products with strong demand through the deployment of internal production capacity. TrendForce believes that from the changes in PC market conditions, it can be seen in rapid changes in demand, purchasing behavior has quickly switched from the former over-ordering strategy to actively cutting orders, inducing supply chains to buck the seasonal trends of previous years. Due to the accelerated recent spread of Omicron in China and under the country’s dynamic zero-COVID policy, mandatory and sudden lockdown and control measures may cause local manufacturers to face multiple and complex supply chain problems, which will be detrimental to market performance.