Polysilicon prices had enlarged in reduction this week, where mono-Si compound feedings and mono-Si dense materials were concluded at a respective mainstream price of RMB 178/kg and RMB 175/kg under an average drop of roughly 7.8%. The drop of polysilicon prices had somewhat widened this week alongside the continuous release and ramp-up of polysilicon capacity, as well as the depleted procurement from crystal pulling plants.
Low-quality polysilicon continues to diverge in prices from mainstream products, and has been seeing a low level of purchases, while most businesses that are procuring frequently at small batches amidst the continuous reduction of polysilicon prices had contributed to the finalization of several new orders this week. Some businesses have started negotiating for their May orders, and prices are expected to continue fall with the arrival of the new round of order signing. An observation on the production and operation of the polysilicon segment this week indicates that two businesses are currently under overhaul. Polysilicon supply should continue to climb as businesses, including Daqo New Energy, GCL, Dongli, Runergy, and East Hope, release capacity and initiate production between May and June.
Wafer prices had slightly fluctuated this week, where M10 and G12 were concluded at a respective mainstream price of RMB 6.25/pc and RMB 8/pc. M10 mono-Si wafers had maintained a slow reduction due to insufficient demand, while G12 mono-Si wafers were largely stabilized in prices thanks to lingering support from demand.
The successive port arrival of imported quartz sand is able to fulfill full-load production among wafer businesses, though downstream cell businesses have not been aggressive in inventory pulls alongside the continuously rising level of wafer inventory, and are relatively resistant towards high-priced resources, which amplify the degree of wafer shipment on a continual basis, while second and third-tier businesses are also constantly lowering their prices in order to fight for orders.
Current mainstream concluded prices have fallen below that of prices previously announced by LONGi and Zhonghuan, and await the new round of prices that will soon be announced by leading businesses.
Cell prices were essentially stabilized this week, where M10 and G12 cells were concluded at a respective mainstream price of RMB 1.07/W and RMB 1.13/W. Supply and demand from upstream and downstream cell sectors have been relatively sturdy lately, with no significant changes to the level of overall inventory.
As demand for TOPCon cells increases, a segment of P-type production lines are currently being upgraded to N-type amidst continuous release and ramp-up of partial new capacity. The significantly risen supply of TOPCon cells could further widen the price difference between P-type and N-type cells. With upstream wafers dropping in prices and climbing in production, the cell segment has welcomed a recovery in profitability, though module makers are also amplifying in sentiment towards suppressing cell prices.
Module prices continued to maintain stability this week, where 182 and 210 mono-Si single-sided PERC modules were respectively concluded at RMB 1.67/W and RMB 1.68/W, while 182 and 210 bifacial double-glass mono-Si PERC modules sat on RMB 1.69/W and RMB 1.7/W respectively.
The continuous price drop from the upstream polysilicon and wafer segments has yet to be effectively transmitted to the module segment, and the end sector remains relatively anticipative towards dropping module prices, where some businesses of integrated production are generating orders by offering lower prices in order to bring up their shares in the domestic market. Integrated module makers are likely to carry on with their increase of demand during May, while the transmittance of price reduction from midstream and upstream sectors are also expected to further pull up end demand.
In terms of auxiliary materials, glass prices were seen with robustness this week, where 3.2mm and 2.0mm were respectively priced at RMB 26/㎡ and roughly RMB 18.5/㎡. As the new round of centralized order signing arrives soon, some module makers have been comparatively cautious on procurement by adequately digesting their previously accumulated inventory, and had thus led to a marginal drop of glass shipment this week compared to that of last week.
According to TrendForce’s research, the global production volume of smartphones in 2022 is projected to reach 1.192 billion units, a YoY decrease of 10.6%, exceeding even the decline seen in the pandemic year.
However, the market for refurbished smartphones is a completely different story. Research institutions have pointed out that Apple’s sales of refurbished smartphones have grown by 16%, and the company now holds nearly half of the refurbished phone market.
The thriving market for refurbished and second-hand smartphones has ignited demand for touch and display integration IC (TDDI) from the repair market since 2H22, and this demand is expected to double to 200 million units in 2023.
But why is the demand for refurbished and second-hand smartphones increasing year after year? There are two possible reasons based on the current overall environment:
The price of refurbished and second-hand smartphones is lower than that of new smartphones.
Most refurbished and second-hand smartphones are refurbished before being sold back to the market. Their functionality and appearance are mostly normal, and unless the user mentions it, it is difficult to tell if it is a second-hand smartphone.
When purchasing refurbished smartphones, most people prioritize high-end models with high price points, regardless of the brand. However, the entry threshold for high-end new smartphones is often high, but this type of smartphone can be obtained at a cheap price in the second-hand market.
Why iPhone is the preferred choice?
However, the support and fluidity of the operating system are often advantages of Apple, and the iOS update support period is quite long, up to 6 years. Although 6 years may not sound long, statistics show that the average device usage cycle for an iPhone user is 3 years, and a support period of 6 years is actually a very long time. Even if you take over someone else’s second-hand smartphone, you don’t have to worry too much about the operating system not being supported.
While the growth of the refurbished smartphone market is good news for consumers, how much benefit does it bring to smartphone manufacturers?
For smartphone manufacturers, most of the profits come from the sale of new smartphones. When consumers purchase second-hand smartphones, the profit margins for manufacturers are reduced. However, manufacturers can still benefit from the demand for refurbished smartphones by participating in the refurbishment process or selling refurbished smartphones themselves.
In the post-Moore’s Law era, chiplet design has been burgeoning as the mainstream architecture.
With the widespread adoption of EUV technology by foundries on process nodes of 5nm and below, the cost of semiconductor fabrication has skyrocketed. The cost of the 5nm process has grown by almost 1x compared to the 7nm process, and the 3nm process is expected to increase by almost 1x compared to the 5nm process.
To address this issue, IC design companies have started to split chip components or connect multiple chips and adopt advanced packaging such as 2.5D/3D IC to integrate multiple chips together.
Compared to traditional chip design methods, chiplet design has superior characteristics such as shorter upgrade cycles, lower costs, and higher yields, which is one of the reasons why chiplet technology is gaining popularity.
AMD’s chiplet design is a representative example. Through close collaboration with TSMC, AMD has fully transitioned its CPUs to chiplets since the 7nm process, with the Ryzen 7000 series CPU and Radeon RX 7000 series graphics cards released in 2022. The latter uses the RDNA 3 architecture and integrates the GCD and MCD produced by the 5nm and 6nm processes respectively, as a result improving overall performance, with a 54% increase in RDNA 3’s Performance per Watt.
Under the leadership of industry leaders such as AMD and Intel, chiplet design has had a significant impact on the entire semiconductor industry – substrates manufacturers in particular.
ABF Substrates Set to Soar
Aside from CPUs, developments in AMD and Intel’s server platforms indicate that the trend towards higher-layer-count and larger-area ABF substrates is expected to continue.
Given the server shipment volume is expected to remain stable and grow steadily in the mid to low single digits for the next 3-5 years, the growth momentum of ABF substrates mainly comes from the increase in layer count and area brought by 2.5D/3D packaging adoption in servers.
Starting in 2020, ABF substrates saw a surge in demand due to the pandemic. The supply-demand gap peaked in 2021, and in the first half of 2022, ABF substrate prices increased while volume increased and gross profit margins hit new highs.
Due to the impact of shortage in ABF substrates in 2020-2021, major substrate manufacturers have initiated large-scale expansion plans, with the expectation that demand for ABF substrates would continue to grow with the upcoming releases of new server platforms and the integration of 2.5D packaging for PC CPUs.
Growing demands with Some Hiccups
However, the moves have been put on hold for now. Since the second half of 2022, due to inventory correction in the overall semiconductor industry and the delayed production time of Intel’s new server platform, there’s been a supply glut in ABF substrates.
Therefore, Unimicron has taken the lead in adjusting its capital expenditure plans, reducing its planned capacity increase for 2023 from about 20% to only 3.5%. AT&S has also tentatively postponed the significant increase in capacity planned for the end of 2024. It is unclear when the expansion will resume or whether the expansion will be scaled back.
This indicates that current substrate manufacturers have not only lowered their demand projections for 2023, but also for 2025-2026. Further adjustments to the expansion plans of other manufacturers will also affect the future market supply-demands dynamics.
Back on Track for Major Growth in 2024
Looking into the future, things are looking up for the ABF substrate industry. In the second quarter of 2023, we can expect the release of new server platforms from AMD and Intel, as well as the completion of PC inventory adjustments.
With expansion plans in place, it’s predicted that global ABF substrate production capacity will only increase by 15-20% in the latter half of 2023, continuing to put pressure on substrate manufacturers, according to TrendForce.
Things are expected to pick up in 2024 with the release of AMD and Intel’s next-generation server platforms, Zen 5 and Birch Stream. Plus, the anticipated introduction of 2.5D packaging for PC CPUs will drive a new wave of demand for ABF substrates. All in all, we can expect a significant rebound for the ABF substrate industry in 2024.
Sodium-ion batteries are burgeoning as a popular alternative to lithium-ion batteries, thanks to the efforts of Chinese automakers who are pushing for its mainstream adoption.
Leading Chinese companies like CATL and BYD are ramping up the production of sodium-ion batteries. In mid-April, CATL and Chery unveiled their new battery brand, “ENER-Q”, which includes full product lines including sodium-ion, iron phosphate lithium, and ternary lithium batteries. Chery’s new energy vehicles will be the first to use CATL’s sodium-ion batteries.
Following CATL, BYD is rumored to start mass production of its sodium-ion batteries in the second half of this year, which will be used in its compact hatchback, the Seagull series. Both the moves have once again sparked discussions about battery technology in the market.
Geopolitical risks fuels Sodium-ion Batteries
Considering market supply and technical stability, lithium-ion batteries and iron phosphate lithium batteries are still the most popular types of batteries for electric vehicles. The former has a higher energy density but contains cobalt and nickel, which drives up costs. The latter has a lower cost but a lower energy density.
Sodium-ion batteries, on the other hand, have been overlooked due to their low energy density compared to traditional lithium-ion batteries.
So, why are companies like CATL and BYD turning to sodium-ion batteries?
Geopolitical risk is a major factor. Most lithium mines are located in countries like the US, Australia, and Canada. In today’s anti-China political climate, these materials could be used as bargaining chips to curb China’s electric vehicle industry. China won’t want to be at the mercy of other countries when it comes to the fate of its EV industry, so developing new technological routes is crucial.
From a mass production perspective, sodium is a more abundant element in the Earth’s crust than nickel, cobalt, or lithium carbonate, with a distribution that’s more evenly spread out. As such, sodium could be a better fit as a positive electrode material in batteries in the long run. Industry experts predict that sodium-ion batteries could even cost 20% less than iron phosphate lithium batteries once it reaches economies of scale.
The Supporting Actor in EV Batteries
However, a closer look into the pros and cons of both the materials may reveal that it’s not a zero-sum game. Instead, their characteristics can complement each other and help to accelerate battery technology development.
CATL’s new sodium-ion battery has an energy density of up to 160Wh/kg, which is comparable to the iron phosphate lithium battery in its Kirin battery system, but still lags behind the 255Wh/kg of ternary lithium batteries.
As a result, CATL is mixing sodium-ion and ternary lithium batteries in Chery’s new energy vehicles to balance cost and performance.
BYD is also expected to use a mix of sodium-ion and iron phosphate lithium batteries. Assuming this is true, it will echo the market’s assumption that sodium-ion batteries are not overturning the battery industry, but rather helping battery manufacturers maintain flexible product portfolios that cater to different market segmentations.
To give an example, CATL’s lithium iron phosphate batteries have been utilized in heavy-duty vehicles like 120-ton ore trucks and marine service vessels since 2022, where charging efficiency and cost take precedence over high energy density.
Therefore, sodium-ion batteries are likely to become a complimentary choice for lithium iron phosphate batteries, as they offer advantages such as high-rate charging, low cost, and high safety. This will definitely give car makers more flexibility in their future product strategies.
OpenAI’s ChapGPT, Microsoft’s Copilot, Google’s Bard, and latest Elon Musk’s TruthGPT – what will be the next buzzword for AI? In just under six months, the AI competition has heated up, stirring up ripples in the once-calm AI server market, as AI-generated content (AIGC) models take center stage.
The convenience unprecedentedly brought by AIGC has attracted a massive number of users, with OpenAI’s mainstream model, GPT-3, receiving up to 25 million daily visits, often resulting in server overload and disconnection issues.
Given the evolution of these models has led to an increase in training parameters and data volume, making computational power even more scarce, OpenAI has reluctantly adopted measures such as paid access and traffic restriction to stabilize the server load.
High-end Cloud Computing is gaining momentum
According to Trendforce, AI servers currently have a merely 1% penetration rate in global data centers, which is far from sufficient to cope with the surge in data demand from the usage side. Therefore, besides optimizing software to reduce computational load, increasing the number of high-end AI servers in hardware will be another crucial solution.
Take GPT-3 for instance. The model requires at least 4,750 AI servers with 8 GPUs for each, and every similarly large language model like ChatGPT will need 3,125 to 5,000 units. Considering ChapGPT and Microsoft’s other applications as a whole, the need for AI servers is estimated to reach some 25,000 units in order to meet the basic computing power.
As the emerging applications of AIGC and its vast commercial potential have both revealed the technical roadmap moving forward, it also shed light on the bottlenecks in the supply chain.
The down-to-earth problem: cost
Compared to general-purpose servers that use CPUs as their main computational power, AI servers heavily rely on GPUs, and DGX A100 and H100, with computational performance up to 5 PetaFLOPS, serve as primary AI server computing power. Given that GPU costs account for over 70% of server costs, the increase in the adoption of high-end GPUs has made the architecture more expansive.
Moreover, a significant amount of data transmission occurs during the operation, which drives up the demand for DDR5 and High Bandwidth Memory (HBM). The high power consumption generated during operation also promotes the upgrade of components such as PCBs and cooling systems, which further raises the overall cost.
Not to mention the technical hurdles posed by the complex design architecture – for example, a new approach for heterogeneous computing architecture is urgently required to enhance the overall computing efficiency.
The high cost and complexity of AI servers has inevitably limited their development to only large manufacturers. Two leading companies, HPE and Dell, have taken different strategies to enter the market:
With the booming market for AIGC applications, we seem to be one step closer to a future metaverse centered around fully virtualized content. However, it remains unclear whether the hardware infrastructure can keep up with the surge in demand. This persistent challenge will continue to test the capabilities of cloud server manufacturers to balance cost and performance.
(Photo credit: Google)