Major Cloud Service Providers (CSPs) continue to see an increase in demand for AI servers over the next two years. The latest projections of TrendForce indicate a global shipment of approximately 1.18 million AI servers in 2023, with a year-on-year growth of 34.5%. The trend is expected to persist into the following year, with an estimated annual growth of around 40.2%, constituting over 12% of the total server shipments.

NVIDIA, with its key products including AI-accelerating GPU and the AI server reference architecture HGX, currently holds the highest market share in the AI sector. However, it is crucial to monitor CSPs developing their own chips and, in the case of Chinese companies restricted by U.S. sanctions, expanding investments in self-developed ASICs and general-purpose AI chips.

According to TrendForce data, AI servers equipped with NVIDIA GPUs accounted for approximately 65.1% this year, projected to decrease to 63.5% next year. In contrast, servers featuring AMD and CSP self-developed chips are expected to increase to 8.2% and 25.4%, respectively, in the coming year.

Another critical application, HBM (High Bandwidth Memory), is primarily supplied by major vendors Samsung, SK Hynix, and Micron, with market shares of approximately 47.5%, 47.5%, and 5.0%, respectively, this year. As the price difference between HBM and DDR4/DDR5 is 5 to 8 times, this is expected to contribute to a staggering 172% year-on-year revenue growth in the HBM market in 2024.

Currently, the three major manufacturers are expected to complete HBM3e verification in the first quarter of 2024. However, the results of each manufacturer’s HBM3e verification will determine the final allocation of procurement weight for NVIDIA among HBM suppliers in 2024. As the verifications are still underway, the market share for HBM in 2024 remain to be observed.

Read more

• [News] Memory Titans Vie for Control in HBM Tech, Who Will Shape the Next-Gen?
• [Insights] NVIDIA Maintains Confidence Despite Export Restrictions, Continues Unveiling New Products

(Photo credit: NVIDIA)

On the 28th, Amazon unveiled two AWS-designed chips, Graviton4, a CPU propelling its AWS cloud services, and the second-gen AI chip Trainium2, tailored for large language models. Both chips boast substantial performance upgrades. With a positive market outlook, Amazon is intensifying its competition with Microsoft and Google for dominance in the AI cloud market. The demand for in-house chips is surging, leading to increased orders for key players like the wafer foundry TSMC and the silicon design and production services company ALCHIP, reported by UDN News.

According to reports, Amazon AWS CEO Adam Selipsky presented the fourth AWS-Designed custom CPU chip, Graviton4, at the AWS re:Invent 2023 in Las Vegas. It claims a 30% improvement in computing performance compared to the current Graviton3, with a 75% increase in memory bandwidth. Computers equipped with this processor are slated to go live in the coming months.

Trainium2, the second-gen chip for AI system training, boasts a computing speed three times faster than its predecessor and doubled energy efficiency. Selipsky announced that AWS will commence offering this new training chip next year.

AWS is accelerating the development of chips, maintaining its lead over Microsoft Azure and Google Cloud platforms. Amazon reports that over 50,000 AWS customers are currently utilizing Graviton chips.

Notably, Amazon’s in-house chip development heavily relies on the Taiwan supply chain, TSMC and ALchip. To produce Amazon’s chips, Alchip primarily provides application-specific integrated circuit (ASIC) design services, and TSMC manufactures with advanced processes.

TSMC consistently refrains from commenting on products for individual customers. Analysts estimate that TSMC has recently indirectly secured numerous orders from Cloud Service Providers (CSPs), mainly through ASIC design service providers assisting CSP giants in launching new in-house AI chips. This is expected to significantly contribute to TSMC’s high utilization for the 5nm family.

In recent years, TSMC has introduced successive technologies such as N4, N4P, N4X, and N5A to strengthen its 5nm family. The N4P, announced at 2023 Technology Symposium, is projected to drive increased demand from 2024 onwards. The expected uptick in demand is mainly attributed to AI, network, and automotive products.

(Image: Amazon)

On the 15th, Microsoft introducing its first in-house AI chip, “Maia.” This move signifies the entry of the world’s second-largest cloud service provider (CSP) into the domain of self-developed AI chips. Concurrently, Microsoft introduced the cloud computing processor “Cobalt,” set to be deployed alongside Maia in selected Microsoft data centers early next year. Both cutting-edge chips are produced using TSMC’s advanced 5nm process, as reported by UDN News.

Amidst the global AI fervor, the trend of CSPs developing their own AI chips has gained momentum. Key players like Amazon, Google, and Meta have already ventured into this territory. Microsoft, positioned as the second-largest CSP globally, joined the league on the 15th, unveiling its inaugural self-developed AI chip, Maia, at the annual Ignite developer conference.

These AI chips developed by CSPs are not intended for external sale; rather, they are exclusively reserved for in-house use. However, given the commanding presence of the top four CSPs in the global market, a significant business opportunity unfolds. Market analysts anticipate that, with the exception of Google—aligned with Samsung for chip production—other major CSPs will likely turn to TSMC for the production of their AI self-developed chips.

TSMC maintains its consistent policy of not commenting on specific customer products and order details.

TSMC’s recent earnings call disclosed that 5nm process shipments constituted 37% of Q3 shipments this year, making the most substantial contribution. Having first 5nm plant mass production in 2020, TSMC has introduced various technologies such as N4, N4P, N4X, and N5A in recent years, continually reinforcing its 5nm family capabilities.

Maia is tailored for processing extensive language models. According to Microsoft, it initially serves the company’s services such as $30 per month AI assistant, “Copilot,” which offers Azure cloud customers a customizable alternative to Nvidia chips.

Borkar, Corporate VP, Azure Hardware Systems & Infrastructure at Microsoft, revealed that Microsoft has been testing the Maia chip in Bing search engine and Office AI products. Notably, Microsoft has been relying on Nvidia chips for training GPT models in collaboration with OpenAI, and Maia is currently undergoing testing.

Gulia, Executive VP of Microsoft Cloud and AI Group, emphasized that starting next year, Microsoft customers using Bing, Microsoft 365, and Azure OpenAI services will witness the performance capabilities of Maia.

While actively advancing its in-house AI chip development, Microsoft underscores its commitment to offering cloud services to Azure customers utilizing the latest flagship chips from Nvidia and AMD, sustaining existing collaborations.

Regarding the cloud computing processor Cobalt, adopting the Arm architecture with 128 core chip, it boasts capabilities comparable to Intel and AMD. Developed with chip designs from devices like smartphones for enhanced energy efficiency, Cobalt aims to challenge major cloud competitors, including Amazon.
(Image: Microsoft)

On October 17th, the U.S. Department of Commerce announced an expansion of export control, tightening further restrictions. In addition to the previously restricted products like NVIDIA A100, H100, and AMD MI200 series, the updated measures now include a broader range, encompassing NVIDA A800, H800, L40S, L40, L42, AMD MI300 series, Intel Gaudi 2/3, and more, hindering their import into China. This move is expected to hasten the adoption of domestically developed chips by Chinese communications service providers (CSPs).

TrendForce’s Insights:

1. Chinese CSPs Strategically Invest in Both In-House Chip Development and Related Companies

In terms of the in-house chip development strategy of Chinese CSPs, Baidu announced the completion of tape out for the first generation Kunlun Chip in 2019, utilizing the XPU. It entered mass production in early 2020, with the second generation in production by 2021, boasting a 2-3 times performance improvement. The third generation is expected to be released in 2024. Aside from independent R&D, Baidu has invested in related companies like Nebula-Matrix, Phytium, Smartnvy, and. In March 2021, Baidu also established Kunlunxin through the split of its AI chip business.

Alibaba, in April 2018, fully acquired Chinese CPU IP supplier C-Sky and established T-head semiconductor in September of the same year. Their first self-developed chip, Hanguang 800, was launched in September 2020. Alibaba also invested in Chinese memory giant CXMT, AI IC design companies Vastaitech, Cambricon and others.

Tencent initially adopted an investment strategy, investing in Chinese AI chip company Enflame Tech in 2018. In 2020, it established Tencent Cloud and Smart Industries Group(CSIG), focusing on IC design and R&D. In November 2021, Tencent introduced AI inference chip Zixiao, utilizing 2.5D packaging for image and video processing, natural language processing, and search recommendation.

Huawei’s Hisilicon unveiled Ascend 910 in August 2019, accompanied by the AI open-source computing framework MindSpore. However, due to being included in the U.S. entity list, Ascend 910 faced production restrictions. In August 2023, iFLYTEK, a Chinese tech company, jointly introduced the “StarDesk AI Workstation” with Huawei, featuring the new AI chip Ascend 910B. This is likely manufactured using SMIC’s N+2 process, signifying Huawei’s return to self-developed AI chips.

2. Some Chinese Companies Turn to Purchasing Huawei’s Ascend 910B, Yet It Lags Behind A800

Huawei’s AI chips are not solely for internal use but are also sold to other Chinese companies. Baidu reportedly ordered 1,600 Ascend 910B chips from Huawei in August, valued at approximately 450 million RMB, to be used in 200 Baidu data center servers. The delivery is expected to be completed by the end of 2023, with over 60% of orders delivered as of October. This indicates Huawei’s capability to sell AI chips to other Chinese companies.

Huawei’s Ascend 910B, expected to be released in the second half of 2024, boasts hardware figures comparable to NVIDIA A800. According to tests conducted by Chinese companies, its performance is around 80% of A800. However, in terms of software ecosystem, Huawei still faces a significant gap compared to NVIDIA.

Overall, using Ascend 910B for AI training may be less efficient than A800. Yet with the tightening U.S. policies, Chinese companies are compelled to turn to Ascend 910B. As user adoption increases, Huawei’s ecosystem is expected to improve gradually, leading more Chinese companies to adopt its AI chips. Nevertheless, this will be a protracted process.

As Jiwei reported, AMD, although trailing NVIDIA in AI, has recently clinched significant deals, earning the trust of two major clients, Oracle and IBM. Oracle plans to integrate AMD’s Instinct MI300X AI chips into their cloud services, complemented by HPC GPUs. Additionally, as per insights from Ming-Chi Kuo, TF International Securities analyst, IBM is set to leverage AMD’s Xilinx FPGA solutions to handle artificial intelligence workloads.

Oracle’s extensive cloud computing infrastructure faces challenges due to a shortage of NVIDIA GPUs. Nonetheless, Oracle maintains an optimistic outlook. They aim to expand the deployment of the H100 chip by 2024 while considering AMD’s Instinct MI300X as a viable alternative. Oracle has decided to postpone the application of their in-house chips, a project with a multi-year timeline. Instead, they are shifting their focus to AMD’s high-performance AI chip, the MI300X, well-regarded for its impressive capabilities.

Reports indicate that Oracle intends to introduce these processor chips into their infrastructure in early 2024.

Similarly, IBM is exploring chip options beyond NVIDIA. Their new AI inference platform relies on NeuReality’s NR1 chip, manufactured on TSMC’s 7nm process. AMD plays a pivotal role in NeuReality’s AI solution by providing the essential FPGA chips. Foxconn is gearing up for AI server production using this technology in the Q4 2023.

Guo also pointed out that, although Nvidia remains the dominant AI chip manufacturer in 2024, AMD strengthens partnerships with platform service providers/CSPs like Microsoft and Amazon while acquiring companies like Nod.ai. This positions AMD to potentially narrow the AI gap with Nvidia starting in 2025. This collaboration also affirms that AMD remains unaffected by the updated U.S. ban on shipping AI chips to China.

(Image: AMD)