TL;DR:
- Dell and Lenovo are lagging behind in AI software development and GPU allocations.
- Dell’s Q3 results show a decline in overall sales but a substantial increase in net income.
- Lenovo faces challenges with declining revenue in its data center business, despite growth in traditional HPC.
- Both companies are impacted by limited GPU availability, hindering their AI server growth prospects.
- The AI server market is witnessing strong demand, but supply constraints, especially in high-end GPUs, are holding OEMs back.
Main AI News:
In the ever-evolving landscape of enterprise technology, original equipment manufacturers (OEMs) have often found themselves at a crossroads. Two decades ago, many OEMs attempted to build formidable software divisions in emulation of IBM, only to change course a decade ago by divesting their software and services businesses. Consequently, they focused their efforts on hardware, relegating the software front. This strategic choice has placed companies like Dell and Lenovo a step behind in the race to lead AI software development and secure GPU allocations from industry leaders Nvidia and AMD.
However, they remain poised to catch the transformative wave of generative AI technology as it sweeps into the world of enterprise data centers. Hewlett Packard, although already making strides with hybrid HPC/AI supercomputers powered by AMD and Nvidia GPUs, still faces hurdles in acquiring sufficient GPU allocations to meet the burgeoning demand for AI training and inference clusters among enterprises.
As we recently attended the AMD MI300 GPU launch event, we couldn’t help but contemplate the financial performance of Dell and Lenovo. Our goal was to gain insights into their data center businesses, incorporating them into a broader model to comprehend the dynamics of the global data center landscape.
Dell reported $22.25 billion in sales for the fiscal third quarter ending in October, a 10% decline from the previous year. However, net income soared by 4.1 times, reaching just over $1 billion. Dell achieved this feat by paying off debts and reducing interest costs by $1 billion while simultaneously cutting costs to offset the impact of higher parts and manufacturing expenses.
Of the total revenue, $12.28 billion came from client devices, a segment that, while not directly relevant to our analysis, significantly enhances Dell’s supply chain leverage against its parts suppliers. Furthermore, if client devices contribute to profits, it alleviates some of the pressure on Dell’s data center business. This business segment has come under strain as enterprises, both large and small, have curtailed server and storage spending to allocate resources to AI development and platforms.
Within Dell’s Infrastructure Solutions Group, the situation is less optimistic, with a consistent 11.7% year-on-year decline in revenue. Both Servers & Networking and Storage divisions have been in a recession for three consecutive quarters. While this is not the longest streak of declines for Servers & Networking, it signifies ongoing challenges for Dell, especially since it began facing competition from Inspur and Lenovo even before the onset of the COVID-19 pandemic.
In the latest quarter, Dell’s Server & Networking division reported sales of $4.66 billion, down 10.5%, while its Storage division recorded sales of $3.84 billion, down 13.2%. When operating costs are factored in, the Infrastructure Solutions Group had an operating income of $1.07 billion, down 22.2% and comprising 12.6% of revenues.
Jeff Clarke, Dell’s Chief Operating Officer, noted that demand for traditional servers had improved, and AI server demand remained robust. However, AI-optimized systems contributed just over $500 million in sales, highlighting the disconnect between demand and supply. The main bottleneck remains the limited availability of high-end GPUs from Nvidia and AMD for OEMs like Hewlett Packard Enterprise, Dell, and Lenovo, in contrast to hyperscalers, cloud builders, and select customers like Tesla.
Despite AI servers accounting for a third of Dell’s orders in fiscal Q3, revenue growth has been constrained by supply limitations. Clarke cited the PowerEdge XE9680 as the fastest-ramping product in Dell’s history, but it is also the most expensive. Still, the constrained availability of top-end GPUs has impeded growth across the server businesses of Dell and other OEMs.
Lenovo’s story closely mirrors that of HPE and Dell, with unique twists attributable to its significant presence in HPC and AI across the United States, Europe, and China. In the second quarter of fiscal 2024, ending in September, Lenovo reported sales of $14.41 billion, down 15.7% year-on-year, with net income of $249 million, a 54% decline. The Infrastructure Solutions Group at Lenovo, responsible for its server, storage, and networking businesses in the data center, faced its second consecutive quarter of revenue decline, registering slightly over $2 billion in revenue, down 23.4% year-on-year but up 4.6% sequentially.
Lenovo, like its peers, grapples with challenges in both the PC and server markets and appears to be affected by companies conserving budgets for AI projects. Despite these obstacles, Lenovo’s traditional HPC business continued to grow in double digits year-on-year, marking five consecutive quarters of growth. Additionally, its storage revenue reached an all-time high, solidifying Lenovo’s position as the third-largest storage vendor globally and the leader in low-end storage devices priced at $25,000 or less.
While Lenovo did not disclose its AI server pipeline, Kirk Skaugen, President of ISG, revealed that Lenovo had allocated $1.2 billion to HPC and AI systems, with an additional $1 billion earmarked for research and development. This investment underscores Lenovo’s commitment to advancing AI solutions for data centers, edge computing, and cloud customers over the next three years.
Conclusion:
The challenges faced by Dell and Lenovo in obtaining GPU allocations for AI servers indicate the ongoing supply-demand imbalance in the AI server market. Despite robust demand, the inability to secure sufficient GPUs hampers growth prospects for these OEMs. This underscores the critical need for expanded GPU production and distribution to support the burgeoning AI market.
