- ASIC servers will hold 26% share of total AI server market in 2024, while GPU-equipped AI servers will occupy approximately 71% of the market.
- NVIDIA maintains a dominant position as leading AI chip supplier for GPU-equipped AI servers, commanding nearly 90% market share in this segment.
In 2024, the demand for advanced AI servers from major cloud service providers (CSPs) and brand clients is anticipated to remain high.
Research conducted by TrendForce indicates that major CSPs are prioritising the procurement of AI servers over general servers, leading to a noteworthy growth disparity between the two segments.
While general server shipments are experiencing a modest annual growth rate of 1.9 per cent, AI server shipments are surging, with an expected share of 12.2 per cent in total server shipments, marking a 3.4 percentage point increase from the previous year.
Significantly, the expansion of production by key players such as TSMC, SK hynix, Samsung, and Micron has alleviated shortages in the second quarter of 2024.
The improvement is evidenced by the reduced lead time for NVIDIA’s flagship H100 solution, dropping from 40–50 weeks to under 16 weeks.
TrendForce forecasts a substantial 20 per cent quarter-on-quarter increase in AI server shipments for the second quarter, revising the annual shipment forecast up to 1.67 million units, reflecting a remarkable 41.5 per cent year-on-year growth.
Prominence of ASIC servers
In terms of market value, AI servers are making a substantial contribution to revenue growth compared to general servers. The projected market value of AI servers in 2024 is expected to surpass $187 billion, with a growth rate of 69 per cent, constituting 65 per cent of the total server market value.
The prominence of ASIC servers is on the rise, with North American CSPs and Chinese companies amplifying their respective ASIC AI solutions. It is anticipated that ASIC servers will hold a 26 per cent share of the total AI server market in 2024, while GPU-equipped AI servers will occupy approximately 71 per cent of the market.
NVIDIA maintains a dominant position as the leading AI chip supplier for GPU-equipped AI servers, commanding nearly 90 per cent market share in this segment.
Evolution in chip specifications
In the broader scope encompassing all AI chips utilised in AI servers (GPU, ASIC, FPGA), NVIDIA’s overall market share stands at around 64 per cent in 2024, with AMD trailing at an 8 per cent market share.
The sustained demand for advanced AI servers is expected to persist into 2025, particularly with NVIDIA’s upcoming Blackwell generation poised to supersede the Hopper platform, driving the need for CoWoS and HBM. The expansion in production capacity of major supplier TSMC’s CoWoS to 550–600K units by the end of 2025 signals an imminent 80 per cent growth rate.
The evolution in chip specifications is noteworthy, as the main chips like NVIDIA’s Blackwell Ultra and AMD’s MI350 are projected to feature up to 288GB of HBM3e by 2025, tripling the current unit usage.
The anticipated surge in demand for HBM underscores the robust growth projections for the AI server market, with overall HBM supply expected to double by 2025 to meet the escalating requirements.