NVIDIA GB300 AI Servers Expected to Ship in H2, According to Supply Chain Sources

This content should not be interpreted as investment advice. The author does not hold stocks in any of the referenced companies.

NVIDIA’s Blackwell GB300 AI Servers Set for Launch in H2 2025

NVIDIA’s latest AI offering, the Blackwell GB300 AI servers, is anticipated to commence shipments in the latter half of this year, potentially overshadowing the release of Apple’s upcoming iPhone. Market analysts are recognizing the GB300 as poised to become the most powerful AI servers globally, prompting Taiwanese contract manufacturers to compete fiercely for orders.

Foxconn, a leading player in contract manufacturing, has confirmed it secured the highest volume of orders for the GB300 servers, which include configurations featuring 72 Blackwell AI GPUs in their top-tier model. Alongside Foxconn, other notable manufacturers like Quanta, Wistron, Wiwynn, and Inventec are also gearing up for this launch.

The urgency to produce the GB300 servers is driven by their critical role in the international AI supply chain, thus placing them ahead of other tech releases, including Apple’s iPhone, according to supply chain feedback. As the primary assembler for NVIDIA, Foxconn is prepared to navigate the transition to mass production of the GB300, albeit with anticipated challenges tied to its resemblance to the previous GB200 model. Given the surge in AI server demand, Foxconn expects that these new product lines will comprise over half of its total server revenue.

Quanta’s shipments of GB300 servers are expected to begin in September, with the company noting that, despite the superior capabilities of the new servers, high-end AI firms may not delay their projects due to ongoing critical model training requirements.

Similarly, Winwynn and Inventec are anticipating shipments by September, aiming for substantial volume increases in the latter part of the year. Management teams from these firms believe this development could enable them to capture a larger share of the market.

The market is currently facing challenges, including tight supply and elevated prices, leading to heightened interest in alternatives to NVIDIA’s GPUs. These alternatives range from custom chips created by major tech firms like Amazon and Alphabet to collaborations with component designers such as Broadcom and Marvell. NVIDIA’s CEO, Jensen Huang, is confident that the performance of NVIDIA’s products stands unrivaled, downplaying the potential impact of these alternatives.

However, reports have emerged indicating that OpenAI might be diverting some of its computational needs to Google’s tensor processing units (TPUs) due to concerns about the rising costs associated with NVIDIA’s products.

For further details, please refer to the source.