Broadcom beats estimates: Artificial Intelligence drives accounts
The company is the eighth largest in the world by capitalisation and a crucial group for the development of the infrastructure underpinning artificial intelligence
3' min read
3' min read
On the one hand, results for the third quarter of the fiscal year 2024- 2025 that exceeded analysts' expectations. On the other hand, solid guidance for the current quarter, but which - at first - did not seem to satisfy everyone. This is how the context of Broadcom's third quarter data can be summarised. A scenario that -at first- induced a lukewarm reaction from the market.
Specifically, earnings per share (EPS) came in at $1.69 on an adjusted basis, compared to $1.65 expected. Revenues, for their part, stood at $15.96 billion, compared to the estimated $15.83 billion. The company, on the other hand, expects revenues of $17.4 billion in the fourth quarter. The figure - according to some agencies - is higher than the 17.02 billion estimated by Wall Street analysts. According to other surveys - on the contrary - it would be lower. In short: a mix of conflicting visions, which initially provoked a lukewarm reaction from traders.
The group's president and CEO, however, indicated that the growth 'is related to the continued momentum of customised Ai accelerators, networking, and VMware'. Above all, 'Artificial Intelligence-related revenue rose 63 per cent year-on-year to $5.2 billion in the quarter,' Hock Tan added. The manager then -during the conference call- emphasised that the outlook for AI-related revenues will improve in 2025-2026. The statement sent the stock into high gear.
Behind the showcase of quarterly results is an articulated strategy
.That said, it must be emphasised that Broadcom does not compete on the consumer front with chatbots or visible applications, but builds the hidden muscles that make those services possible. In data centres hosting ChatGPT, Copilot or Gemini, its chips and networks are crucial components. The company has chosen to focus strongly on infrastructural artificial intelligence, designing custom chips for hyperscalers, the cloud giants that handle the most massive computing loads. These are ASICs and XPUs, tailor-made accelerators that offer ultra-high performance and lower power consumption than general-purpose GPUs, thanks to a design targeted on the specific needs of those who have to train gigantic neural networks. In parallel, Broadcom develops networking solutions such as Jericho3-AI and optical technologies capable of connecting thousands of computing units, breaking down bottlenecks in AI clusters.
The second pillar of the strategy is expansion into infrastructure software. The acquisition of VMware, completed in 2023, transformed Broadcom into a hybrid chip-software group, with revenues now almost equally divided between the two segments. Software has grown to weigh about 42% of revenues, up from 22% pre-acquisition, and offers stable recurring streams through subscriptions and long-term enterprise contracts. The company has chosen to focus on around 600 mission-critical customers, mainly large regulated entities, betting on their loyalty and the difficulty of replacing solutions that are considered critical.
