GP Bullhound's weekly review of the latest news in public markets.

This week's GP Bullhound Tech Thoughts Newsletter, brought to you by Inge Heydorn and Ofelia Aspemyr, highlights Nvidia's AI roadmap, Qualcomm and AMD's AI PC solutions, and product launches from Intel and AMD.

Market: Economic indicators continue to present a mixed picture, with US numbers on the weak side. Volatility in single equity names has increased lately, especially around results. On the tech side, hardware exposed to AI continues to perform well, while software struggles, leading to several forecast downgrades from companies. Prolonged decision-making, reduced seat growth, and tougher financing contribute to this trend. However, TSMC reported strong May figures earlier today, with sales coming in at NT$229.62 billion (30.1% Y/Y).

Portfolio: We have added to our positions in TSMC and AMD this week.

Weekly Conclusion: Hardware with AI exposure, especially Nvidia, is performing well, while software has started the year weakly, and we wouldn't be surprised if this trend continues for a while. Many software companies are waiting for AI demand to lift their sales forecasts.

Anonymous Quote of the Week: “Never seen such a big crowd at Computex, and the mix of overseas visitors is a lot more; this event has gone from a Taiwan-centric PC event to an international one.”

Results:

HPE (not owned): Reported sales of USD 7.3bn, a clear beat, with an EPS of USD 0.42, 3 cents ahead. Topline growth was driven by strong server demand, especially AI servers, and a recovery in traditional server sales. However, gross margin fell 310 basis points Q/Q and Y/Y due to weaker Intelligent Edge sales and higher AI server sales. Intelligent Edge sales fell 9% Q/Q and 19% Y/Y. AI system sales totaled USD 900m in the quarter, up more than 100% Q/Q, aided by improved supply chain conversion, notably Nvidia H100 supply. Lead times for H100 solutions are now down to between 6 and 12 weeks. Enterprise AI sales are ramping up fast and now constitute 15% of total AI sales. The order backlog declined slightly, but the company states it's lumpy and that the pipeline is much stronger than the backlog. HPE is raising its sales forecast for the second half of the year on the back of strong AI sales and a recovery in Intelligent Edge sales. Our view is that HPE trends are similar to those seen at Dell: strong sales and pipeline but with margin contraction due to a tough pricing environment. HPE is notably more focused on selling Nvidia solutions than its counterparts.

Foxconn Technology (not owned): Reported May sales growth of 22% Y/Y and 7.68% M/M. The company has guided for both Q/Q and Y/Y growth in 2Q, now stating that AI server sales will drive better-than-expected sales. AI server sales, cloud, and networking product sales were up strongly M/M, and computing sales were also up strongly in May due to new products. Components and smart consumer electronics were flat M/M. Our view: Another sign of strength for Nvidia as they are the main product for Foxconn's AI server business.

UMC (not owned): Reported May sales flat M/M but up 3.89% Y/Y, indicating a more stable environment for mature production nodes.

Computex:

Nvidia (owned): Hosted its own presentation at a university ahead of the fair, providing a glimpse of their roadmap. Nvidia plans to roll out new AI accelerators annually, increasing the pace of development. In 2025, they will roll out Blackwell Ultra, and in 2026, the Rubin platform will be introduced, featuring HBM4 memory solutions. The company provided limited data on upcoming models but signaled a fast pace of innovation.

AMD (owned): CEO Lisa Su released several new products at the event. On the PC side, the company released the desktop processor Ryzen 9000 and the notebook AI PC solution AI300, which integrates a Zen CPU core, GPU core, and NPU, with a TOPS performance of 50, surpassing Microsoft's AI requirements of 40 TOPS. The AI300 will compete with Qualcomm's X Elite. On the data center side, AMD released the 5th generation EPYC processor Turin, set to launch in 2H 2024 and produced by TSMC 3nm processes. AMD has gained market share from Intel over the last 12 months. AMD also unveiled its AI accelerators roadmap, with new releases planned annually. The MI325X will be released by the end of this year, featuring up to 288GB of HBM3E memory with 6 TB/s of memory bandwidth. In 2025, the MI350 based on CDNA 4 architecture and 3nm with HBM3E memory will be released, followed by the MI400 in 2026 based on the new CDNA Next architecture.

Intel (not owned): Released two main products at the event: the Sierra Forest Xeon CPU for data centers and the Lunar Lake PC CPU. The Xeon will be produced using Intel 3 process, which must run smoothly to compete with AMD, which has gained market share over the last year. Lunar Lake, Intel's AI PC processor, will be produced by TSMC at 4nm and 6nm (improved 5nm and 7nm). Pat Gelsinger also discussed their AI GPU Gaudi 3, which will hit the market in 3Q and sell for around USD 125,000, significantly cheaper than Nvidia and AMD solutions. Intel also announced a partnership with Apollo Global on an USD 11bn deal for its Fab 34 plant in Ireland, with Intel holding a 51% majority stake and Apollo holding 49%.

Qualcomm (not owned): Focused their presentation on Snapdragon X Elite, which leads in power efficiency over Intel and AMD. X Elite with Copilot will enhance AI capabilities and user interaction with PCs. Qualcomm has support from leading OEMs like Acer, Asus, Dell, HP, Lenovo, Microsoft, and Samsung. Our view: Qualcomm appears to have momentum with its product, presenting a new competitor for Intel and AMD.

ARM (not owned): CEO Rene Haas boldly stated that ARM processors will hold more than 50% of the Windows PC market share by 2029, adding to the market share they already have through Apple.

ASML (owned): CFO, quoted by Bloomberg, stated that all three major clients will receive a High NA EUV tool in 2024, with the first tool shipped to Intel at the beginning of the year. Our view: This is not new information, but it is positive in the long term as it allows clients to start test production on the tool and prepare for sub-1.8nm production in a few years.

For enquiries, please contact: Inge Heydorn, Partner, at [email protected] and Ofelia Aspemyr, Associate, at [email protected]

About GP Bullhound: GP Bullhound is a leading technology advisory and investment firm, providing transaction advice and capital to the world’s best entrepreneurs and founders. Founded in 1999 in London and Menlo Park, the firm today has 12 offices spanning Europe, the US and Asia. For more information, visit www.gpbullhound.com