Quick Thoughts: NVDA – An Inventory Glut, A China Trade War, and a Cloud CAPEX Intermission Walk into a Bar …
SEE LAST PAGE OF THIS REPORT Paul Sagawa / Tejas Raut Dessai
FOR IMPORTANT DISCLOSURES 203.901.1633 /.901.1634
psagawa@ / email@example.com
January 28, 2019
- NVDA’s pre-announcement combined with the already weak guidance given for the quarter projects 4Q19 sales $1.2B or 35% short of the November consensus, with obvious implications for margins.
- NVDA’s 3Q19 disappointment and guide down were tied to a glut of graphics card inventory in the wake of the cryptocurrency melt down. Management is now pointing to weak Chinese demand and a break in cloud spending for the further weakness.
- In Nov., we expressed concern that the inventory glut would take longer than a quarter to resolve but remained bullish on 2019 datacenter spending. This, and INTC’s guidance, suggest datacenter investment will start the year on a down note.
- Historically, chip companies have shown very poor visibility into end market demand. Given very strong uptake of cloud services and the shift to AI, and extremely lumpy spending patterns, we believe hyperscale platforms will return to aggressive CAPEX sooner rather than later.
- The future is still bright for NVDA – i.e. paradigm shift to public cloud and GPU accelerated processing, rise of automotive ADAS, etc. – but compares will be tough until 2H19.
Another big shoe dropped on Nvidia investors Monday morning. After having guided down 4Q18 revenues by $500M in its November conference call, management pre-announced a further $700M shortfall, with commensurate margin impact, as the company struggles with clearing a channel inventory glut amidst trade war impacted China demand and an apparent lull in hyperscale datacenter investment. Bang! NVDA shares are trading down almost 15%, 35% lower than the company’s value on the eve of its 3Q18 earnings announcement less than 3 months ago. At the time, we raised an eyebrow – our experience says channel inventory gluts are never resolved in a single quarter – but unfortunately decided to ride it out. Our bad.
The datacenter demand weakness is another wrinkle. While Xilinx was just fine – it has started to ride the 5G wave – Intel had already given us warning that the big cloud hyperscalers were stepping away from the table to digest. This shows Nvidia was not immune, and we cannot count on Google, Amazon and Alibaba to cover up for the bolus of unbought graphics cards sloppily over-shipped into the channel.
Still, this is not a fundamental change in the narrative for the stock. The cloud guys may be taking a break, but they will be back, and most likely, sooner rather than later. Historically, their spending has been lumpy, but this marks the first time in recorded history that cloud capex has been down sequentially for two quarters in a row (Exhibit 1).
Exh 1: Quarterly cloud capex spending trend for major cloud players, 1Q13 – 3Q18
While it looks like Intel and Nvidia have been told to expect a third drop in 1Q19 and the YoY compare for 2Q19 is crazy difficult, we still expect very strong capex for the second half – there is NO sign that the enterprise shift to the cloud or the adoption of AI is slowing down. Moreover, the long, slow adoption of Autonomous Driver Assistance Systems (ADAS) into passenger vehicles has just begun, and Nvidia stands to be a prime beneficiary. These are the reasons to hold the name.
Nonetheless, it is very unlikely that management will be able or willing to sound an all clear three months from now. Semiconductor companies have very poor visibility into end market demand, and Nvidia may be worse than most – they did willingly feed that nasty glut of graphics cards despite the “crypto-pocalypse” of the last 6 months. When demand picks up again, they will be the last to tell us. Given that, we suspect that Nvidia will trade sideways for a couple of quarters, at least until the hyperscalers pull out their checkbooks again. Our advice? Sell it on a bounce, then look to buy it back this summer.