Investors will be looking to Nvidia's management team for a lot of answers later this week, both in the short term and about where the semiconductor company will be in more than a year.
In the short term, questions are being asked about the company's new Blackwell chip family, especially given reports that initial shipments may be slightly delayed. And before Blackwell starts shipping, Wall Street is still speculating whether the anticipation for the product is affecting demand for Nvidia's current Hopper family.
Longer term, investors will want to understand the potential of Blackwell once it is in full production, as well as the product roadmap beyond Blackwell.
Here are the five most important takeaways from Nvidia’s second-quarter earnings report on Wednesday:
How strong is the demand and supply for Hopper?
Investors have been concerned for some time about a possible "demand vacuum" for the Hopper, as some customers may buy fewer Hoppers as they wait for the more powerful Blackwell series.
Last quarter, Chief Executive Officer Jensen Huang successfully allayed concerns about a demand vacuum, and Evercore ISI’s C.J. Muse expects he may do the same this quarter.
Muse continued: “We hope he can paint a picture of continued data center growth over the next 12 months, which will fill the gap before entering full Blackwell production.”
Rosenblatt's Hans Mosesmann will be watching the supply of Hopper closely. The "wild card" in Nvidia's earnings report is "the extent to which Hopper is supply constrained," he wrote recently. That will determine the extent to which he expects Nvidia to beat expectations.
What are the expectations for early Blackwell shipments?
Analysts aren't too concerned about a possible short delay in the initial Blackwell shipments, but they will certainly be looking to Huang to comment on the expected timeline.
"We did have concerns when we first heard about supply chain and customer issues with the new Blackwell chips, including a potential three-month delay due to overheating risks, design issues, and some packaging issues," Melius Research analyst Ben Reitzes wrote in a research note to clients. "However, it appears Blackwell system shipments in the April 2025 quarter are still likely to be quite strong."
Some analysts believe that as long as the potential delay of Blackwell does not last too long, Nvidia will not actually lose business, but rather revenue may shift to subsequent quarters.
Raymond James analyst Srini Pajjuri believes there could be a gross margin benefit in the interim. "We expect any delays to drive sales of Hopper GPUs in the near term, which could actually benefit gross margins." He and his team estimate that Hopper's gross margins are 3% to 6% higher than Blackwell's.
How much will the performance exceed expectations?
While the consensus estimate is for $28.7 billion in revenue, up from $13.5 billion a year earlier, many analysts expect Nvidia to ultimately report sales of $30 billion or more.
Melius Research's Reitzes, for example, will be watching to see if Nvidia can beat estimates by $2 billion in the latest quarter.
Nvidia has beaten revenue estimates in each of the past five years, according to FactSet data. In the April quarter, it beat estimates by about $1.5 billion.
From an earnings perspective, Nvidia has missed adjusted EPS estimates in just one of the past 20 quarters, with its most recent report delivering a roughly 9% beat.
Analysts expect adjusted earnings per share of 65 cents, up from 27 cents a year ago. According to a FactSet report on August 16, Nvidia is expected to be the company with the largest annual earnings growth in the S&P information technology sector.
How will Blackwell's production be going next year?
Aside from the issue of possible delays in Blackwell shipments in the short term, the bigger topic is how the new chip will affect Nvidia once it goes into full production.
Raymond James’ Pajjuri is optimistic. The supply chain appears to be building as much as 50,000 to 70,000 GB200 NVL racks of capacity for the next calendar year, “which translates into $100 billion to $150 billion in potential revenue at the systems level alone,” he wrote.
“We wouldn’t be overly optimistic based solely on supply chain numbers, but we see room to move higher in the consensus estimate for annual data center revenues through 2025.” When he released his report last Tuesday, the consensus estimate was $146 billion; now it’s $148 billion.
While Jefferies analyst Blayne Curtis is optimistic about the short-term outlook, he is more cautious about the medium-term outlook.
"Over a month ago, we noted that market expectations for Blackwell's timing and content were overly optimistic and that we thought it would be difficult to reach $5.00 in earnings per share in fiscal 2025, but we continue to believe our expectations (which were approximately $1 billion higher in July and October) are achievable, but likely not exceedable," he wrote Thursday.
So what about 2026?
Yes, it's something Nvidia investors have been thinking about for more than a year.
Late last year, some on Wall Street questioned whether 2024 would be a “peak” year for Nvidia, Reitzes said. “Concerns about a ‘peak year’ for calendar year 2025 could soon resurface,” he wrote. “It would be good to hear what Jensen Huang says on this call about whether Blackwell’s successor will still follow a one-year upgrade cycle.”
He noted that Rubin products could lay the foundation for further revenue growth in calendar year 2026. (The FactSet consensus calls for $194 billion in revenue in calendar year 2026, up from an estimated $164 billion in calendar year 2025 and the $116 billion expected this year.)
"One issue Nvidia needs to address is concerns about power availability -- and how to help customers find affordable and available power," Reitzes wrote. "Blackwell and Rubin's innovations will help alleviate shortages and space constraints, but the view is that customers will continue to buy more products and use a lot of power."
The article is forwarded from: Jinshi Data