Nvidia’s second-quarter earnings, due this Wednesday, will land exactly two years after the AI boom kicked the company into a new era.
Back in late 2022, when OpenAI launched ChatGPT, Nvidia was mostly known for its gaming chips. Since then, the company has grown into the biggest name in AI infrastructure, with revenue up more than three times and profits quadrupled, according to CNBC.
Last month, Nvidia became the first company in history to cross a $4 trillion market cap. The stock is up 33% this year alone, ending last week at $177.99. Since the AI boom started, the company’s shares have gained twelvefold.
But despite that massive rise, the rate of growth has dropped. After five straight quarters of triple-digit gains in 2023 and early 2024, Q1 growth this year dropped to 69%. Analysts now expect the company to post 53% growth in Q2, hitting $45.9 billion in revenue.
AI data centers now dominate Nvidia’s business. In Q1, 88% of the company’s total sales came from its data center segment. Nvidia also reported that 34% of last year’s total revenue came from just three unnamed buyers. The company now makes up 7.5% of the entire S&P 500.
In late July, every other major tech firm reported earnings and updated their capital expenditure forecasts. The result: combined spending of $320 billion this year on AI infrastructure and data centers, not counting OpenAI.
With AI capex still climbing, Nvidia’s role stays central. Analysts estimate the company takes in about half of total AI infrastructure spend.
But there are warnings. Sam Altman, OpenAI’s CEO, said last week, “Investors as a whole are overexcited about AI,” and called it a potential “bubble.” Wall Street is watching to see what Nvidia CEO Jensen Huang says on Wednesday.
The most important product line in Nvidia’s pipeline is Blackwell. The architecture includes standalone GPUs and full systems with 72 GPUs combined.
Nvidia said in May that Blackwell sales hit $27 billion, making up 70% of total data center revenue. That’s up from $11 billion the quarter before. GPT-5 was trained on Nvidia’s earlier Hopper chips. Experts expect even more demand as Blackwell rolls out further.
The company warned last year that the limit wasn’t demand, but how many chips partners can build. Blackwell Ultra is scheduled to begin shipping in late 2025. Meanwhile, Nvidia denied reports from Asia that Rubin, its 2027 chip line, is already facing production issues.
Jensen Huang has also been trying to manage the China situation. Trump confirmed this month that Nvidia agreed to pay 15% of its China AI chip revenue to the U.S. government in exchange for export licenses. That deal covered the company’s H20 chip, which was designed for the Chinese market. Trump said he initially asked for 20%, but Jensen talked him down.
Nvidia said in May that the H20 could have added $8 billion to Q2 revenue. But that was before the U.S. added license restrictions and China started pushing its cloud providers to use domestic chips like those from Huawei. As a result, Nvidia excluded H20 from its Q2 guidance and is unlikely to include it in Q3 either.
KeyBanc analysts said including H20 could raise guidance by $2 to $3 billion. But they expect the company to skip it, just like AMD did in early August.
Nvidia is already working on a Blackwell-based AI chip for China, but it would still need Trump’s approval. Most of Nvidia’s big customers already posted solid earnings this season. They also plan to increase capex, which bodes well for Nvidia’s numbers.
Options pricing shows expectations of a 6% move either way after the report. Also, the S&P 500 trades at about 22 times forward earnings, higher than its 10-year average. Nvidia trades at 34 times, below its own five-year average of 39.
Still, the Street isn’t flinching. Over the last week, at least nine analysts raised their price targets. The average now sits at over $194, roughly 9% above Nvidia’s $178 close on Friday.
Sign up to Bybit and start trading with $30,050 in welcome gifts