For Nvidia For investors, the past two years have been a joyride, but lately it has been more of a rollercoaster.

As the primary beneficiary of the rise of artificial intelligence, Nvidia has seen its market capitalization expand roughly ninefold since the end of 2022. But after hitting a record high in June and briefly becoming the world’s most valuable public company, Nvidia proceeded to lose nearly 30% of its value over the next seven weeks, shedding roughly $800 billion in market capitalization.

Now, it is in the midst of a rally that has taken the stock to within about 6% of its all-time high.

The chipmaker is set to report quarterly results on Wednesday, so volatility in its stock is the main concern on Wall Street. Any hint that demand for artificial intelligence is waning or that a leading cloud customer is tightening its belt slightly could translate into a significant drop in revenue.

“It’s the most important stock in the world right now,” EMJ Capital’s Eric Jackson said on CNBC’s “Closing Bell” last week. “If they lay an egg, it would be a major problem for the entire market. I think they’re going to surprise to the upside.”

Nvidia’s report comes weeks after its large-cap tech peers released their results. The company’s name was present at all of those conferences with analysts, such as Microsoft, Alphabet, Goal, Amazon and Tesla Everyone spends a lot of money on Nvidia graphics processing units (GPUs) to train AI models and run massive workloads.

Over the past three quarters, Nvidia’s revenue has more than tripled year over year, with the vast majority of the growth coming from its data center business.

Analysts expect a fourth consecutive quarter of triple-digit growth, but at a reduced pace of 112% to $28.7 billion, according to LSEG. From here, year-over-year comparisons become much more difficult, with growth expected to slow in each of the next six quarters.

Investors will be paying particular attention to Nvidia’s guidance for the October quarter. The company is expected to show growth of around 75% to $31.7 billion. Upbeat forecasts will suggest that Nvidia’s deep-pocketed customers are signaling a continued willingness to open their wallets for AI development, while a disappointing forecast could raise concerns that infrastructure spending has become excessive.

“Given the sharp rise in hyperscale capital spending over the past 18 months and the strong near-term outlook, investors are often questioning the sustainability of the current path of capital spending,” Goldman Sachs analysts, who rate the stock a buy, wrote in a note last month.

Much of the pre-report optimism — the stock is up nearly 10% in August — is due to comments from major customers about how much they continue to shell out for Nvidia-based data centers and infrastructure. Shares rose 1.5% on Tuesday, closing at $128.30.

Last month, the chief executives of Google and Meta enthusiastically backed the pace of their developments, saying that underspending was a bigger risk than overspending. Former Google CEO Eric Schmidt recently told Stanford students, in a since-deleted video, that he was hearing from major tech companies that they “need $20 billion, $50 billion, $100 billion worth of processors.”

But even as Nvidia’s profit margins have expanded lately, the company still faces questions about the long-term return on investment that customers will see from their purchases of devices that cost tens of thousands of dollars each and are ordered in bulk.

During Nvidia’s last earnings call in May, CFO Colette Kress provided data suggesting that cloud providers, which account for more than 40% of Nvidia’s revenue, would generate $5 in revenue for every $1 spent on Nvidia chips over four years.

More such statistics are likely to be released in the future. Last month, Goldman analysts wrote after a meeting with Kress that the firm would share more ROI metrics this quarter “to instill investor confidence.”

The other big question facing Nvidia is the timeline for its next-generation AI chips, dubbed Blackwell. The Information reported earlier this month that the company is facing production issues, which will likely delay large shipments until the first quarter of 2025. Nvidia said at the time that production was on track to ramp up in the second half of the year.

The report came after Nvidia CEO Jensen Huang surprised investors and analysts in May by saying the company will see “a lot” of revenue from Blackwell this fiscal year.

While Nvidia’s current generation of chips, called Hopper, remains the premium choice for deploying AI applications like ChatGPT, competition is emerging from Advanced MicrodevicesGoogle and a handful of startups, putting pressure on Nvidia to maintain its performance lead through a smooth update cycle.

Even with a potential Blackwell delay, those revenues could be pushed back to a future quarter, boosting current Hopper sales, especially of the new H200 chip. The first Hopper chips went into full production in September 2022.

“That timing shift doesn’t matter much as customer supply and demand have quickly pivoted toward H200,” Morgan Stanley analysts wrote in a note this week.

Many of Nvidia’s biggest customers say they need the extra processing power of Blackwell chips to train more advanced next-generation AI models, but they’re willing to take what they can get.

“We expect Nvidia to reduce the emphasis on its Blackwell B100/B200 GPU allocation in favor of increasing its Hopper H200 offering” in the second half of the year, HSBC analyst Frank Lee wrote in an August note. He has a buy rating on the stock.

Correction: Colette Kress is Nvidia’s chief financial officer. In an earlier version, her name was misspelled.