Nvidia is likely to report on Wednesday that its second-quarter revenue more than doubled. But investors used to its blockbuster results will be expecting even more from the artificial intelligence chip giant.

A beat or a miss on Wall Street expectations could either stoke or shatter an AI rally on Thursday, a day after Nvidia reports earnings for the May-July period.

The company’s shares have surged more than 150% this year, adding $1.82 trillion to its market value and lifting the S&P 500 to new highs.

The stock is valued at about 37 times its forward earnings, compared with an average of around 29 for the top six tech companies on the benchmark index that includes the chipmaker.

Tech heavyweights, including Microsoft, which are spending heavily to build out their AI infrastructure, have been buying Nvidia’s powerful graphic processing units that allow large amounts of computing quickly. These chips are difficult to replace in present-day datacenters, which has sharply boosted Nvidia’s fortunes.

Nvidia is expected to have recorded a year-over-year jump of about 112% in second-quarter revenue to $28.68 billion, according to LSEG data as of Aug. 23.

But its adjusted gross margin likely dropped more than 3 percentage points to 75.8% from the first quarter, burdened by the cost of a production ramp-up to meet growing demand.

“They’re not only a benchmark for chips, but they’re also a benchmark for AI as a whole,” said Daniel Morgan, senior portfolio manager at Synovus Trust, which owns shares in big U.S. tech firms, including Nvidia.

“If Nvidia misses, (investors are) going to sell off every company in AI.”

Some investors are concerned about the company’s ability to meet lofty expectations and have questioned the pace of spending on AI by Nvidia’s largest customers.

These worries led to a 20% slump in Nvidia’s stock through much of July and early August, though a recent recovery has left the stock just about 5% below its record high in June.

There may be more trouble brewing around potential production delays of Nvidia’s next generation Blackwell AI chips. CEO Jensen Huang said in May the chips would ship in the second quarter, but analysts have flagged design hurdles that could push the timeline.

This means revenue growth might take a hit in the first half of next year, research group SemiAnalysis said. Margins could also get squeezed if Nvidia’s chip contractor TSMC raises fees, a possibility that the Taiwanese firm hinted at recently.

Nvidia is likely to forecast a 75% surge in third-quarter revenue to $31.69 billion, LSEG data showed, ending its five-quarter run of triple-digit growth and reflecting tough comparisons from a year ago when it surged about 206% to $18.12 billion.

For the past three quarters, Nvidia’s growth exceeded 200%.

“We’re reaching the law of large numbers here, once a company gets to a certain size, it just physically can’t keep up the same growth,” said Michael Schulman, chief investment officer at Running Point Capital.

Some analysts said Nvidia could offset much of the hit from the delay in Blackwell chips by substituting those orders with its prior generation Hopper chips. The Hopper family of processors is not as powerful or lucrative as Blackwell, but it is sufficient for most AI-related applications.

Investors will also seek updates on AI processors for the China market, where sales of its most advanced chips are barred by the U.S. government.

Nvidia’s China-focused processors, reportedly called H20 and less powerful than its best chips, could help the company gain business over the next few quarters in a major market where domestic champion Huawei has emerged as a competitor.

There are also mounting antitrust concerns about the company’s practices. US regulators are probing whether Nvidia pressured cloud providers to buy multiple products, and if it is trying to bundle its networking equipment with their sought-after AI chips.