(Reuters) -Micron shares dropped 6% in premarket trading on Thursday after the chipmaker's current-quarter revenue forecast failed to impress investors looking for outsized results powered by the AI demand surge.

The chipmaker forecast fourth-quarter revenue of $7.6 billion, plus or minus $200 million, in line with analysts' average estimate even as the company said it had "sold out" its high-bandwidth memory (HBM) chips for this year and the next.

Micron is one of the few providers of HBM chips that power the world's most advanced AI systems, allowing the company to cash in on surging demand for semiconductors and driving up its stock 67% so far this year.

"Anything less than fantastic is not good enough when your share price got multiplied by three in just about 18 months," said Ipek Ozkardeskaya, senior analyst at Swissquote Bank in a note.

The chipmaker is set to lose $10 billion at current share price levels of $133.25. Peers Nvidia ( NVDA ), Broadcom ( AVGO ), Intel ( INTC ) and Qualcomm ( QCOM ) were also down 0.2-2% in premarket trading.

"The market reaction underscores the high expectations for every company that is part of the AI ecosystem," said analysts at Saxo Bank.

Some analysts, however, were positive about the firm's end markets, after the company beat estimates for third-quarter revenue.

Goldman Sachs's analysts view the stock's pullback "as an opportunity to add to positions" as the brokerage continues to see market share gains for the company in the lucrative HBM chip market.

Piper Sandler's Harsh Kumar had a similar view. "At a high level, end markets for MU continue to improve with demand increasing and supply still relatively tight."

"We envision that these conditions will continue to persist at least through the vast majority of 2025 as well," Kumar added.

At least two brokerages raised their price targets following the results. Micron has a 12-month forward price-to-earnings ratio of 17.07, compared with AI darling Nvidia's ( NVDA ) 40.22 and the industry median of 23.46, according to LSEG data.

(Reporting by Kanchana Chakravarty in Bengaluru; Editing by Saumyadeb Chakrabarty)