Chipmaker Intel Corp.’s stock gained almost 8% in the extended trading session today after it delivered an upbeat forecast, beat expectations on profit and sales, talked up new customers for its nascent foundry business, and revealed growing traction in artificial intelligence.

The company reported third-quarter earnings before certain costs such as stock compensation of 41 cents per share, blowing past Wall Street’s projection of just 22 cents per share. Revenue for the period fell by 8% from a year earlier to $14.16 billion, but this was still ahead of analysts’ consensus estimate of $13.53 billion.

All told, Intel delivered a net profit of $297 million during the quarter, down from a profit of $1.02 billion a year earlier. It also reported a gross margin of 45.8%, flat year-over-year.

As for its guidance, it said it’s looking at fourth-quarter revenue of between $14.6 billion and $15.6 billion, streets ahead of the $14.4 billion analyst target. It’s also modeling earnings of 44 cents per share, compared with the 33-cent-per-share forecast.

On a conference call, Intel Chief Executive Pat Gelsinger (pictured) said that despite seeing inventory burn in the data center server market and wallet share shifts between central processing units and accelerators in the last few quarters, the market now appears to be normalizing. He also expressed confidence regarding the position of Intel and its Xeon CPUs in the rapidly growing market for AI.

“Training of these large models is interesting, but the deployment of those models, the inferencing use of those models, is what we believe is truly spectacular for the future,” Gelsinger said. “And… some of that will run on the accelerators, but a huge amount of that is going to run on Xeons.”

In recent years, Intel has been pushing its nascent foundry business that makes computer chips for other companies, and Gelsinger told analysts of the progress being made there. He said the company has now secured commitments for three large customers for its 18A foundry process technology, having previously only disclosed one. He didn’t name those customers.

“The other thing that we saw this quarter, which was a little bit unexpected, was this huge surge in interest for AI customers and Intel’s advanced packaging technology,” the CEO revealed.

According to Gelsinger, Intel remains on track to catch up with its chief rival Taiwan Semiconductor Manufacturing Co. by 2025. The company has conceived a plan known as “five nodes in four years” that aims to advance its chip manufacturing process to match its rival. During the quarter, it revealed it has made progress by beginning the mass production of chips using extreme ultraviolet lithography, or EUV, the most advanced semiconductor manufacturing technology on the market, at its Fab 34 plant in Leixlip, Ireland.

“While many thought our ambitions were a bit audacious when we began our ‘five nodes in four years’ journey roughly two and a half years ago, we have increasing line of sight toward achieving our goal,” he said.

Today’s results and the optimism of Intel’s shareholders show that the company’s decision to bring Gelsinger onboard to lead the company is beginning to pay off, said Charles King of Pund-IT Inc. “While the company often confounds those focused on quarterly performance, the size and breadth of Intel’s lines of business requires carefully focusing on long term efforts, goals and strategies,” he explained.

King said Gelsinger deserves kudos for the acceleration of Intel’s foundry business, with its acquisition of three new customers. “If Intel is correct in claiming that its 18A chip manufacturing process will be industry leading in terms of performance in 2025, it could signal a shakeup of assumptions and leadership in the semiconductor industry,” he pointed out.

Breaking down the numbers from the most recent quarter, Intel’s client computing group, which makes chips for laptops and personal computers, delivered sales of $7.9 billion, down 3% from a year earlier. The Data Center and AI business, which manufacturers server chips, saw sales fall 10% to $3.8 billion. Within this segment, Intel admitted it’s facing some “competitive pressure.”

King said the strength of the client computing business is a big surprise, as many in the industry claim that the PC market is moribund. “Intel attributed its success to PC vendors finally normalizing their inventories and said that server vendors are progressing similarly,” King said.

The foundry services business remains a very small part of Intel overall, delivering revenue of just $311 million, though it did grow by 300% from a year earlier. That was partly the result of a major customer making a prepayment, the company said.

Earlier today, Intel’s Mobileye, which is a publicly traded subsidiary for self-driving car chips, reported its sales rose 18%, to $530 million. Finally, the network and edge division, which sells networking chips and low-powered processors, saw sales drop 32%, to $1.5 billion.

Not everyone was pleased with Intel’s performance. Holger Mueller of Constellation Research Inc. said the 8% revenue decline shows that the company remains the “sick man of the chip industry” and still has plenty of work to do. “Pat Gelsinger and team did a good job by delivering to expectations, and they even posted a slight profits,” the analyst said. “However, investors will be watching closely over the next few quarters to see if Gelsinger can keep the company in the black.”

Earlier this month, Intel revealed that it intends to streamline its business by treating its programmable chip unit, which is currently a part of the Data Center and AI group, as a separate business. The plan is ultimately to spin out the unit as a separate company through an initial public offering, probably in the next two to three years.

Intel’s Programmable Solutions Group makes field-programmable gate arrays or FPGAs, which can be programmed for specific use cases by customers after they have been shipped out. FPGAs are commonly used in data centers, telecommunications, video encoding, aviation and other industries, and can also be used to run some artificial intelligence algorithms.

Photo: SiliconANGLE