Intel cuts spending, eyes data center recovery after mixed Q2
CEO Lip-Bu Tan says the company's layoff plan is mostly complete as Intel tries to rebound after financial woes in recent years.
Shrinking semiconductor giant Intel Corp. will nix planned European factories and slow U.S. construction, the company said Wednesday after its latest earnings report showed flat revenue year over year for the quarter, and a loss on earnings per share.
Intel's second quarter 2025 results beat revenue expectations with $12.86 billion but reported a net income loss of $2.9 billion. Intel's foundry business, which launched last year, suffered an operating loss of $3.17 billion on $4.4 billion in revenue.
CEO Lip-Bu Tan said the company would cancel manufacturing operations in Germany and Poland while consolidating testing and assembly operations in Costa Rica into existing sites in Vietnam and Malaysia. The company will also slow down construction of a $28 billion chip factory in Ohio as it searches for customers.
"There are no more blank checks," Tan said about its foundry business as part of a memo to employees on Thursday. "Over the past several years, the company invested too much, too soon -- without adequate demand. In the process, our factory footprint became needlessly fragmented and underutilized. We must correct our course."
Mass layoffs near end
Tan said Intel's deep layoff plan was nearly complete, with 21,400 total job cuts expected by the end of 2025. The company will slim its workforce to 75,000 -- from the most current headcount of 96,400. The chipmaker said it has reduced the number of "management layers" by 50%.
"We are making hard but necessary decisions to streamline the organization," Tan said in his memo to employees. "These actions are critical to strengthening our competitive position going forward, but it means we are saying goodbye to valued colleagues."
Jack Gold, president and analyst at J. Gold Associates, in an email said the completion of layoffs "should help with profitability both with fewer impairment charges, as well as simply reduced costs of payroll. It's sad for the workers that are let go, but Intel needed to prune the payroll."
Focus on data center CPUs
Part of the course correction, the company said, will be to wrestle back market share from data center rival Advanced Micro Devices by focusing on its x86 chip business.
While Intel is still the leader in data center CPUs, AMD has captured significant market share in just a few years with its Epyc processor line. In the third quarter of 2024, AMD for the first time outsold Intel in the data center with $3.5 billion in revenue compared with Intel's haul of $3.3 billion over the same period.
"Companies like Intel don't turn on a dime, so it's going to take some time for the direction to emerge and play out," said Stephen Sopko, an analyst with HyperFrame Research. "So much of what Lip-Bu is doing is sorting through a storied company looking for value … AMD in the data center has thrived with great products, but also by capitalizing on Intel missing the rapid expansion of AI."
Tan plans on addressing the competitive shortfall by reintroducing simultaneous multi-threading (SMT) -- a technique that improves CPU performance by allowing multiple threads to execute simultaneously on a single processor. Intel moved away from SMT for its Lunar Lake design, opting for increased power efficiency to target mobile and AI PC use cases. That decision was made during former CEO Pat Gelsinger's tenure. Gelsinger retired in December.
"Moving away from SMT puts us at a competitive disadvantage," Tan wrote. "Bringing it back will help us close performance gaps."
Tan said future chip designs will require his final approval.
"Can Intel take back market share? Yes, of course," said Matt Kimball, an analyst at Moor Insights & Strategy. "The biggest impact to market share numbers would be in that cloud/hyperscale segment that has very little loyalty or stickiness to a certain vendor and is far more focused on performance per watt and performance per dollar."
Renewed AI focus
Intel has struggled with its AI workload processing business, with GPU maven Nvidia and AMD offerings outselling in large-scale AI model training.
"The company is exposed on the AI acceleration front," Kimball said. "Nvidia is at a $4 trillion market cap by owning the market."
Tan said the company will address its AI shortfalls with an emphasis on inference and agentic AI.
"Our starting point will be emerging AI workloads -- then we will work backward to design software, systems and silicon that enable the best customer outcomes," he said in his memo.
That seems like a good plan, Kimball said, noting that there's more space to compete in areas like inferencing. "That's where the long-term play is," he said.
Shane Snider, a veteran journalist with more than 20 years of experience, covers IT infrastructure at Informa TechTarget.
