Intel’s incoming CEO Lip-Bu Tan is contemplating substantial modifications to its chip manufacturing processes and artificial intelligence plans prior to his return to the business on Tuesday, according to two sources familiar with Tan’s perspective, in a comprehensive effort to rejuvenate the struggling technology giant.

The revised strategy includes reorganizing the company’s AI initiatives and implementing workforce reductions to rectify what Tan perceives as an inefficient and cumbersome middle management tier. According to reports, one of Tan’s primary objectives is to overhaul the company’s production facilities, which previously exclusively produced chips for Intel but have now been adapted to manufacture semiconductors for external clients, including Nvidia.

The plans are under formulation and may possibly be subject to alteration, according to these sources.

Intel (INTC.O) shares increased over 8% during mid-day trading on Nasdaq.

During a town hall meeting subsequent to his recent appointment as CEO, he informed staff that the company must undertake “difficult decisions,” as reported by two individuals informed about the discussion.

Dylan Patel, a specialist in the semiconductor sector, stated that a significant issue during Pat Gelsinger’s tenure as Intel CEO, which concluded in December, was his overly accommodating demeanor. “He was reluctant to terminate numerous middle management personnel as required,” he stated.

Tan, 65, former CEO of the semiconductor design software company Cadence and technology investor, served on Intel’s board until his resignation in August. Upon his return as CEO, Tan is poised to assume control of the American icon following a decade of poor decisions made by three predecessors, during which the company failed to manufacture chips for smartphones and neglected the burgeoning demand for AI processors, enabling competitors Arm Holdings and Nvidia (NVDA.O) to dominate both sectors.

In 2024, Intel (INTC.O) had an annual loss of $19 billion, marking its first such deficit since 1986.

In the short term, Tan intends to enhance the performance of Intel Foundry, a manufacturing division that produces chips for design firms like Microsoft (MSFT.O) and Amazon (AMZN.O), by actively soliciting new clients, as reported by sources.

The company will resume initiatives to manufacture chips for AI servers and explore opportunities beyond servers in many domains, including software, robotics, and AI foundational models.

An Intel official stated, “Lip-Bu will dedicate considerable time to engaging with customers, partners, and employees as he joins the team and collaborates closely with our leadership to strategically position the business for future success.”

Intel refrained from providing more comments or facilitating an interview with Tan. Walden Catalyst, Tan’s venture firm, did not provide a response to inquiries for comment.

Initially, Tan’s idea seems to refine Gelsinger’s approach. The centerpiece of Gelsinger’s turnaround plan was to transform Intel into a contract chip manufacturer that would compete with Taiwan Semiconductor Manufacturing Co. (2330.TW), or TSMC, which counts Apple (AAPL.O), Nvidia and Qualcomm (QCOM.O), as customers.

Gelsinger committed tens of billions of dollars to build factories in the U.S. and Europe to make chips for both Intel and outside customers, but he was forced to scale back those ambitions as the market for Intel’s core products cooled.

BETTING ON AI

Tan has been a vocal internal critic of Gelsinger’s execution, according to the two sources familiar with Tan’s plans.

For most of its history, Intel has manufactured chips for only one client—itsself. When Gelsinger became CEO in 2021, he prioritized manufacturing chips for others but fell short of providing the level of customer and technical service as rival TSMC, leading to delays and failed tests, former executives have told Reuters.

Tan’s views were shaped by months of reviewing Intel’s manufacturing process after the board in late 2023 appointed him to a special role overseeing it, according to a regulatory filing.

In his assessment, he expressed frustration with the company’s culture, sources told Reuters, saying it had lost the “only the paranoid survive” ethos enshrined by former CEO Andy Grove. He also came to believe that decision-making was slowed down by a bloated workforce, Reuters reported.

Tan presented some of his ideas to Intel’s board last year, but they declined to put them into place, according to two people familiar with the matter. By August, Tan abruptly resigned over differences with the board, Reuters reported.

When he returns as CEO this week, he will lay fresh eyes on Intel’s workforce, which was slashed by roughly 15,000 to almost 109,000 at the end of last year, the sources said.

Beyond the cuts, Tan has little choice but to make Intel’s existing manufacturing operation work in the short run. Intel’s next generation of advanced chips equipped with AI features, called Panther Lake, will depend on its in-house factories using a new set of techniques and technologies Intel calls “18A.”

Intel’s financial success this year is tied to strong sales of the forthcoming chip.

Tan signaled in a memo Intel published Wednesday that he plans to keep control over the factories, which remain financially and operationally separate from the design business, and restore Intel’s position as a “world-class foundry.”

Intel’s contract manufacturing operation can succeed if Tan wins over at least two large customers to produce a high volume of chips, industry analysts and Intel executives told Reuters.

Part of the effort to lure large customers will involve improving Intel’s chip manufacturing process to make it easier for potential customers like Nvidia and Alphabet’s (GOOGL.O) Google to use.

Intel has demonstrated improvements in its manufacturing processes in recent weeks and has attracted interest from Nvidia and Broadcom, which have launched early test runs, Reuters reported. Advanced Micro Devices is also evaluating Intel’s process.

Tan is expected to work on ways to improve output, or “yield” to deliver higher numbers of chips printed on each silicon wafer as they move to volume manufacturing of its first in-house chip using the so-called 18A process this year.

The goal is to move to an annual release schedule of AI chips, similar to Nvidia, but that will take years. It will be at least 2027 before Intel can develop a compelling new architecture for a first AI chip, according to three industry sources, and one person familiar with Intel’s progress.