Intel CEO bets on agentic AI and foundry revival as chipmaker eyes 10A and 7A process nodes

Intel's CEO reveals early hiring challenges as bankruptcy concerns deterred top talent

Recovering chipmaker looks beyond 14A to 10A and 7A process nodes in foundry comeback bid When Lip-Bu Tan took over as Intel CEO, the company's balance sheet was so dire that potential recruits turned him down flat, worried they'd be joining a chipmaker on the verge of going bust. Speaking at the JP Morgan Global Technology, Media and Communications Conference on Tuesday, he said: "I tried to recruit some talent. They said 'It's almost a bankrupt company, why should I join you?'" Fixing that became his first priority. That effort has since paid off as Lip-Bu secured equity investment from the Trump administration, which converted funds from the CHIPS program in exchange for a stake. He also drew on long-standing personal relationships, with Nvidia CEO Jenson Huang committing $5 billion and Softbank's Masayoshi Son - a former Intel board member - signing on as a backer. "So far, knock on wood, I made money for them, and they're quite happy," Lip-Bu said. The stronger balance sheet has since enabled Intel to buy back a stake it had sold to Apollo, reducing earnings-per-share dilution in the process, he added. Now, a year into the job, Lip-Bu is betting on agentic AI, inference workloads and a bold chipmaking roadmap to complete Intel's revival - looking beyond the upcoming 14A process node to future 10A and 7A chipmaking technology. When asked for a progress report on the process technology Intel uses to manufacture its products, Lip-Bu said the recently introduced 18A is seeing a 7 percent per month yield improvement, and the next-gen 14A node is "ahead of schedule" compared to the end of the year target. "And now I'm starting to look at the 10A, 7A, the roadmap," Lip-Bu said. "People don't go to you just for one node. They're looking for the roadmap for the future. So we want to build a long-term business. And then we can drive the efficiency, the defect density, and then we can go to that Rule of 45, how to drive the operating efficiency, the profitability, cash generation." Intel's 18A and 14A processes refer to 18 and 14 angstroms, and as there are 10 angstroms in 1nm, this could imply that the company is working toward a sub-nanometer process technology with 7A. The Register asked Intel if it was willing to disclose when these process nodes would likely come into use, and we will update this story if we get a response. However, Lip-Bu did mention that for 14A, "my risk production is 2028 and volume production in 2029 is about the same time as A14 for TSMC," implying that 10A and 7A are unlikely to be used to make chips before 2030. On Intel's efforts to reinvent its internal chipmaking operation as a foundry service for third-party customers, he confirmed it is engaged with multiple clients, but declined to identify any, saying disclosure was up to the customer. Manufacturing used to be Intel's strength, but it lost its way, and has never really been in the foundry service business, Lip-Bu said. For this reason, he recently poached Shawn Han, a veteran with three decades' experience at Samsung Foundry, to operate as SVP of Foundry Services. Customers are also asking about Intel's 18AP process node, an enhanced, performance-focused variant of 18A, he claimed. Previously, Intel's foundry biz planned to offer the 14A node as its first mainstream commercial offering. To show how keen potential customers are, Lip-Bu claimed they are even willing to help Intel with pre-payments on wafer substrate materials. "Some of the substrate material is very [short], they're all asking us to prepay the substrate commitment. And we ask our customers, if you are serious to use our EMIB-T [packaging technology], can you help me on the substrate prepay? They jump on it," he claimed. "So they show the commitment, they really want our technology. And this is not a few million, it's billions in the next few years." Lip-Bu expects AI to be Intel's route to recovery after a disastrous few business years, as agentic AI and inference workloads look set to favor CPUs rather than the GPUs that have been making Nvidia's fortune. "It used to be that training is 1 CPU to 8 GPUs. And now in the agentic AI with all the agents, startups all tell me, Lip-Bu, CPU actually is more useful, even single-threaded," he said. "So I can start to see not just my wishful thinking, customers have said to me, Lip-Bu, more like 1:1. And now even some of them tell me it's 4:1. So 4 CPU to 1 GPU, for the inference and agents. And so CPU [is in] high demand, and I try to make sure that we can meet the requirement from the customer." Intel still has some way to go. As The Register pointed out earlier this year, the chipmaker lost $267 million on revenues of $52.9 billion during 2025, compared to an $18.8 billion loss the year before. Maybe it will even turn a profit this year. ®

Intel CEO Is Building A 'New Intel' That Works At 'Speed Of Light' - Intel (NASDAQ:INTC)

Inside Lip-Bu Tan's Push To Reinvent Intel Intel has compressed its management hierarchy from as many as 12 layers down to five, with engineering leaders now reporting directly to Tan. The company has also adopted a "bad news first" philosophy that requires problems to be surfaced within 24 hours, while eliminating unnecessary meetings and tightening accountability standards internally. Tan contrasted the changes with Intel's previous environment, where some decisions could take up to a year. The turnaround push appears to extend beyond management culture and into customer relationships as well. According to Intel, the company worked directly through "14 areas" where one major customer believed Intel had previously fallen short, part of a broader effort to rebuild trust in its foundry and engineering operations. Intel Says Foundry Recovery Is Starting To Gain Credibility The cultural overhaul comes as Intel tries to convince customers and investors that its foundry business is finally stabilizing. The company said its 18A process node is now in volume production, supporting the Panther Lake ramp, while yields are improving roughly 7% per month and tracking ahead of Intel's year-end internal targets. Intel also said it has secured roughly 200 design wins tied to the node. Tan reiterated that foundry is fundamentally a service business where trust is earned through execution, cycle time, yield quality and reliable delivery -- not simply ambitious roadmap promises. Intel's roadmap now extends beyond 18A toward 14A, 10A and 7A development, with 14A risk production targeted for 2028 and volume manufacturing expected in 2029, broadly aligned with TSMC's A14 timing. The company has also embraced external EDA partners like Cadence Design Systems and Synopsys while cutting roughly two-thirds of its prior internal EDA budget. Intel Thinks AI's Next Phase Could Favor CPUs Again While Wall Street remains heavily focused on GPUs, Intel argued that the next wave of AI workloads may increasingly benefit CPUs. Management said inferencing surpassed training workloads in the second half of 2025, while emerging "agentic AI" systems are becoming far more CPU-intensive than traditional AI training environments. According to Intel, the historical CPU-to-GPU ratio of roughly 1:8 in training clusters is now approaching parity in inference-heavy deployments, with some customers reporting ratios as high as 4:1. During the CNBC Mad Money interview, Tan reinforced Intel's turnaround ambitions, saying the company previously "used to have leadership in data center" before losing it -- and that he is now trying to bring that leadership back. Intel also signaled that it sees "physical AI" -- including robotics and digital workers -- as the next major strategic frontier beyond agentic AI, with the company building capabilities spanning silicon, software and systems engineering. For now, though, Intel's biggest challenge may be proving that a company long criticized for moving too slowly can suddenly start operating at "speed of light." Image via Shutterstock Market News and Data brought to you by Benzinga APIs To add Benzinga News as your preferred source on Google, click here.