Newly installed Intel CEO Lip Bu Tan (LBT) held his first earnings call, covering the company’s 1Q25 finances and discussing its outlook. The company plans to cut costs by reorganizing and reducing capital expenditures (capex). Product plans are unchanged, but what was omitted from the 1Q25 review revealed more than what was discussed.
Does the Arrow Fly?
- Specifically, Arrow Lake wasn’t mentioned. Arrow is Intel’s current main PC processor and sits alongside Lunar Lake, which addresses laptops requiring greater power efficiency and integration. Arrow’s absence from the conversation indicates that it isn’t affecting the company’s financials, for better or worse.
- The older Raptor Lake PC processor sold unexpectedly well, mitigating the typical Q1 seasonal decline. Raptor is lower priced, and, being mature, OEMs have reduced manufacturing costs. Thus, they can offer systems at a lower price than those based on Arrow, Lunar, or Meteor Lake (a generation between Raptor and Arrow).
- Intel is better off selling Raptor than the newer chips despite its lower price. A monolithic design fabbed internally, it’s less expensive to produce than more recent chiplet-based PC processors, which are outsourced wholly or partially to TSMC. Lunar, in particular, has a lower gross margin percentage because its package incorporates DRAM.
- Intel Foundry Services (IFS) benefited from Raptor sales, and revenue for the ordinarily soft first quarter rose 8% sequentially.
- The Raptor bump might not last. At a minimum, Raptor sales will plateau as Intel 7 fabs reach full capacity. The company expects a shift to lower-margin and outsourced products (i.e., Meteor and newer processors). Moreover, the company expects Windows 10’s end of life, PCs bought during Covid needing to be refreshed, and AI capabilities will drive demand. Only post-Raptor chips accelerate AI processing.
Waiting for the AI PC to Ignite
- The AI PC story has holes. Intel states commercial users but not consumers are demanding AI PCs. But why? The only reason Intel gave was that they were future-proofing their fleets. However, Arrow and Meteor have anemic acceleration compared with Lunar and products like AMD Strix Point. Raptor’s demand suggests that a market swath (perhaps mostly consumers) prefers lower prices to AI performance. Enterprises may join them if an economic downturn depresses a refresh cycle and makes cheaper machines more attractive than future-proof ones.
- Panther Lake and Nova Lake will replace Lunar and Arrow, respectively. Intel reiterated the plan for these chips that it disclosed in the 4Q24 earnings call. The Panther computing die will employ the Intel 18A process, but other dice in this chiplet design will employ a TSMC process. The first model, a high-end SKU, will ship this year; the rest of the lineup will arrive in early 2026, followed by Nova. Intel will use both 18A and a TSMC process for Nova’s computing die, depending on the model. However, serving a range of designs, Nova will consume more Intel 18A wafers than Panther. The company aims to fulfill 70% of production in house, but this figure ignores the graphics and other dice.
Data-Center Dynamics Mirror the PC Market
- First-quarter data-center sales exceeded expectations, and Intel reported strong demand for older-generation parts, the same dynamic seen in the PC market. Hyperscalers drove this dynamic.
- Intel reiterated that Clearwater Forest, an efficiency-core Xeon employing the 18A process, which the company rescheduled for 2026, is on track. Because it’s a niche product, we believe Intel may cancel it.
- Overall, Intel exceeded its guidance for Q1. The company stated it hasn’t seen buying patterns change but contradictorily said it believes Q1 revenue benefited from customers purchasing in anticipation of tariffs.
- IPU sales will double in 2025. Google is the main customer of these I/O processors, which we call data-processing units (DPUs).
A New AI Strategy Starts to Form
- Based on its experience funding AI startups, LBT will refine Intel’s AI strategy. Details have yet to be determined.
- Enterprise customers will be the AI business’s focus; LBT stated, “Our goal will be to take our integrated system and platform view to develop full-stack AI solutions that enable more accuracy, power efficiency, and security for our enterprise customers.” By contrast, most data-center AI businesses prioritize hyperscalers because, as Willie Sutton said, that’s where the money is.
- Jaguar Shores, a rack-scale design, remains the centerpiece of Intel’s roadmap.
Setting IFS Straight
- Intel Foundry Services remains an important part of Intel’s strategy. Drawing on his experience helming Cadence, LBT recognizes IFS is a customer-service business and must embrace standard EDA tools and practices, understanding that customers’ design flows vary—blindingly obvious observations that IFS apparently was unable to make heretofore.
- TSMC also remains an important partner. On good terms with C. C. Wei, the foundry’s chairman and CEO, and Morris Chang, its founder and long-time leader, LBT states he is seeking to collaborate.
- The Intel 14A process design kit (PDK) is already sampling, the company stated at the IFS symposium following the earnings call. The 14A process will have fewer innovations than 18A but will build on that node’s gate-all-around transistors and backside power technologies. The company also disclosed new 18A variants.
Tariffs Bite
- The 2026 outlook is murky. The first quarter likely exceeded expectations as customers pulled in purchases in anticipation of tariffs. The direct (higher costs) and indirect (depressed economy) tariff effects are unpredictable. Intel foresees a market contraction.
- The second quarter will be comparatively weak, following the strong Q1. Owing to the heightened uncertainty, the company offered a wider-than-normal revenue range—down 2% to 12% sequentially.
- Data-center revenue will decline faster than client computing (PCs).
- Foundry revenue will decline, exacerbated by Intel 7 capacity constraints.
- Costs will increase with rising 18A startup costs and as lower-margin and outsourced products (e.g., Lunar and Arrow) account for a greater sales proportion.
CEO LBT Opens the Second Envelope
- Restructuring is coming. Flattening the organization to reduce management layers, breaking down silos, and eliminating bureaucracy are LBT’s priorities. To help understand how bad bureaucracy had gotten, in an open letter to employees, he explained, “I’ve been surprised to learn that, in recent years, the most important KPI for many managers at Intel has been the size of their teams.” He went on, “There is no way around the fact that these critical changes will reduce the size of our workforce.”
- Intel plans to reduce operating expenses (opex) by an additional $500 million compared with the 4Q24 earnings call’s targets. Next year will see a further $1 billion reduction. However, LBT vows to bring back critical talent lost in previous layoffs and early-retirement waves, acknowledging Intel’s recent brain drain.
- Layoffs are inevitable. A company, ideally, implements them as a surprise and in a single wave; otherwise, morale suffers. Unfortunately, talk of delayering and bureaucracy elimination telegraphs impending cuts, and Intel’s financial circumstances render them necessary.
- Free cash flow (FCF) for the quarter was negative 3.7 billion, a huge amount and more than twice last quarter’s negative FCF.
- Capex cuts are also coming. Intel will reduce spending by an additional $2 billion compared with its previous plan. Capex net of Chips Act grants won’t change much, however.
Bottom Line
Intel’s strategy is little changed from Pat Gelsinger’s plans. What differs, however, is execution. More incisive and purposeful than his predecessors, LBT promises faster and better decision making and cost reductions. External factors—tariffs—escalate the urgency to remake a bloated company and make a turnaround harder.