CPU Returns to the AI Main Stage: Intel's Q2 Earnings Mark a Pivotal Battle

Intel stands at a critical turning point. As AI workload structures shift from training to inference, and then to agents, the strategic position of CPUs in data centers is quietly recovering, leading the market to reprice the investment logic for this veteran chip giant.

After the U.S. market close on July 23, Intel will release its second-quarter earnings.

This is not an ordinary quarterly performance disclosure. With the stock price having risen more than 160% year-to-date and valuation remaining high at approximately 87 times forward earnings, Wall Street demands that management deliver on revenue growth, Gross Profit Margin improvement, and sustainable growth guidance. Furthermore, they must provide concrete evidence that demand for data center CPUs and the transformation of the foundry business have both entered a quantifiable realization phase. Most analysts bet that this performance will exceed the company's original guidance, with HSBC raising its target price by 100% to $200, marking the highest target among Wall Street analysts.

However, the market's high expectations themselves constitute a risk. After Intel announced better-than-expected outlook in the first quarter, its stock rose 19% in after-hours trading, subsequently surging sharply driven by the AI chip boom and expectations for domestic manufacturing policies. But as of July 17, the stock price had retreated significantly from its June highs to around $95, while the Philadelphia Semiconductor Index fell more than 20% from its peak during the same period, entering a technical bear market. Whether the earnings report becomes a catalyst for a new upward trend or evolves into a "sell the news" event will largely depend on Intel's ability to demonstrate its execution capability in converting AI demand into deliverable capacity and sustainable profits.

Fundamental Improvements, but High Barriers Already Set

Intel's first-quarter revenue was $13.6 billion, a 7% year-over-year increase, surpassing its own expectations for the sixth consecutive quarter, laying a fundamental starting point for this earnings report.

Non-GAAP earnings per share (EPS) were $0.29, a significant increase from $0.13 in the same period last year; adjusted net profit grew 156% to $1.5 billion; operating margin improved from 5.4% to 12.3%; and adjusted R&D and administrative expenses decreased by 9% year-over-year, indicating progress in restructuring and cost control.

For the second quarter, Intel's official guidance is revenue between $13.8 billion and $14.8 billion, with a midpoint of $14.3 billion; non-GAAP Gross Profit Margin of approximately 39%, and non-GAAP EPS of $0.20. This guidance itself conveys positive signals of strong server demand and improved product supply.

But the real threshold is significantly higher than this.

According to reports, the current market is not satisfied with Intel merely "meeting" guidance, but requires management to simultaneously prove revenue upside, Gross Profit Margin improvement, and sustainable growth in future performance guidance. If data center business revenue is strong but Gross Profit Margin remains around 39%, the market may judge that incremental revenue is offset by manufacturing costs or supply bottlenecks; only if the average selling price of the Xeon product mix and total revenue growth simultaneously drive margin expansion will the revival of CPU demand truly be reflected in the profit statement.

Notably, Intel still recorded a GAAP net loss—a GAAP net loss of $3.7 billion in the first quarter, or $0.73 per share, mainly due to large provisions related to restructuring. This means that the current forward P/E ratio of approximately 87 is essentially a speculative bet on the accelerated improvement of the profitability slope, rather than a safe-margin investment based on current free cash flow or static valuation.

CPU Renaissance: AI Agents Reshape the Structure of Computing Demand

Over the past two years, the AI narrative was almost monopolized by GPUs, with CPUs once relegated to a supporting role in the AI arms race. However, as AI workloads shift from model training to inference, and then to agents (Agentic AI), the market is re-recognizing the irreplaceability of CPUs in system architecture.

Agent AI requires continuous execution of tool calls, code running, retrieval, database access, task scheduling, memory management, and multi-process orchestration, causing the bottleneck status of CPUs in end-to-end throughput to rise again. According to Investing.com citing data from Tom's Hardware, the configuration ratio of CPUs to GPUs in AI servers has moved from 1:8 towards 1:4, and may converge to 1:1 in agent scenarios. This means that the same GPU clusters require more server CPUs, directly boosting Xeon demand, delivery cycles, and pricing power.

Intel CEO Lip-Bu Tan stated at J.P. Morgan's 54th Annual Global Technology, Media and Communications Conference that Intel 18A has supported the mass production of Panther Lake, with yields improving by about 7% monthly, exceeding Intel's internal expectations; as the focus of AI computing infrastructure shifts from training to inference, the configuration ratio of CPUs to GPUs is accelerating towards 1:1, and may even reach 4:1.

First-quarter Data Center and AI business revenue was $5.1 billion, a 22% year-over-year increase, significantly faster than the 1% growth rate of the Client Computing business, confirming this structural shift. NVIDIA's choice of Intel Xeon 6 as the host processor for its DGX Rubin NVL8 system, and Google's announcement to expand the use of Intel processors for cloud workloads and collaborate on developing custom infrastructure processing chips, both endorse Intel's presence in AI data centers.

TSMC's second-quarter performance also provided cross-validation at the industry level. TSMC's second-quarter revenue in U.S. dollars reached $40.2 billion, at the upper end of guidance, with a Gross Profit Margin of 67.7%, and expects third-quarter revenue to further rise to between $44.6 billion and $45.8 billion; management emphasized that AI demand has multi-year sustainability and pointed out that AI agents are driving a rebound in computing demand such as data center CPUs, indicating that AI infrastructure is upgrading from simply stacking GPUs to a systemic engineering expansion involving GPUs, CPUs, and advanced packaging.

Furthermore, Intel recently raised server CPU prices, with the new suggested customer price for the data center-grade Xeon 6980P rising to $13,955, an increase of $1,495 from $12,460 in 2025, validating the market judgment that "demand exceeds supply" rather than simple cost pass-through.

Foundry Transformation: Order Narrative Still Needs Yield and Cash Flow Verification

The chip foundry business is both the determinant of Intel's valuation ceiling and the current largest source of risk.

First-quarter Intel Foundry division revenue was approximately $5.4 billion, a 16% year-over-year increase, but about $5.25 billion of this was eliminated internally, indicating that the vast majority still comes from providing manufacturing services for Intel's own products, with a very low proportion of external customer revenue; the division recorded an operating loss of about $2.44 billion in the same period, further expanding from about $2.32 billion in the same period last year.

18A is the technical pivot of this transformation. In June, Intel announced that 18A-P entered risk production, stating that compared to 18A, it offers a 9% performance improvement at the same power consumption, or an 18% power reduction at the same performance, helping to enhance external customers' design willingness. However, risk production is not equivalent to large-scale mass production, nor does it mean that economic yields have been met. According to Investing.com, rumors that "wafer-to-wafer yield fluctuations have been resolved" still mainly come from supply chain research rather than official disclosures by Intel. Therefore, statements regarding yield, capacity ramp-up, customer tape-outs, and mass production timelines during the earnings Q&A session may determine the stock price direction more than the current EPS.

On July 13, Intel announced a €5 billion (approximately $5.7 billion) expansion investment in its Leixlip manufacturing campus in Ireland, used to update facilities, increase manufacturing capacity, and improve production efficiency for Xeon 6 and future Xeon series CPUs, utilizing Intel 3 process technology and existing cleanroom space without building new fabs. This investment has strategic significance against the backdrop of rising server chip demand, but it also means the foundry business remains in a loss-making period, with continued pressure on capital expenditures.

Potential foundry relationships with Apple, Tesla, SpaceX, xAI, and NVIDIA should be viewed as long-term option value, not as deterministic revenue that can be capitalized in advance. Customer testing of 18A or evaluation of packaging services is fundamentally different from formally entering mass production; scaled production may take several years. The most reasonable investment framework is to view these relationships as upside call options for the foundry business, while basing recent valuations on verifiable data center CPU revenue, external foundry orders, and chip capacity and yield data.

Technical Analysis: $94 Double Bottom Support Becomes Key

From a technical perspective, Intel's stock price is testing the important double bottom support level in the $94 to $96 range on the daily chart. Recent closing prices have fallen below the 50-day simple moving average ($109.68) and broken through the previous resistance level of $107.21, which had acted as support.

The Relative Strength Index (RSI) is currently around 37, indicating strong downward momentum and approaching oversold territory, but there is no clear reversal signal yet. Analysis suggests that if the $94 support level holds, the stock price is expected to rebound to $107.20, and then challenge $116.50; if the closing price falls below $94, the double bottom pattern will break, with lower support levels in the $84.50 and $75.40 regions respectively.