More GPUs, Busier CPUs! Intel's Q2 Data Center AI Revenue Soars 59% — CEO Chen: Demand to Stay Strong Through 2028

Intel (INTC-US) recently released its Q2 2026 fiscal earnings, delivering the fastest revenue growth in nearly 15 years, fueled by rising AI demand. The results challenge the long-held market belief that 'GPU adoption will squeeze CPU demand.'

Analysts note that as AI development shifts from model training to inference, AI agents, and multi-agent applications, server CPUs are rapidly regaining importance as a critical component of AI infrastructure.

Intel's Q2 revenue reached $16.128 billion, up 25% year-over-year—the fastest growth in 15 years. Non-GAAP earnings per share (EPS) were $0.42, a significant improvement from a $0.10 loss per share in the same period last year.

The company's Data Center and AI Group generated $6.262 billion in revenue, a 59% YoY increase, with operating profit surging to $2.474 billion. Intel forecasts Q3 revenue between $15.8 billion and $16.8 billion.

Intel CEO Lip-Bu Tan stated that the sustained growth in AI-driven computing demand is the primary driver of the company's performance. CPUs, ASICs, advanced packaging, and foundry services will be the four pillars supporting Intel's ability to meet future AI demand.

AI Enters the Inference Era: CPUs Reemerge as Key Data Center Resources

For the past few years, demand for large language model training has driven a GPU procurement boom, with cloud providers directing capital expenditures toward GPUs. Server CPUs were often treated as secondary components.

However, as AI applications move into large-scale inference deployment, extensive CPU involvement is required for data preprocessing, model routing, network communication, permission management, task scheduling, and tool invocation before and after GPU operations—reigniting the strategic value of CPUs.

The rise of AI agents and multi-agent applications has further amplified this trend. A single user request may now involve multiple rounds of model inference, database queries, program execution, and external tool calls. CPUs, memory, and networking now handle more context management and control tasks, shifting data center procurement strategies from maximizing single-chip compute power to improving overall system utilization.

Intel's Q2 results reflect this shift. The Data Center and AI Group's revenue grew 59% YoY—far outpacing the 13% growth in client computing and edge AI businesses. Operating profit surged from $633 million last year to $2.474 billion, with profit growth exceeding revenue growth.

Additionally, due to increased shipments, improved product mix, and supply constraints, the Xeon 6 processor has become one of the fastest-ramping products in Intel's history.

Multiple studies also show that AI performance can no longer be measured solely by GPU count.

Researchers testing multi-GPU large model inference systems found that insufficient CPU allocation leads to core startup delays, communication bottlenecks, and token generation latency. Increasing CPU resources improved first-token latency by 1.36x to 5.4x across different system configurations.

Another study on five AI agent workloads revealed that in some cases, CPU-handled tool processing accounted for up to 90.6% of total latency.

While these findings may not apply universally, they clearly indicate that CPUs have become a critical factor in overall AI system efficiency amid the rapid adoption of AI inference and agent applications.

In response, Intel's leadership has raised its server CPU demand forecast, projecting double-digit growth in global server CPU shipments for 2026 and 2027, with momentum expected to continue through 2028.

Xeon Drives Profitability: Intel Targets AI Platform Opportunities

The CPU business recovery has brought much-needed profitability improvements to Intel.

Intel's Q2 non-GAAP gross margin reached 41.8%, a 12.1 percentage point increase from last year. Non-GAAP R&D and administrative expenses decreased to $3.967 billion, down 8% YoY.

Compared to a $503 million non-GAAP operating loss last year, Intel posted a $2.77 billion operating profit this quarter—driven by revenue expansion, yield improvements, and cost controls.

In product strategy, Intel continues to strengthen Xeon's role in heterogeneous computing architectures.

In Intel's proposed AI system architecture, Xeon handles host computing and task orchestration, SambaNova's RDU performs partial inference, and NVIDIA's Blackwell GPU focuses exclusively on AI acceleration—creating a collaborative,分工ed model.

Moreover, Xeon 6+ will be the first server processor to adopt Intel's 18A process.

Analysts note that Intel is not attempting to replicate NVIDIA's GPU ecosystem. Instead, it leverages its x86 software foundation, enterprise customer relationships, and on-premises deployment experience to provide data exchange, scheduling, and management capabilities across different AI accelerators.

If Xeon can effectively integrate AI model inference, enterprise databases, legacy business software, and data governance workflows in the future, Intel's revenue streams could expand into networking equipment, custom ASICs, software optimization, and advanced packaging.

However, competition remains intense. AMD continues to gain server CPU market share, cloud providers are accelerating adoption of Arm architecture and in-house chips, and GPU vendors are actively integrating CPU and AI accelerator products.

Intel's leadership acknowledges that some products still lag behind competitors in single-threaded and multi-threaded performance and energy efficiency, requiring ongoing improvements.

Strong AI Demand, But Supply Capacity Is the Biggest Challenge

Compared to demand, Intel's greater challenge lies on the supply side.

Management stated that server CPU demand already exceeds supply capacity. Wafers, substrates, and memory remain the primary bottlenecks in AI infrastructure expansion. New production capacity is expected to come online in late Q3 and Q4, meaning short-term revenue will still be constrained by ramp-up speed.

While supply shortages enhance order visibility and pricing power, they also force the company to increase investment.

Intel has raised its 2026 capital expenditure to over $20 billion, with further increases expected in 2027. Investments will focus on Intel 3, 18A, and 18A-P processes, advanced packaging, substrates, and memory supply chains.

However, high investment brings cash flow pressure.

Intel's Q2 operating cash flow was $7.006 billion, but adjusted free cash flow remained negative at $8.419 billion. The company attributes this to changes in partner investments, capital expenditures, and finance lease payments. While AI demand has improved operating profits, the benefits have not yet fully translated into free cash flow.

Meanwhile, foundry operations remain Intel's largest source of losses.

The foundry business generated $5.765 billion in Q2 revenue, up 31% YoY, but posted an operating loss of $2.089 billion—down from $3.168 billion last year, but still substantial.

The 18A-P process has entered risk production, and some Panther Lake products have begun volume production. Future yield improvements, external customer adoption, and capacity utilization will determine when the foundry business can further reduce losses.

Overall, Intel's latest earnings signal a shift from demand validation to manufacturing capability validation. The rapid development of AI inference and agent applications has reestablished CPUs as a core element of AI infrastructure, offering Intel new growth opportunities.

However, whether AI business potential can truly translate into long-term profitability depends on the timely execution of the 18A process, advanced packaging, foundry expansion, and supply chain scaling.

For Intel, GPUs ignited the AI compute race—but whether CPUs can lead the company back to growth will depend on manufacturing capability and supply efficiency in the coming quarters.