Intel Corp. (Intel (INTC)) laid out its next AI move on Monday, and it's a classic underdog strategy: go where the giant isn't looking. Instead of trying to beat Nvidia (Nvidia (NVDA)) at the training game—where Nvidia's GPUs are basically the only option—Intel is targeting AI inference, the process of running already-trained models. And it's doing it with cheaper parts and in-house manufacturing.

The centerpiece is a new GPU called Crescent Island, which Intel plans to start shipping in limited quantities by the end of 2026. Kevork Kechichian, who runs Intel's data-center group, told the Financial Times on Monday that the company learned from the disappointing Gaudi AI chip program and decided not to go head-to-head with Nvidia in training. Instead, it's focusing on inference, where customers care more about cost than raw performance.

And cost is the key selling point. Unlike Nvidia's and AMD's (AMD (AMD)) chips, which rely on expensive high-bandwidth memory (HBM) and liquid cooling, Intel's Crescent Island uses cheaper LPDDR5 memory and air cooling. That could make it a lot more affordable for companies running AI models at scale. Kechichian also said Intel is exploring whether it can sell a version of the chip in China while complying with U.S. export restrictions, citing demand for products at that price point.

Intel is also doubling down on manufacturing its own chips. Under CEO Lip-Bu Tan, the company plans to build Crescent Island in its own foundries, rather than outsourcing to Taiwan Semiconductor (TSMC (TSM)). That could further lower costs and give Intel more control over its supply chain. It's all part of a broader turnaround plan that investors have largely cheered, following leadership changes and cost-cutting.

But while Intel is trying to carve out a niche in AI infrastructure, Nvidia is coming after its home turf: the PC. At Computex, Nvidia unveiled the RTX Spark Superchip, a new processor developed with MediaTek and paired with Microsoft's (Microsoft (MSFT)) Windows. Nvidia CEO Jensen Huang described it as a reinvention of the personal computer, designed to power AI agents that can perform tasks across applications with minimal human input. The chip targets AI developers, creators, and gamers initially, before expanding to a broader consumer base, according to the Financial Times.

That's a direct challenge to Intel, AMD, Qualcomm (Qualcomm (QCOM)), and Apple (Apple (AAPL)) in the PC processor market. Huang said Nvidia and Microsoft spent three years developing the platform, and he's betting that AI assistants will become the primary way users interact with their devices. For Intel, which is already struggling to regain momentum in AI infrastructure, this adds a new competitive front in its core client computing business.

So where does that leave Intel stock? The next big catalyst is the July 23, 2026 (estimated) earnings report. Analysts expect earnings per share of 19 cents, up from a loss of 10 cents a year ago, and revenue of $14.40 billion, up from $12.86 billion. The stock carries a Hold rating with an average price target of $77.65, though recent analyst moves have been bullish: Citigroup raised its target to $130 on May 18, Benchmark to $140 on the same day, and Mizuho to $124 on May 12.

Technically, Intel shares are pulling back after a powerful 12-month run. In premarket trading Monday, the stock was down 4.87% at $109.10. It's still 29.2% above its 50-day moving average ($82.45) and 116.6% above its 200-day moving average ($49.17), with a golden cross from August 2025 still intact. But the near-term picture is softer: shares are 7.5% below the 20-day moving average ($115.16), and the MACD is below its signal line with a negative histogram, signaling fading momentum. The key support level to watch is $102.50, where buyers previously stepped in.

In plain English, Intel's long-term trend is still up, but the stock needs to find a footing before the next leg higher. For now, the story is about a company trying to fight on two fronts: cheaper AI chips against Nvidia, and a PC defense against Nvidia's new offensive. It's a tough spot, but at least Intel has a plan.