Let me say something that will sound wrong at first.
Intel’s stock is up 210% over the past year. They just locked in a multiyear deal with Google. They bought back full ownership of their Ireland fabrication facility for $14.2 billion. They joined Elon Musk’s Terafab project alongside SpaceX and Tesla. By every short-term measure, Intel is having a moment.
And I still think you’ll forget them in 15 years.
Not because the comeback isn’t real. It is. But because of what this Google deal actually reveals about the box Intel is building itself into.
Here’s the deal: Google and Intel announced an expanded multi-year collaboration on April 9th. Intel’s Xeon processors will continue powering Google Cloud across AI inference and general-purpose workloads. They’re deepening co-development of custom Infrastructure Processing Units — IPUs — that handle networking, storage, and security functions at hyperscale. Google’s AI infrastructure chief called Intel’s roadmap a source of confidence for meeting growing performance demands.
That last sentence is the one to read carefully.
Google’s infrastructure chief is expressing confidence in Intel’s roadmap. Meaning Intel’s roadmap — the decisions Intel makes about what to build and when — now answers, at least partly, to what Google needs. Multiple processor generations, aligned to one customer’s requirements. That’s not a partnership of equals. That’s how you become infrastructure. Useful, necessary, and invisible.
This is how platform companies work. Apple ditched Intel for their own M-series chips the moment they had the capability to do so. Amazon’s Graviton processors are eating Intel’s data center market. Microsoft is moving toward custom ARM. Google already has TPUs for training and is now co-developing IPUs with Intel for inference. Each of these moves follows the same logic: own the silicon, own the economics.
Intel is not being destroyed by this trend. They’re being absorbed into it. They become the company that builds the components that power other companies’ platforms. That’s a real business. It generates real revenue. But it is not the Intel that defined computing for thirty years — the company whose architecture everyone else had to support, whose roadmap the industry followed, whose brand you knew even if you never thought about chips.
The current recovery is genuine because the AI deployment era genuinely needs CPUs. The industry spent three years obsessing over training compute — GPUs, NVIDIA, H100s. But running AI models in production at scale requires orchestration, inference infrastructure, and exactly the kind of general-purpose compute Intel specializes in. That tailwind is real and Intel is riding it correctly.
The problem is the ceiling. Intel is winning on inference-era CPU demand. But they’re not building a platform. They’re supplying components to companies that are. And the companies buying those components — Google, Microsoft, Amazon — are all simultaneously developing the next generation of silicon that will eventually reduce their dependency on Intel again. Google’s TPUs already handle training. The IPU co-development is for inference. What comes after inference?
Whatever it is, Google will build it themselves. That’s what platform companies do.
So yes, Intel is back. The numbers are real, the momentum is real, and anyone who shorted this company at its lows is having a painful year. But there is a difference between a recovery and a reinvention. Intel is recovering. They are not reinventing their position in the stack.
In 15 years, Intel will probably still exist. They’ll have revenue. They’ll have customers. But they’ll be the company powering someone else’s AI infrastructure, the way a utility powers someone else’s building.
Necessary. Profitable. Forgotten.
The Google deal is not the beginning of the comeback. It’s the ceiling of it.