Six months ago, Intel was the punchline of every semiconductor analyst note. The stock had cratered. The CEO was gone. The foundry business was bleeding cash. Nvidia and AMD had taken the AI spotlight, and Intel looked like the Blackberry of chips: a former giant watching the future happen without it.

Then Google showed up with a multiyear deal and a very large checkbook.

On Thursday, Google and Intel announced an expanded partnership for Google Cloud to deploy Intel Xeon 6 processors across its AI infrastructure. The deal covers AI inference, cloud computing, and general-purpose workloads. The companies will also deepen their co-development of custom infrastructure processing units (IPUs), specialized chips that handle data center management tasks. Intel declined to share pricing, which in corporate speak means the number is large enough to be competitive intelligence.

Intel stock surged on the news. And for good reason: this is not charity. This is Google making a calculated bet that CPUs are about to have their revenge.

Here is the industry shift that makes this deal matter. The AI boom so far has been a GPU story. Training large language models requires massive parallel processing, and Nvidia GPUs are the best tool for that job. But the industry is now transitioning from training AI to deploying AI. Running trained models at scale, what the industry calls inference, does not always need a $40,000 GPU. It often needs a good CPU.

"AI is reshaping how infrastructure is built and scaled," Intel CEO Lip-Bu Tan said in the announcement. "Scaling AI requires more than accelerators. It requires balanced systems. CPUs and IPUs are central to delivering the performance, efficiency and flexibility modern AI workloads demand."

Translation: everyone spent two years buying GPUs. Now they need CPUs to actually run the things they built. And Intel has been making data center CPUs longer than most AI companies have existed.

Google has used Intel Xeon processors for nearly three decades. That relationship predates Google Cloud, Android, and YouTube. It predates Gmail. The fact that Google is doubling down on Intel for AI workloads specifically signals that Intel has Xeon 6 is competitive enough for the inference era.

The timing is also telling. SiFive just raised $400 million to build RISC-V data center CPUs, with Nvidia as an investor. Arm just built its first in-house chip. The CPU market is suddenly the hottest space in semiconductors because AI inference is creating massive new demand. Intel, for all its problems, still ships more data center CPUs than anyone.

For Google, the logic is straightforward: diversification. Google builds its own TPUs for AI training. It uses Nvidia GPUs for heavy workloads. Now it is locking in Intel CPUs for the inference and general computing layer. That is three silicon suppliers for three different needs, and none of them has enough leverage to hold Google hostage on pricing.

For Intel, this is validation at the exact moment the company needed it most. Lip-Bu Tan took over as CEO promising a comeback built on execution rather than hype. A multiyear deal with the world is largest cloud provider is not proof of concept. It is proof of orders.

The AI chip war was supposed to be Nvidia versus everyone. It is turning into something much more interesting: a three-way split between GPUs for training, CPUs for inference, and custom silicon for everything in between. Intel just got handed the role it was born for. Whether it can execute is another question entirely.