Skip to content
Advertisement
Oracle Cloud plans to offer Arm AGI CPU instances in 2027

Oracle Cloud plans to offer Arm AGI CPU instances in 2027

Arm's first self-designed chip is headed to Oracle Cloud Infrastructure, where it will join Ampere and NVIDIA Vera processors

Arm’s first homegrown chip now has a confirmed cloud destination. On October 28, 2026, Arm revealed that Oracle Cloud Infrastructure (OCI) plans to launch Arm AGI CPU instances in fall 2027.

For a company long associated with the designs inside other people’s chips, that is a notable moment. Arm is now shipping silicon with its own name on the box, and one of the major cloud providers wants to rent it out.

What Oracle is planning

OCI already offers Arm-based computing. Its current lineup includes Ampere processors and NVIDIA’s Vera CPUs.

The AGI CPU would be the third Arm option on that menu. Oracle is positioning it for two kinds of jobs: general-purpose cloud workloads and agentic AI services.

Agentic AI refers to systems that do more than answer questions. They plan, call tools, and carry out multistep tasks. That kind of software generates a lot of ordinary computing work around the AI model itself, which is where a strong general-purpose CPU earns its keep.

The timing is a long runway. Instances are slated for fall 2027, roughly a year after the announcement. In cloud infrastructure, that is a normal lead time, not a delay.

Inside the AGI CPU

Arm first unveiled the AGI CPU in March 2026. It is the company’s first self-designed silicon, and it is built on Arm’s Neoverse IP architecture.

Advertisement

The headline specs are hefty. The chip packs up to 136 Neoverse V3 cores and is manufactured on TSMC’s 3nm process.

Power is capped at a 300W envelope. That figure tells data center operators how much electricity and cooling each chip will demand, which is a very real line item when you run thousands of them.

On the plumbing side, the AGI CPU supports 12 DDR5 memory channels and PCIe Gen6 connectivity. Twelve lanes means less traffic when 136 cores are all hungry at once.

PCIe Gen6 is the connection standard that links the CPU to accelerators, storage, and networking gear. In an AI-heavy data center, where the CPU often coordinates GPUs and other hardware, a faster link helps keep everything fed.

Then there is the performance pitch. Arm claims the AGI CPU delivers more than 2x the performance of equivalent x86 systems in specific configurations.

That qualifier is doing real work. “Specific configurations” means the advantage applies to particular setups Arm has tested, not every workload a customer might throw at it.

How Oracle got here

Oracle did not arrive at this decision overnight. The company joined the Arm AGI CPU ecosystem on June 2, 2026, during COMPUTEX.

So the arc runs like this: Arm unveils the chip in March, Oracle signs on to the ecosystem in June, and the deployment plan surfaces in late October. For enterprise hardware, that is a fairly brisk courtship.

What this means

For Arm, OCI is a validation point. Designing your own silicon is one thing. Getting a hyperscale-class cloud to plan instances around it is the kind of commitment that tells other buyers the chip is meant for production, not just press releases.

It also puts Arm in an unusual spot. OCI’s Arm lineup will include chips from Ampere and NVIDIA, both built on Arm technology, alongside Arm’s own processor. Arm is effectively sharing a shelf with companies that build on its architecture.

For the x86 camp, the more than 2x performance claim is the obvious pressure point. If independent testing backs up even part of that advantage in real cloud workloads, x86 vendors will face sharper questions about performance and efficiency from data center buyers.

There are open questions, too. Pricing for the instances has not been disclosed. Neither have independent benchmarks comparing the AGI CPU against Ampere and Vera on OCI’s own infrastructure.

The fall 2027 date is also a plan, not a launch. Until instances actually go live, the AGI CPU’s cloud story remains a promising roadmap item rather than a proven product.

What to watch between now and then: whether additional cloud providers announce similar plans, whether Arm publishes broader performance data beyond its specific configurations, and how Oracle prices the new instances against its existing Arm lineup.

Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.