OpenAI is looking to hire a principal network engineer/architect for its centi-billion-dollar Stargate project.

The previously unreported hire, part of OpenAI’s Industrial Compute team, is expected to "define and evolve OpenAI’s global network architecture."

In a job listing, the company said that the new hire will "set the routing, topology, and control-plane strategy that shapes how OpenAI’s network operates across clouds, PoPs, and self-built data center locations."

Historically, OpenAI primarily used Microsoft's Azure cloud to train and inference its models, as part of an early investment deal by the hyperscaler. But, as the company's ambitions and compute requirements grew, it has significantly expanded its compute portfolio, primarily through its Stargate joint venture.

Launched earlier this year with a $500 billion spending target, and backed by Oracle, MGX, and SoftBank, the Stargate project aims to deploy tens of gigawatts of data center compute around the world.

Specifics of what constitutes Stargate and what are standalone cloud contracts are lacking, with OpenAI employees themselves telling SDxCentral that many of those details are still being defined internally.

Core to OpenAI's newer compute strategy is a record deal with Oracle, expected to be as much as $300 billion over five years. The company is also using Google Cloud, and has committed to spend as much as $22.4 billion on CoreWeave services. Alongside this, OpenAI is reportedly considering spending as much as $100 billion on backup cloud servers.

In late September, OpenAI also signed a letter of intent to deploy "at least" 10GW of AI data centers with Nvidia hardware, backed by as much as $100 billion from Nvidia.

How much of these overlapping commitments and investments will come in the form of known Stargate projects remains unclear, but the company has announced six U.S.-based Stargate data center projects, as well as facilities in the United Arab Emirates (UAE), U.K., and Norway.

With a mixture of upcoming self-build, cololocated wholesale contracts with the likes of Nscale, and cloud deals with Oracle, the company's direct control over its networking infrastructure is expected to vary from project to project.

Stargate's first data center site, a campus in Abilene, Texas, went live earlier this week at 200MW of IT capacity. It is expected to grow to 1.2GW, connected by a single integrated network fabric over Oracle Cloud.

The principal network engineer is expected to develop the "multicloud interconnect strategy and BGP policy, including peering, security posture, route leak mitigation, and interoperability across heterogeneous networks."

They will also "model and plan for evolving topologies to support new workload patterns (e.g., multiterabit east–west, low-latency fan-in/fan-out)."

The network is expected to connect millions of GPUs and AI systems worldwide, with OpenAI expected to launch its own inference chip co-developed with Broadcom as soon as next year.

As training runs grow larger, multidata hall clusters are expected to shift to multicampus training, requiring advances in network infrastructure.

The role reports to Anuj Saharan, OpenAI's industrial compute lead. On LinkedIn, Saharan noted: "Stargate requires the world's largest network, come help build it from scratch!"

The job offers $393,00 to $495,000 plus equity, working out of San Francisco or Seattle.