OpenAI Chip Hire #2 Clive Chan Joins Anthropic

Who is Clive Chan, and why does his move matter?

Clive Chan, the second hardware hire on OpenAI's custom chip team, left the company and joined Anthropic during the week of June 7, 2026. Chan posted the announcement on X (formerly Twitter) on June 7, confirming his first week at Anthropic had already begun.

Chan is a hardware and AI infrastructure engineer. Before OpenAI, he worked at Google on machine learning infrastructure, at SpaceX on liquid rocket engine projects, and at Tesla on Autopilot deep-learning infrastructure — including GPU optimization, cluster scheduling, and training infrastructure. He joined OpenAI in January 2024.

What did Chan work on at OpenAI?

At OpenAI, Chan was part of the company's in-house chip development program. His work covered matrix multiplication (Matmul), performance analysis (Roofline Analysis), and broader hardware architecture, according to AlphaMatch's reporting on the departure.

He described the team he left in strong terms. "The density of hardware talent on that team is extraordinary, and I don't think there's a better chip design team anywhere," Chan wrote on X, as reported by Data Center Dynamics.

What is the OpenAI–Broadcom chip project?

OpenAI is co-developing a custom AI chip with Broadcom. Under the arrangement announced in an October 2025 blog post, OpenAI handles chip and system design while Broadcom manages accelerator and network system development and deployment.

You might also like

Intel EMIB Targets TSMC CoWoS in AI Packaging

EU Court Upholds €4.1B Android Fine Against

Amazon Designs Custom AI Chips for Echo, Fire TV

Microsoft Launches $2.5B AI Unit With 6,000 Staff

The first batch of racks is planned for the second half of 2026. The full project is expected to run through the end of 2029. Chan pointed to that October 2025 blog post as the primary public reference for the program's status, and added that the public "will soon see more relevant progress."

Here's what we know so far about the timeline Chan confirmed publicly:

Why did Chan leave for Anthropic?

Chan said he "hadn't been able to shake the pull to climb a new mountain from the bottom again." He cited Anthropic's talent, values, and ambition as the factors that drew him in. He also noted that within days of joining, he had already felt "an extremely intense work rhythm."

His closing statement was direct: "It's time to build."

What chips does Anthropic currently use?

Anthropic currently runs on a mix of third-party hardware. The company uses Nvidia GPUs, Google TPUs, and Amazon Trainium chips. Anthropic has not announced any plans to develop its own silicon. Reuters reported in April 2026 that the company was considering building its own chips, but no official announcement has followed.

This context matters for builders and developers tracking the AI memory chip shortage — companies like Anthropic that rely entirely on external silicon are directly exposed to supply constraints.

Is this part of a broader talent shift from OpenAI to Anthropic?

Chan's move follows a pattern. Andrej Karpathy, a co-founder of OpenAI, also announced he was joining Anthropic in the weeks before Chan's post. The two departures happened close together in mid-2026.

Anthropic closed a Series H funding round of up to $65 billion on June 1, 2026. That pushed its post-investment valuation to approximately $965 billion.

For context on how AI infrastructure investment is reshaping the hardware stack, the Micron AI memory supercycle and the broader push by AI labs to reduce dependence on third-party silicon are directly connected trends. Chan's move also echoes the kind of talent competition driving stories like SpaceX vets raising $50M for data center infrastructure plays.

Developers building on top of Anthropic's models should also watch how the company's infrastructure choices evolve — custom silicon decisions affect latency, cost, and availability at the API level. The IBM 0.7nm Nanostack story shows how quickly the underlying compute landscape is shifting for all AI providers.

What happens next?

Chan confirmed he could not disclose details about OpenAI's chip program beyond what the company has already made public. He pointed to the October 2025 OpenAI–Broadcom blog post as the key reference, and said the public "will soon see more relevant progress." The first batch of Broadcom chip racks remains on schedule for the second half of 2026.