The architect of the Transformer just joined OpenAI — weeks before its IPO

Noam Shazeer, one of the eight researchers who co-authored the 2017 paper that underpins virtually every modern large language model, is leaving Google DeepMind for OpenAI. A former Trump administration AI policy official is coming aboard the same week.

Written by Lena Armitage · Bureau Tech · Jun 19, 2026

The Transformer's co-inventor is switching sides

Noam Shazeer is one of eight researchers credited on "Attention Is All You Need," the 2017 Google paper that introduced the Transformer — the neural network architecture that underpins GPT-4, Gemini, Claude, and essentially every other large language model (LLM) that matters commercially today. He has now left Google DeepMind to join OpenAI, according to reporting by TechCrunch.

That's a meaningful defection. Google has spent years and considerable capital trying to reassert itself in AI after OpenAI's ChatGPT launch caught the company flat-footed in late 2022. Losing one of the architects of the foundational technology — to the company that embarrassed them — is not a good look, whatever Google's internal bench strength.

It's worth being precise about what Shazeer's hire does and doesn't signal. Brilliant researchers don't automatically translate into better products; OpenAI already employs a deep roster of top-tier AI scientists. What the hire does signal, clearly, is that OpenAI can still attract talent at the very top of the field — and that it's willing to make that point loudly in the weeks before going public.

A policy hire that isn't subtle

In the same week, OpenAI brought on Dean Ball, who worked on AI policy during the first Trump administration. Ball has written and spoken extensively on AI governance, generally from a perspective skeptical of heavy-handed federal regulation.

The timing is not coincidental. An IPO puts OpenAI under a different kind of scrutiny than it has faced as a private company — from the SEC, from institutional investors doing due diligence, and from a Congress that has shown increasing interest in AI oversight. Having someone with established relationships in the current administration's policy orbit is a straightforward risk-management move.

What Ball will actually do at OpenAI — and how much influence he'll have over the company's public policy positions — isn't clear from the available reporting.

What this looks like from the outside

Taken together, the two hires paint a picture of a company that knows exactly what story it wants to tell investors: we have the best researchers, and we know how to operate in Washington.

Whether that story holds up under IPO scrutiny is a separate question. OpenAI's path to profitability, its unusual corporate structure (it converted from a nonprofit to a capped-profit entity and is now pursuing a full for-profit conversion), and the competitive dynamics of a market where Google, Anthropic, Meta, and others are all spending aggressively — none of that changes because of two new hires, however prominent.

But optics matter in IPO windows, and OpenAI clearly understands that.

Key takeaways

Noam Shazeer co-authored 'Attention Is All You Need' (2017), the paper that introduced the Transformer architecture now used in GPT, Gemini, Claude, and most other leading AI systems.
Shazeer had been at Google DeepMind; his departure is a significant talent win for OpenAI and a notable loss for Google at a moment of intense competition between the two companies.
Dean Ball, a former AI policy official in the Trump administration, joins OpenAI in the same week — a hire that reads as deliberate positioning ahead of regulatory scrutiny tied to an IPO.
Both hires arrive in the run-up to OpenAI's anticipated IPO, where technical pedigree and policy relationships carry real weight with institutional investors and regulators.
The timing is notable, but the strategic impact of individual hires — however prominent — on a company of OpenAI's scale is difficult to assess from the outside.

FAQ

What is the Transformer architecture and why does it matter?

The Transformer is a neural network design introduced in a 2017 paper by researchers at Google. It solved key limitations of earlier sequence-modeling approaches by using a mechanism called 'self-attention,' which allows models to weigh the relevance of different parts of an input simultaneously rather than processing it sequentially. Virtually every major large language model in commercial use today — including OpenAI's GPT series, Google's Gemini, and Anthropic's Claude — is built on this architecture.

Who is Noam Shazeer?

Shazeer is one of the eight co-authors of 'Attention Is All You Need,' the foundational 2017 paper that introduced the Transformer. He spent much of his career at Google and was most recently at Google DeepMind before joining OpenAI.

Who is Dean Ball and what is his background?

Dean Ball is a policy researcher and writer who worked on AI policy during the first Trump administration. He has been a vocal commentator on AI governance, generally favoring approaches that limit prescriptive federal regulation of AI development.

When is OpenAI's IPO expected?

OpenAI has not announced a firm IPO date as of this reporting. The company has been publicly preparing for a public offering, and the timing of these hires suggests it is in an active pre-IPO phase, but specific dates have not been confirmed.

Does hiring prominent researchers guarantee better AI models?

Not directly or automatically. Research talent contributes to long-term capability development, but the relationship between individual hires and near-term product outcomes at a company of OpenAI's scale is not straightforward. The hire is more immediately significant as a signal about OpenAI's ability to attract top talent — and as a competitive loss for Google DeepMind.

Citations

OpenAI is bringing on some big guns in the lead-up to its IPOOpenAI hired Transformer co-inventor Noam Shazeer from Google DeepMind and former Trump AI policy official Dean Ball in the same week, ahead of its IPO.
Attention Is All You Need (Vaswani et al., 2017)Noam Shazeer is one of eight co-authors of the 2017 paper that introduced the Transformer architecture, which underpins most modern large language models.
TechCrunch AI coverage feedSource feed used for research aggregation on this story.