General · EdgepediaSeptember 17, 20267 min read

Aidan Gomez

Aidan Gomez is a Canadian computer scientist who co-authored the 2017 paper "Attention Is All You Need", which introduced the transformer architecture underlying modern large language models, and who is the co-founder and chief executive officer of the enterprise AI company Cohere.¹^•² He was a 20-year-old intern at Google when the transformer paper was written, and he founded Cohere in 2019 with Nick Frosst and Ivan Zhang.²^•³

Key fact	Detail
Known for	Co-author of "Attention Is All You Need" (2017), the paper that introduced the transformer architecture¹
Role at the time	20-year-old Google intern on the Google Brain team when the paper was written²
Company	Co-founded Cohere in 2019 in Toronto with Nick Frosst and Ivan Zhang; serves as CEO³
Education	Bachelors, University of Toronto (Roger Grosse); doctorate, University of Oxford (Yarin Gal and Yee Whye Teh, 2018–2023)⁴
Cohere funding	$500 million round in August 2025 at a $6.8 billion valuation, led by Radical Ventures and Inovia Capital⁵
Cohere revenue	Crossed $100 million in annualized revenue as of May 2025⁶
Policy positions	Rejects extinction scenarios as science fiction; advocates evidence-based risk frameworks, independent testing, and sovereign AI for middle powers⁷

Education and early career

Gomez studied computer science at the University of Toronto as an undergraduate student of Roger Grosse.⁴ He then joined the Google Brain team, where he worked under the mentorship of Geoffrey Hinton and Łukasz Kaiser; his own website describes him as an intern of both, and as a student researcher working with Jakob Uszkoreit and Jeff Dean.⁸

In 2018 he began a doctorate at the University of Oxford under Yarin Gal of the Oxford Applied and Theoretical Machine Learning group and Yee Whye Teh of the Oxford Centre for Statistical Machine Learning, completing it in 2023.⁴ He was awarded a 2019 Open Philanthropy AI Fellowship, which supported his research for five years, and is a Clarendon Scholar.⁴ Before Cohere, he led a group of researchers at FOR.ai, a research group he founded that provided resources, mentorship, and collaboration between academia and industry.⁴^•⁹

The transformer paper

In 2017 Gomez joined the team of Google researchers writing "Attention Is All You Need", which proposed a neural network architecture, the transformer, based solely on attention mechanisms and dispensing with recurrence and convolutions entirely.¹^•¹⁰ He was one of eight co-authors, listed alongside Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Lukasz Kaiser, and Illia Polosukhin.¹ Gomez and his colleagues raced to finish the paper for a major AI conference, at one point sleeping in the office to make the deadline.¹⁰

The paper reported strong machine translation results: 28.4 BLEU on the WMT 2014 English-to-German task, and a single-model state-of-the-art 41.8 BLEU on English-to-French after 3.5 days of training on eight GPUs, a small fraction of the training cost of earlier models.¹ In a 2024 interview with the Associated Press, Gomez explained why the architecture spread: transformers scale very well, can be trained across tens of thousands of chips, and use operations that GPUs are tailored for, making them faster and more efficient than what came before.² Looking back, he told TIME in 2023 that he had no appreciation for what was to come: the team was focused on building something good at translation, and the consequences of the work surprised him.¹⁰

Cohere

Gomez co-founded Cohere in 2019 in Toronto with Nick Frosst and Ivan Zhang, colleagues from his FOR.ai research days.³^•⁹ The company builds large language models for enterprise customers rather than broad consumer foundation models, with offices in Toronto, San Francisco, New York, and London, and collaborations with companies including Google, Amazon, Oracle, and Microsoft.³^•⁵ Gomez has described the founding motivation as frustration that the models' potential for transformational change in business was not being realized, so the founders decided to do it themselves.¹²

Data privacy as the differentiator. Gomez identifies data privacy as the principal challenge for enterprises adopting generative AI: customers need to customize models without leaking data to competitors. Cohere's platform can be deployed privately within a customer's virtual private cloud, so that, in his words, not even Cohere can see the data.¹² The company also promotes retrieval augmented generation, a technique that lets a model query a customer's proprietary databases to make its answers more relevant and reliable.¹² In a January 2025 conversation at Davos, he said Cohere focuses purely on enterprise, in particular private deployments.¹³

Growth and funding. Between 2021 and 2024 Cohere raised nearly $1 billion across its Series A through D rounds.¹⁴ In August 2025 it raised a further $500 million at a $6.8 billion valuation, led by Radical Ventures and Inovia Capital with participation from AMD Ventures, Nvidia, PSP Investments, and Salesforce Ventures, and appointed former Meta AI research head Joëlle Pineau as chief AI officer.⁵ The company crossed $100 million in annualized revenue as of May 2025, roughly double the level at the start of that year, driven by demand from regulated sectors.⁶ In 2025 Cohere launched North, a turnkey AI platform for workplace productivity with agentic capabilities.¹⁴

Efficiency argument. When the Chinese startup DeepSeek released its R1 model in early 2025, Gomez said it validated Cohere's strategy that spending billions of dollars a year is not necessary to produce competitive top-tier technology, though he argued DeepSeek was not enterprise-ready because businesses want customized models and are cautious about sensitive data.¹⁵

Aleph Alpha merger. In 2026 Cohere announced a proposed merger with the German AI firm Aleph Alpha, supported by the German and Canadian governments, to create what the companies describe as the first transatlantic sovereign AI firm. The combined company would operate as Cohere with more than 1,000 staff, headquarters in Berlin and Toronto, and a combined valuation of around $20 billion, subject to regulatory approvals.¹⁶ "No government or enterprise should have to choose between capable AI and control over their technology," Gomez said of the deal.¹⁶

Views on AI risk and policy

Gomez has positioned himself against what he calls doomsday tropes, telling the Associated Press his goal is technology that is "additive to humanity".² In his writing and interviews he acknowledges real risks such as cyber threats but dismisses extinction scenarios as science fiction, calling instead for evidence-based risk frameworks, independent testing, and mandatory transparency requirements.⁷ He has criticized proposals from some AI executives, including Anthropic's Dario Amodei, that he says would grant temporary antitrust relief to a small group of frontier labs, describing such ideas as a "cartel" move and arguing that safety standards should be set by governments and a broader group of stakeholders rather than a few dominant companies.⁷

On sovereign AI, Gomez argues that sovereignty does not mean every country builds the entire supply chain, which he calls impossible and colossally inefficient, but that middle powers such as Canada, Germany, and France should pool resources to maintain a diverse supply chain so no single nation holds too much power.⁷ He has advocated building data center capacity in Canada, citing the country's clean energy grid, and argues that Canada invented foundational AI technology through publicly funded research and should commercialize it rather than cede leadership.⁷ In written testimony to a 2023 U.S. Senate AI Insight Forum, he described Cohere's independence, noting that unlike other leading developers it had not taken multi-billion dollar funding from a tech giant, and emphasized that a diverse AI ecosystem is critical to innovation and competition.¹¹

On enterprise adoption, his advice to businesses is concentration: companies often spread themselves too thin across AI pilots when they should make a handful of big bets with a strong strategy and leadership mandate.¹⁷ Looking at the technology's next steps, he has pointed to continual learning, the ability of models to learn from experience and improve over time by interacting with the world.⁷

References

Topic: Encyclopedia › Technology and the built world › Computing and digital systems › Modern AI: foundation models, generative AI, and the AI industry › AI companies, people, and products › AI founders and executives › Applied AI startup founders

Initially written Sep 17, 2026 · Reviewed: Sep 30, 2026; Oct 15, 2026 · Edited: Oct 11, 2026; Oct 15, 2026 · Last review: Oct 15, 2026

Notice something wrong?