Aidan Gomez
Last reviewed
May 31, 2026
Sources
No citations yet
Review status
Needs citations
Revision
v2 · 3,518 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Suggest edit
CtrlK
Last reviewed
May 31, 2026
Sources
No citations yet
Review status
Needs citations
Revision
v2 · 3,518 words
Add missing citations, update stale details, or suggest a clearer explanation.
Aidan N. Gomez (born 1996) is a British-Canadian computer scientist and technology executive who is the co-founder and chief executive officer of Cohere, a Toronto-based enterprise artificial intelligence company.[1][2] He is one of the eight co-authors of the 2017 paper "Attention Is All You Need", which introduced the Transformer neural network architecture that underpins most contemporary large language models, including those behind ChatGPT, Gemini and Claude.[3][4] Gomez was a 20-year-old intern at Google Brain when he contributed to the paper, making him the youngest of its eight authors.[1][5]
After working at Google Brain and pursuing doctoral studies at the University of Oxford, Gomez co-founded Cohere in 2019 together with Nick Frosst and Ivan Zhang.[1][6] Under his leadership, the company has focused on enterprise and government deployments of generative AI rather than consumer products, raising approximately US$1.6 billion in venture capital by late 2025 and reaching a valuation of about US$7 billion.[7][8][9]
Gomez is widely cited as a sceptic of the artificial general intelligence (AGI) framing that dominates much of the AI industry, instead promoting what Cohere has summarised in the slogan "ROI over AGI."[10][11] He was named to the inaugural TIME100 AI list of the most influential people in artificial intelligence in 2023[12] and was elected to the board of directors of electric vehicle manufacturer Rivian Automotive in April 2025.[13]
| Born | 1996 |
| Nationality | British-Canadian |
| Education | University of Toronto (BSc, computer science and mathematics); University of Oxford (DPhil, computer science, 2024) |
| Known for | Co-author of "Attention Is All You Need"; co-founder and CEO of Cohere |
| Title | CEO and co-founder, Cohere |
| Company | Cohere |
Aidan Gomez grew up in Brighton, Ontario, a small town of roughly twelve thousand people on the north shore of Lake Ontario in eastern Ontario, Canada.[14] He holds British and Canadian citizenship and is sometimes described in profiles as British-Canadian.[1] In interviews he has described an early interest in computers and science fiction, and has credited the open culture of online machine-learning communities and tutorials with helping him learn the field before formal study.[14][38]
Gomez studied computer science and mathematics at the University of Toronto, completing his bachelor's degree there.[1][15] During his undergraduate studies he worked with Roger Grosse and became interested in machine learning research, which led to a research internship at Google's Toronto office in the summer of 2017, where he was mentored in part by Łukasz Kaiser of Google Brain.[15][16] At Toronto he was also part of the broader community of students who interacted with Geoffrey Hinton's group at the Vector Institute, which would later supply early angel investors and the founding team of Cohere.[21][20]
In 2018, Gomez began doctoral studies at the University of Oxford as a member of the Oxford Applied and Theoretical Machine Learning (OATML) group, supervised by Yarin Gal in the Department of Computer Science and by Yee Whye Teh in the Department of Statistics.[15] He held a Clarendon Scholarship and was an Open Philanthropy AI Fellow during his time at Oxford.[15] His research focused on understanding and improving neural networks and their applications, with particular attention to model sparsity, dropout, reversible architectures and Bayesian uncertainty.[15][16] Gomez paused his Oxford studies to launch Cohere in 2019 but ultimately completed his DPhil in 2024.[1]
In 2017 Gomez joined Google Brain in Toronto as a research intern while still an undergraduate. There he was attached to a project led by researchers Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Łukasz Kaiser and Illia Polosukhin that was attempting to replace the recurrent and convolutional components of contemporary sequence-to-sequence models with a mechanism based purely on attention.[3][4]
The resulting paper, "Attention Is All You Need," was first posted to arXiv on 12 June 2017 (arXiv:1706.03762) and presented at the 31st Conference on Neural Information Processing Systems (NeurIPS) in December 2017.[3][4] The paper introduced the Transformer architecture, which uses multi-head self-attention layers in an encoder–decoder configuration and dispenses with recurrence. The authors are listed in a deliberately randomised order, with a footnote noting that they were equal contributors; Gomez is the sixth name on the list.[3]
At 20 years old, Gomez was the youngest of the eight co-authors.[1][5][12] The Transformer architecture went on to become the foundation of essentially every leading large language model of the late 2010s and 2020s, including the GPT family, Llama and Cohere's own Command series.[4][17]
While at Google Brain Gomez also co-authored "One Model to Learn Them All" (2017) with Łukasz Kaiser and others, a paper that explored training a single neural network to handle multiple tasks across modalities.[1][18]
After the publication of the Transformer paper, Gomez and his Oxford colleague Ivan Zhang continued collaborating through FOR.ai, an open online research collective that they helped co-found and that published papers on topics including text-based generative adversarial networks and model-serving efficiency.[19]
In 2019 Gomez left Google Brain and co-founded Cohere in Toronto with Zhang and Nick Frosst, a former Google Brain researcher who had worked in the Toronto lab of Turing Award winner Geoffrey Hinton.[1][6][20] The company's stated mission was to make natural language processing safe and accessible to businesses, distinguishing itself from consumer-facing competitors by selling its models as enterprise infrastructure rather than as chatbots for end users.[21]
Cohere emerged from stealth in 2021 with public APIs for text generation, embedding and classification.[21] In September 2021 the company announced a US$40 million Series A round led by Index Ventures, with participation from Section 32 and Radical Ventures and angel investments from leading AI researchers including Geoffrey Hinton, Fei-Fei Li, Pieter Abbeel and Raquel Urtasun; Mike Volpi of Index Ventures joined the board.[21][22]
Under Gomez's leadership Cohere has developed several families of foundation models and developer tools aimed primarily at enterprise customers. The company's flagship text-generation models are the Command series, including the open-weight Command R and Command R+ models released in 2024 and the more recent Command A (command-a-03-2025), a model the company says delivers higher throughput than its predecessors while running on as few as two GPUs and supports state-of-the-art translation across 23 languages.[23][24] Cohere also produces Embed, a family of multilingual text- and image-embedding models, and Rerank, a foundation model dedicated to re-ordering search results to improve retrieval accuracy.[25][26]
Cohere has also developed enterprise products that combine its models with retrieval and agent capabilities. Cohere Compass is a private embedding and indexing system for enterprise data, while North is a secure generative-AI workspace and agent platform launched for organisations such as banks and government departments.[23]
In June 2022 Cohere established a non-profit research arm called Cohere For AI (renamed Cohere Labs in April 2025), led until September 2025 by Sara Hooker, a former Google Brain researcher.[27][19] Cohere Labs is best known for the Aya project, a multilingual large language model and dataset effort that engaged more than 3,000 researchers from 119 countries and released open-weight models supporting more than 100 languages, including many languages not previously covered by major open-source models.[28][27] In October 2024 the lab released Aya Expanse in 8-billion- and 32-billion-parameter versions, open-weight multilingual models optimised for 23 languages and distributed on the Hugging Face and Kaggle platforms; Cohere reported that Aya Expanse 32B outperformed several larger contemporaries such as Llama 3.1 70B on multilingual benchmarks.[44][45]
Alongside Gomez, Frosst and Zhang, Cohere's senior leadership has expanded as the company has grown. Martin Kon, the former chief financial officer of YouTube, joined the company as president and chief operating officer in December 2022. In August 2025 Cohere appointed Joëlle Pineau, the long-time head of Meta's Fundamental AI Research (FAIR) laboratory, as its first chief AI officer, and promoted Phil Blunsom, an Oxford professor who had previously served as chief scientist, to chief technology officer.[46][47]
Following its 2021 Series A, Cohere closed a US$125 million Series B in February 2022 led by Tiger Global, with participation from Radical Ventures, Index Ventures and Section 32.[29] In June 2023 the company raised a US$270 million Series C led by Inovia Capital at a valuation of about US$2.2 billion, with participation from Nvidia, Oracle, Salesforce Ventures, DTCP, Index Ventures, Mirae Asset, Schroders Capital, Thomvest Ventures and SentinelOne.[30][31] In July 2024 Cohere raised US$500 million at a valuation of US$5.5 billion in a round that included Cisco, AMD, Fujitsu, PSP Investments and the Export Development Canada arm.[32]
In August 2025 the company announced another US$500 million round at a valuation of US$6.8 billion, led by Radical Ventures and Inovia Capital with participation from AMD Ventures, Nvidia, PSP Investments, Salesforce Ventures and the Healthcare of Ontario Pension Plan.[7][33] A US$100 million second close in September 2025 lifted the post-money valuation to about US$7 billion.[8][9] By that point Cohere had raised approximately US$1.6 billion in total venture funding.[7][8] In its August 2025 funding announcement Cohere disclosed that the company had crossed US$100 million in annualised revenue earlier in 2025, with subsequent reporting from CNBC and Crunchbase News indicating that annualised revenue had grown further by late 2025 as enterprise contracts ramped up.[33][7]
Cohere has built its commercial business around partnerships with established enterprise software vendors. In June 2023 Oracle and Cohere announced a strategic partnership under which Cohere's models would be embedded in Oracle Cloud Infrastructure's generative AI service and in Oracle's Fusion Cloud, NetSuite and industry applications; Oracle also became an investor.[34][35] Cohere's models have been integrated into Salesforce products and used to support the Notion AI workspace assistant, where Cohere Rerank powers part of Notion's enterprise search.[36][37] In 2023 the company also announced a consulting and deployment partnership with McKinsey & Company's QuantumBlack AI practice to bring generative AI to McKinsey's enterprise clients.[38]
Cohere has expanded into financial services with deployments such as North for Banking at the Royal Bank of Canada, into sovereign and defence applications with partners including Bell Canada, Thales Group and Saab AB, and into market research through its 2025 acquisition of the Vancouver-based startup Ottogrid.[32][2] In January 2025 Cohere and the Royal Bank of Canada announced a partnership to co-develop North for Banking, a generative AI platform that RBC has positioned as the first bank-wide system of its kind in Canada and which is being used internally for tasks such as policy and procedure review.[52] In July 2025 Cohere agreed a partnership with Bell Canada under which Bell will host Cohere's models and North agent-building system on its Canadian compute infrastructure and resell them to government and enterprise customers.[53] Through other agreements Cohere's models are also available on Amazon SageMaker, Google Cloud's Vertex AI and Oracle Cloud Infrastructure.[2]
In December 2024 the Government of Canada announced that Cohere would receive up to C$240 million through the new Canadian Sovereign AI Compute Strategy, the first investment under that programme. The funds support a roughly C$725 million project to build dedicated AI compute capacity in Canada, including a multi-billion-dollar data centre intended to come online in 2025 in partnership with private investors.[39][40] The investment was finalised in March 2025.[40]
On 24 April 2026 Cohere announced an agreement to merge with the German AI company Aleph Alpha, in a transaction that media reports described as valuing the combined entity at approximately US$20 billion.[48][49] The deal, which is subject to regulatory approval, was backed by the German retail conglomerate Schwarz Group with around €500 million in structured financing.[49] Speaking to TechCrunch, Gomez said that Aleph Alpha's focus on small language models, European languages and tokenizers was "really complementary" to Cohere's emphasis on larger general-purpose models, and framed the merger as a transatlantic effort to build sovereign AI capability for highly regulated sectors including the public sector, finance, defence and healthcare.[48][49]
In February 2025 a coalition of fourteen news and magazine publishers, including The Atlantic, Condé Nast, Forbes, The Guardian, Politico, Vox Media, McClatchy and the Los Angeles Times, filed a lawsuit (Advance Local Media LLC v. Cohere Inc.) in the U.S. District Court for the Southern District of New York, alleging that Cohere had used at least 4,000 of their copyrighted works to train its models and that its systems reproduced "substitutive summaries" of their articles.[50] Cohere has characterised the suit as "misguided and frivolous" and asserted that it follows responsible practices for training data.[50] In November 2025 Judge Colleen McMahon largely denied Cohere's motion to dismiss, allowing the publishers' copyright claims to proceed.[51]
In addition to the Transformer paper, Gomez has co-authored a number of research papers in machine learning. Notable work includes "The Reversible Residual Network: Backpropagation Without Storing Activations" (2017), which introduced a memory-efficient training technique allowing very deep networks to be trained without storing all intermediate activations, and "Targeted Dropout" (2018), proposed as a structured pruning method for neural networks that selectively drops weights expected to be unimportant.[16][41] His Oxford research with Yarin Gal and Yee Whye Teh addressed neural network sparsity, uncertainty and training efficiency.[15][16] Gomez also co-authored "One Model to Learn Them All" (2017) with Łukasz Kaiser, Ashish Vaswani, Noam Shazeer and others, an early demonstration that a single neural network could be trained jointly to perform tasks across modalities including speech recognition, image classification and machine translation.[18]
After founding Cohere, Gomez has continued to appear as a senior or supervisory author on Cohere Labs publications, including papers describing the Aya multilingual models and the Command R family.[28][45] His Google Scholar profile records that the "Attention Is All You Need" paper has accumulated more than 100,000 citations, making it one of the most cited papers in artificial intelligence and a foundational reference in machine learning curricula.[41][17]
The Transformer architecture introduced in that paper has come to underpin virtually all leading large language models, including OpenAI's GPT series, Google's Gemini family, Anthropic's Claude, Meta's Llama and Cohere's own Command models, as well as protein-structure systems such as AlphaFold 2 and many vision and multimodal models.[4][17] Gomez has frequently been asked in interviews whether he expected the paper to have such an impact; he has typically responded that the team was excited about the architecture's elegance but did not anticipate that it would become the dominant backbone of generative AI a decade later.[11][38]
Gomez is a frequent speaker at AI conferences and on technology podcasts and is widely identified in business media as one of the most prominent advocates for the enterprise-oriented approach to generative AI. He has repeatedly argued that the most economically significant uses of large language models will not be consumer chatbots or speculative attempts to build artificial general intelligence, but rather "boring" enterprise applications such as automating document review, customer-support triage and code documentation.[11][10]
In interviews he has criticised the prevailing "AGI" framing of progress in AI as a distraction and has emphasised that Cohere positions itself differently from competitors such as OpenAI and Anthropic by focusing on private-deployment software licences rather than on consumer products operating on negative unit economics.[10][11] Cohere's chief AI officer Joëlle Pineau, who joined the company from Meta in 2025, has described the company as "a very low drama company" with a slogan of "ROI over AGI."[10]
Gomez has also commented publicly on AI policy and on the role of Canada in the global AI economy, advocating for sovereign compute capacity and arguing that Canada has the talent base to remain a centre of AI development if it invests in domestic infrastructure.[39][42] At the World Economic Forum's annual meetings in Davos he has been a regular panellist on the future of generative AI, telling reporters in 2024 that he expected enterprise deployments to "take off" as customers moved beyond proofs of concept and warning in 2026 that companies were spreading themselves too thinly across AI projects rather than focusing on a small number of bets.[54][55] In discussions of AI regulation Gomez has argued that, because language is a "horizontal" technology, it should be regulated vertically, by existing sector regulators in finance, health care and other fields, rather than through a single overarching AI law.[55]
The McKinsey & Company "Humans behind AI" profile of Gomez describes him as motivated less by speculative claims about superintelligence than by a desire to make AI a reliable productivity tool for organisations, summarising his view that "the most exciting use cases" of large language models will be in well-understood enterprise workflows.[38]
In April 2025 Rivian Automotive announced that Gomez had been elected as an independent Class II director, citing his AI expertise as the company expands its use of artificial intelligence in vehicles and manufacturing.[13][43]