The Architects of AI, A Small Circle of Minds Shaping Our Future
As global AI leaders convene in New Delhi for the AI Impact Summit, it is difficult to ignore the web of interconnected companies and common mentors that bind them together. Sam Altman, the founder of OpenAI, is in India for the summit, as is Anthropic’s Dario Amodei. Also in town is Yann LeCun, a mentor to many AI leaders. Their presence is a reminder that the world’s most consequential artificial intelligence research is being conducted by a strikingly small circle of scientists—people who studied under the same mentors, taught in the same rooms, and kept recruiting from each other’s labs.
To understand where AI is going, one needs to understand a simple fact: virtually every major breakthrough of the past 15 years can be traced back to a handful of key figures and three institutions: Stanford University, the Massachusetts Institute of Technology, and the University of Toronto. The larger AI ecosystem is deeply interconnected, with DeepMind, OpenAI, and Stanford/MIT serving as the primary hubs, characterised by close, often fluid, talent flows.
The Transformer That Changed Everything
If Alan Turing’s 1950 paper laid the foundation for AI and language processing, it was Google’s transformer paper—”Attention Is All You Need,” written by eight Google researchers in early 2017—that set the stage for all the foundational large language models that we have today. The transformer architecture is the common link in the development of LLMs and generative AI tools. Every chatbot, every image generator, every AI system that can hold a conversation traces its lineage back to that paper.
The eight authors of that paper have since scattered across the industry, founding companies, joining labs, and training the next generation. Their work demonstrated that a simple idea—allowing models to pay “attention” to different parts of input data—could unlock capabilities that had eluded researchers for decades.
The PayPal Parallel
There is a historical parallel that helps explain the current structure of the AI industry. After eBay acquired PayPal in 2002, a close-knit group of former PayPal founders and employees—including Peter Thiel, Elon Musk, and Reid Hoffman—went on to found or fund many of Silicon Valley’s most influential companies. Their ventures include YouTube, LinkedIn, Tesla, Yelp, SpaceX, and Palantir. The PayPal Mafia, as they came to be known, demonstrated how a concentrated group of talent could shape an entire industry.
Something similar is happening in AI. OpenAI was founded by Elon Musk, Sam Altman, Ilya Sutskever, Greg Brockman, Wojciech Zaremba, and John Schulman. Thiel was the main backer and mentor of Altman, providing a large part of the $5 million that made up Altman’s initial venture fund. When OpenAI launched in 2015, Thiel was one of the founding investors, pledging $1 billion.
The most notable startups founded by OpenAI alumni include brother-sister duo Dario Amodei and Daniela Amod ei of San Francisco-based Anthropic. The siblings left OpenAI in 2021 to form their own startup with a focus on AI safety. Anthropic has rapidly grown to become OpenAI’s biggest rival. OpenAI co-founder and chief scientist Ilya Sutskever left OpenAI in May 2024 after he was reportedly part of a failed coup to replace CEO Sam Altman. Shortly afterward, Sutskever co-founded Safe Superintelligence. Mira Murati, OpenAI’s CTO, left last year to found Thinking Machines. Arvind Srinivas worked as a research scientist at OpenAI for a year until 2022, when he left to co-found AI search engine Perplexity.
Google DeepMind is a major “founder factory” in the AI industry, with over 200 former employees having gone on to establish their own startups, with OpenAI coming in a close second. Its co-founder, Mustafa Suleyman, is now CEO of Microsoft AI.
The Intellectual Godfathers
Behind these companies and founders stand a small number of intellectual godfathers. Geoffrey Hinton (University of Toronto/Google), Yann LeCun (New York University/ex-Meta), and Yoshua Bengio (Université de Montréal) shared the Turing Award in 2018 for conceptual and engineering breakthroughs that made deep neural networks a critical component of computing.
Hinton’s students include Ilya Sutskever (OpenAI/Safe Superintelligence) and Alex Krizhevsky (AlexNet). LeCun, Meta’s chief AI scientist until January this year, was in fact a postdoctoral researcher under Hinton. The lineage is direct and traceable.
After the success of AlexNet—a deep neural network that crushed the competition in the 2012 ImageNet challenge—Hinton, Sutskever, and Krizhevsky formed a company to commercialise their work. In 2013, Google acquired the company for $44 million and hired all three. Hinton stayed at Google Brain for the next decade while maintaining his faculty position at the University of Toronto. Sutskever worked at Google Brain for a few years before co-founding OpenAI in 2015. Krizhevsky eventually left Google to work at startups before becoming a research advisor at various AI companies.
The Ng Connection
Now consider Andrew Ng. He founded Google Brain in 2011, the research division that spawned countless AI leaders. He was chief scientist at Baidu, where he built an AI group of 1,300 people. He co-founded Coursera, which has taught millions of people machine learning. He continues to teach at Stanford as an adjunct professor. And critically, he has been a post-doctoral adviser, collaborator, and mentor to many of the field’s rising stars.
In 2012, Ilya Sutskever spent time with Ng as a postdoc after completing his PhD at Toronto under Hinton. That brief connection marked the start of a new era. Ng has also taught Sam Altman at Stanford.
At Stanford, Fei-Fei Li created ImageNet—the massive dataset of 14 million labelled images that made AlexNet possible. Without her dataset, Hinton’s work at Toronto wouldn’t have had the fuel it needed. She was also the PhD advisor of computer vision expert Andrej Karpathy. He interned at Google Brain (2011), Google Research (2013), and DeepMind (2015)—gaining exposure to the world’s three largest AI research operations. When he graduated in 2015, he joined OpenAI as a founding member. He left the company to join Tesla in 2017 to lead its autopilot program. Karpathy is also well-known for his YouTube videos explaining core AI concepts. He left Tesla in 2024 to found Eureka Labs, a startup building AI teaching assistants.
The Circular Funding Web
The interconnectedness of AI extends beyond people to capital. AI funding cycles are typically structured through a combination of capital and rounds, with a big portion of funding coming from a mix of big tech players (hyperscalers) and private equity firms.
Microsoft is one of the largest backers, with a multi-billion dollar partnership with OpenAI and investments in Anthropic and Anysphere. Chipmaker Nvidia has been a major investor across the downstream AI ecosystem, with equity exposure to Anthropic and Anysphere. Google and Google Ventures have invested in Anthropic and AI21 Labs, while Meta has invested in Scale AI. Amazon has backed Anthropic and invested in robotics player Figure AI.
Among funds, SoftBank led a massive $40 billion investment into OpenAI in 2025, even as Andreessen Horowitz has investments in OpenAI and Mistral AI. Y Combinator is another major player in early-stage funding of startups.
These interlinked funding plans have now catalysed an industry-wide term—circular trading—that is being attributed to the AI bubble. This comes amid a pickup in large partnership announcements across AI model developers, hyperscalers, and chip companies. The recent Stargate Project, announced in January 2025, brings together OpenAI, SoftBank, Oracle, and MGX as initial equity funders to build AI infrastructure for OpenAI at a cost of $500 billion over four years. OpenAI has also committed to spending $300 billion on data centre capacity from Oracle, while Nvidia has committed $10 billion to Anthropic and $100 billion to OpenAI. The money flows in circles, but it is building the infrastructure of the future.
The Stakes
The small circle of minds shaping AI has produced remarkable progress, but it also raises questions. When so much power is concentrated in so few hands, who decides the direction of the technology? Who sets the ethical boundaries? Who benefits?
The interconnectedness of the AI world is both a strength and a vulnerability. It enables rapid progress because ideas flow quickly among trusted collaborators. But it also means that groupthink can set in, that alternative approaches can be neglected, that the field can become insular.
As the AI Impact Summit convenes in New Delhi, these questions are worth keeping in mind. The architects of AI are building our future. Understanding who they are, how they are connected, and what they value is essential for anyone who wants to have a say in what that future looks like.
Q&A: Unpacking the AI Ecosystem
Q1: What is the significance of Google’s 2017 transformer paper “Attention Is All You Need”?
The transformer paper introduced a new architecture for neural networks that allowed models to pay “attention” to different parts of input data. This innovation proved crucial for developing large language models (LLMs) and generative AI tools. Virtually all modern AI systems—from ChatGPT to Gemini—are built on transformer architecture. The paper’s eight authors have since scattered across the industry, founding companies and training the next generation of AI researchers, making it the foundational document of the current AI era.
Q2: How does the “PayPal Mafia” parallel help explain the current AI landscape?
After eBay acquired PayPal in 2002, a close-knit group of former PayPal employees went on to found or fund many of Silicon Valley’s most influential companies (YouTube, LinkedIn, Tesla, etc.). Similarly, a small circle of AI researchers from OpenAI, Google DeepMind, and Stanford/MIT are now founding and funding the next generation of AI startups. Alumni from these organizations have created Anthropic, Safe Superintelligence, Thinking Machines, Perplexity, and dozens of other companies, demonstrating how concentrated talent pools can shape an entire industry.
Q3: Who are the intellectual godfathers of modern AI, and what is their relationship to current leaders?
Geoffrey Hinton, Yann LeCun, and Yoshua Bengio shared the 2018 Turing Award for their work on deep neural networks. Hinton’s students include Ilya Sutskever (OpenAI/Safe Superintelligence) and Alex Krizhevsky. LeCun was a postdoctoral researcher under Hinton. Andrew Ng, who founded Google Brain and co-founded Coursera, has mentored countless AI leaders and taught Sam Altman at Stanford. Fei-Fei Li created ImageNet and advised Andrej Karpathy. The lineage is direct and traceable—today’s leaders studied under yesterday’s pioneers.
Q4: What is “circular trading” in the context of AI funding, and why does it matter?
“Circular trading” refers to the interlinked funding relationships among AI model developers, hyperscalers (big tech), and chip companies. For example, Microsoft backs OpenAI, which uses Microsoft’s cloud and buys Nvidia chips. Nvidia invests in Anthropic, which uses Amazon’s cloud and Nvidia chips. SoftBank invests in OpenAI, which partners with Oracle. These circular flows of capital create a tightly coupled ecosystem where companies are simultaneously customers, investors, and partners. Critics worry this creates an AI bubble, while proponents argue it’s necessary to fund the massive infrastructure required for AI development.
Q5: What are the implications of AI research being concentrated in a small circle of interconnected people and institutions?
The concentration has both benefits and risks. Benefits include rapid progress through trusted collaborations, shared knowledge, and efficient talent development. Risks include potential groupthink, neglect of alternative approaches, and concentration of power over a transformative technology in very few hands. Understanding who these people are, how they are connected, and what they value is essential for anyone who wants to have a say in shaping AI’s future. The interconnectedness is both a strength and a vulnerability.
