The top 10 organizations employed 22% of people shaping AI in 2025: 51% of company researchers and 21% of university researchers
Data as of
- At the top 10 organizations (2025)
- 22%
- Top 10 share in companies (2025)
- 51%
- Top 10 share in institutes and labs (2025)
- 36%
- Top 10 share in universities (2025)
- 21%
How concentrated is the talent shaping AI, and is the field consolidating into a handful of hubs or spreading out across more organizations? Across 5,265 people with a dated work position in 2025 at 2,447 active parent organizations, we estimate that 1,167 (22%) worked at the 10 largest organizations, and that 56 organizations employed at least 50% of the field. Within each sector in 2025, the top 10 companies employed 51% of all company researchers (911 of 1,782 across 723 companies), the top 10 institutes and labs employed 36% of their sector (463 of 1,295 across 594 organizations), and the top 10 universities employed 21% of academic researchers (754 of 3,635 across 1,130 universities).
Key findings
- Across all 71 complete calendar years with at least 50 placed people (1955–2025), the share working at the annual top 10 organizations reached an all-time high of 41% in 1955 (21 of 51 people across 47 organizations) and an all-time low of 21% in 2006, 2007 and 2008 (735 of 3,534 across 1,377 organizations in 2006, 765 of 3,678 across 1,425 organizations in 2007 and 787 of 3,791 across 1,468 organizations in 2008). From 2006 to 2020, the overall top-10 share rose to 26% (1,375 of 5,237 across 2,230 organizations), and from 2020 to 2025 it fell to 22% (1,167 of 5,265 across 2,447 organizations).
- Comparing within-sector top-10 shares in 1970 (the first five-year mark with at least 25 active organizations in every sector) and 2025, the university top-10 share was 40% in 1970 (95 of 237 across 121 universities) and 21% in 2025 (754 of 3,635 across 1,130 universities), and the institute and lab top-10 share was 39% in 1970 (18 of 46 across 39 organizations) and 36% in 2025 (463 of 1,295 across 594 organizations). Within companies, the top-10 share was 57% in 1970 (24 of 42 across 28 companies), reached a high of 64% in 1979 (58 of 91 across 44 companies), and moved to 51% in 2025.
- In 2025, reaching at least 50% of each sector's placed people required 10 of 723 active companies (1% of active companies), 30 of 594 active institutes and labs (5%), and 52 of 1,130 active universities (5%). At the 1979 corporate concentration high, five companies employed at least 50% of all 91 company researchers.
- Inside the overall annual top 10, the three companies in the 2019 top 10 (679 people) moved above the six universities in the top 10 (673 people) for the first time, and top-10 companies have employed more people than top-10 universities in every year from 2019 through 2025 (714 people across five companies versus 331 across three universities in 2025).
- Between 1979 and 2025, the share of company researchers at the top three companies (Google, Microsoft and Meta) moved from 45% (41 of 91) to 31% (559 of 1,782). Over the same 46 years, companies ranked 4th through 10th moved from 22% (20 people) to 22% (388 people), and companies outside the top 10 moved from 36% (33 people across 34 companies) to 49% (871 people across 713 companies).
Sector concentration from 1970 to 2025
Across all sectors combined, the annual total was 51 placed people at 47 organizations in 1955, peaked at 5,360 people in 2023 and 2,447 active organizations in 2025, and stood at 5,265 people at 2,447 organizations in 2025. In 1955, the series start, the top 10 organizations employed 41% (21 of 51, or 41.2%), the all-time high, across 47 organizations. In 2006, 51 years after the start of the series, active organizations stood at 1,377, the share of placed people at the top 10 organizations (735 of 3,534 in 2006) reached its all-time low of 21% in 2006, 2007 and 2008, at 20.8% in each year, and the number of organizations needed to employ at least 50% of placed people was 16 in 1955 and 61 in 2006 (peaking at 68 in 1999).
Four lines compare the share of people at the top 10 organizations within companies (57% in 1970 and 51% in 2025), institutes and labs (39% in 1970 and 36% in 2025), universities (40% in 1970 and 21% in 2025), and across all organizations combined (32% in 1970 and 22% in 2025).
Source: People of Superintelligence career records from biographies and OpenAlex, as of 22 October 2026. Subsidiaries and research labs roll up to their parent organization; education roles are excluded.
Data table
| Year | Placed people | Active orgs | All top 10 | Companies top 10 | Universities top 10 | Institutes top 10 | Orgs for 50% of people |
|---|---|---|---|---|---|---|---|
| 1955 | 51 | 47 | 21 (41%) | 11 of 12 (92%) | 17 of 26 (65%) | 12 of 15 (80%) | 16 |
| 1970 | 309 | 188 | 99 (32%) | 24 of 42 (57%) | 95 of 237 (40%) | 18 of 46 (39%) | 27 |
| 1975 | 488 | 254 | 154 (32%) | 40 of 64 (63%) | 137 of 375 (37%) | 39 of 69 (57%) | 31 |
| 1980 | 714 | 362 | 210 (29%) | 64 of 104 (62%) | 183 of 547 (33%) | 56 of 105 (53%) | 35 |
| 1985 | 1,065 | 531 | 297 (28%) | 93 of 181 (51%) | 269 of 826 (33%) | 70 of 164 (43%) | 42 |
| 1990 | 1,576 | 685 | 416 (26%) | 127 of 246 (52%) | 367 of 1,225 (30%) | 108 of 271 (40%) | 49 |
| 1995 | 2,104 | 901 | 490 (23%) | 155 of 308 (50%) | 424 of 1,679 (25%) | 149 of 356 (42%) | 61 |
| 2000 | 2,679 | 1,123 | 615 (23%) | 225 of 493 (46%) | 501 of 2,099 (24%) | 177 of 479 (37%) | 66 |
| 2005 | 3,396 | 1,323 | 758 (22%) | 245 of 541 (45%) | 659 of 2,798 (24%) | 256 of 657 (39%) | 58 |
| 2006 | 3,534 | 1,377 | 735 (21%) | 250 of 554 (45%) | 628 of 2,890 (22%) | 258 of 671 (38%) | 61 |
| 2010 | 4,060 | 1,591 | 872 (21%) | 336 of 713 (47%) | 687 of 3,268 (21%) | 297 of 803 (37%) | 59 |
| 2015 | 4,653 | 1,868 | 1,093 (23%) | 525 of 1,036 (51%) | 804 of 3,620 (22%) | 321 of 938 (34%) | 61 |
| 2020 | 5,237 | 2,230 | 1,375 (26%) | 951 of 1,655 (57%) | 881 of 3,834 (23%) | 438 of 1,188 (37%) | 50 |
| 2025 | 5,265 | 2,447 | 1,167 (22%) | 911 of 1,782 (51%) | 754 of 3,635 (21%) | 463 of 1,295 (36%) | 56 |
In 1970, when all three sectors had at least 25 active organizations (121 universities, 28 companies and 39 institutes and labs), the top 10 universities employed 40% of academic researchers (95 of 237) and the top 10 companies employed 57% of corporate researchers (24 of 42). At universities, the number of active institutions was 121 in 1970 and 1,130 in 2025, and the share of academic researchers at the top 10 universities was 40% in 1970 and reached a low of 21% in 2025 (754 of 3,635 across 1,130 universities). In research institutes and standalone labs, active organizations numbered 39 in 1970 and 594 in 2025, and the top-10 share was 39% in 1970 (18 of 46), 58% at its 1977 high (45 of 78), 33% at its 2016 low (320 of 980 across 452 organizations), and 36% in 2025 (463 of 1,295).
In companies, active employers numbered 270 in 2003, 283 in 2006 and 44 in 1979, and the share of company researchers working at the 10 largest companies was a low of 43% (227 of 530) in 2003, 45% (250 of 554) in 2006, and a high of 64% (58 of 91) in 1979. At that 1979 peak, five companies employed at least 50% of all corporate AI researchers, compared with 25 universities needed to reach 50% of academic researchers and 8 organizations needed in institutes and labs.
The rise in concentration from 2006 to 2020 and the decline since 2020
From 2006 to 2020, the overall top-10 share rose by 5 percentage points, from 21% (735 of 3,534) to 26% (1,375 of 5,237), and the top-5 share was 14% (500 people) in 2006 and 19% (970 people) in 2020. Across the last 16 complete calendar years (2010–2025), the overall top-10 share rose from 21% (872 of 4,060) in 2010 to 26% in 2020 and fell to 22% (1,167 of 5,265) in 2025.
Annual lines from 2010 to 2025 show the share of all placed people at the 10 largest organizations overall (21% in 2010, 26% in 2020, and 22% in 2025) and the shares at companies (7% in 2010 and 14% in 2025), universities (11% in 2010 and 6% in 2025) and institutes (5% in 2010 and 3% in 2025) within the overall top 10.
Source: People of Superintelligence career records from biographies and OpenAlex, as of 22 October 2026. A person with positions in multiple top-10 sectors in the same year counts in each sector line and once in the combined top-10 line.
Data table
| Year | Placed people | Active orgs | All top 10 | Companies in top 10 | Universities in top 10 | Institutes in top 10 |
|---|---|---|---|---|---|---|
| 2010 | 4,060 | 1,591 | 872 (21%) | 3 orgs: 283 (7%) | 4 orgs: 429 (11%) | 3 orgs: 188 (5%) |
| 2011 | 4,177 | 1,645 | 924 (22%) | 3 orgs: 313 (7%) | 4 orgs: 440 (11%) | 3 orgs: 195 (5%) |
| 2012 | 4,286 | 1,694 | 964 (22%) | 3 orgs: 336 (8%) | 5 orgs: 506 (12%) | 2 orgs: 145 (3%) |
| 2013 | 4,397 | 1,736 | 1,003 (23%) | 3 orgs: 353 (8%) | 5 orgs: 535 (12%) | 2 orgs: 141 (3%) |
| 2014 | 4,516 | 1,812 | 1,026 (23%) | 2 orgs: 318 (7%) | 6 orgs: 600 (13%) | 2 orgs: 145 (3%) |
| 2015 | 4,653 | 1,868 | 1,093 (23%) | 2 orgs: 357 (8%) | 6 orgs: 625 (13%) | 2 orgs: 149 (3%) |
| 2016 | 4,795 | 1,959 | 1,166 (24%) | 3 orgs: 502 (10%) | 6 orgs: 625 (13%) | 1 org: 89 (2%) |
| 2017 | 4,928 | 2,044 | 1,206 (24%) | 3 orgs: 541 (11%) | 6 orgs: 635 (13%) | 1 org: 89 (2%) |
| 2018 | 5,049 | 2,107 | 1,280 (25%) | 3 orgs: 604 (12%) | 6 orgs: 662 (13%) | 1 org: 89 (2%) |
| 2019 | 5,146 | 2,173 | 1,347 (26%) | 3 orgs: 679 (13%) | 6 orgs: 673 (13%) | 1 org: 91 (2%) |
| 2020 | 5,237 | 2,230 | 1,375 (26%) | 3 orgs: 705 (13%) | 6 orgs: 666 (13%) | 1 org: 84 (2%) |
| 2021 | 5,298 | 2,275 | 1,388 (26%) | 3 orgs: 730 (14%) | 6 orgs: 661 (12%) | 1 org: 87 (2%) |
| 2022 | 5,340 | 2,343 | 1,346 (25%) | 3 orgs: 714 (13%) | 6 orgs: 625 (12%) | 1 org: 86 (2%) |
| 2023 | 5,360 | 2,425 | 1,302 (24%) | 3 orgs: 668 (12%) | 5 orgs: 541 (10%) | 2 orgs: 160 (3%) |
| 2024 | 5,337 | 2,424 | 1,244 (23%) | 4 orgs: 693 (13%) | 5 orgs: 522 (10%) | 1 org: 80 (1%) |
| 2025 | 5,265 | 2,447 | 1,167 (22%) | 5 orgs: 714 (14%) | 3 orgs: 331 (6%) | 2 orgs: 160 (3%) |
In 2006, the overall top 10 consisted of six universities (Massachusetts Institute of Technology, Carnegie Mellon University, Stanford University, University of California, Berkeley, Harvard University and University of Oxford, employing 477 distinct people, or 13% of all placed people), two companies (Microsoft and IBM, with 163 people, or 5%) and two institutes (CNRS and Max Planck Society, with 112 people, or 3%). In 2020, the six universities in the top 10 employed 666 people (13% of all placed people), the three companies in the top 10 (Google, Microsoft and Meta) employed 705 distinct people (13%), and the one institute (CNRS) employed 84 people (2%).
The sector crossover inside the overall top 10 occurred in 2019, when the three top-10 companies (679 people) moved above the six top-10 universities (673 people). Companies in the annual top 10 have employed more people than universities in the top 10 in every year since 2019. By count of organizations in the top 10, 2025 is also the first year since 1958 with at least as many companies as universities in the overall top 10: five companies (Google, Microsoft, Meta, NVIDIA and OpenAI, with 714 distinct people, or 14% of all placed people), three universities (Massachusetts Institute of Technology, Carnegie Mellon University and Stanford University, with 331 people, or 6%) and two institutes and labs (Allen Institute for AI (Ai2) and CNRS, with 160 people, or 3%). Between 2020 and 2025, the overall top-10 share fell from 26% (1,375 of 5,237) to 22% (1,167 of 5,265), and the number of organizations needed to employ at least 50% of placed people rose from 50 in 2020 to 56 in 2025.
Inside each sector's top 10 in 2025
After moving from 64% in 1979 to 51% in 2025, the corporate top-10 share stands 30 percentage points above the university top-10 share (51% versus 21%). In 2025, the top 10 companies employed 911 of 1,782 company researchers (51%), led by Google with 321 (18% of all company researchers), Microsoft with 127 (7%) and Meta with 119 (7%).
Horizontal bars show the 10 companies with the most people shaping AI in 2025, led by Google with 321 (18% of all 1,782 company researchers). Together they employ 911 distinct people (51% of company researchers).
Source: People of Superintelligence career records from biographies and OpenAlex, as of 22 October 2026. Parent companies include their subsidiary research labs.
Data table
| Sector (2025) | Rank | Organization | People | Share of sector | Cumulative sector people | Share of all placed |
|---|---|---|---|---|---|---|
| Companies | 1 | 321 | 18% | 321 (18%) | 6% | |
| Companies | 2 | Microsoft | 127 | 7% | 448 (25%) | 2% |
| Companies | 3 | Meta | 119 | 7% | 559 (31%) | 2% |
| Companies | 4 | NVIDIA | 94 | 5% | 651 (37%) | 2% |
| Companies | 5 | OpenAI | 79 | 4% | 714 (40%) | 2% |
| Companies | 6 | Amazon | 57 | 3% | 770 (43%) | 1% |
| Companies | 7 | Anthropic | 50 | 3% | 813 (46%) | 1% |
| Companies | 8 | DeepSeek | 44 | 2% | 855 (48%) | 1% |
| Companies | 9 | Alibaba | 35 | 2% | 889 (50%) | 1% |
| Companies | 10 | Apple | 33 | 2% | 911 (51%) | 1% |
| Institutes and labs | 1 | Allen Institute for AI (Ai2) | 82 | 6% | 82 (6%) | 2% |
| Institutes and labs | 2 | CNRS | 78 | 6% | 160 (12%) | 1% |
| Institutes and labs | 3 | Chinese Academy of Sciences | 68 | 5% | 227 (18%) | 1% |
| Institutes and labs | 4 | Max Planck Society | 52 | 4% | 279 (22%) | 1% |
| Institutes and labs | 5 | Vector Institute | 48 | 4% | 327 (25%) | 1% |
| Institutes and labs | 6 | Inria | 46 | 4% | 351 (27%) | 1% |
| Institutes and labs | 7 | RIKEN | 46 | 4% | 396 (31%) | 1% |
| Institutes and labs | 8 | Helmholtz Association of German Research Centres | 25 | 2% | 419 (32%) | 0% |
| Institutes and labs | 9 | Mila | 25 | 2% | 444 (34%) | 0% |
| Institutes and labs | 10 | DFKI | 19 | 1% | 463 (36%) | 0% |
| Institutes and labs | 11 | Munich Center for Machine Learning | 19 | 1% | 477 (37%) | 0% |
| Universities | 1 | Massachusetts Institute of Technology | 120 | 3% | 120 (3%) | 2% |
| Universities | 2 | Carnegie Mellon University | 116 | 3% | 232 (6%) | 2% |
| Universities | 3 | Stanford University | 105 | 3% | 331 (9%) | 2% |
| Universities | 4 | University of California, Berkeley | 77 | 2% | 402 (11%) | 1% |
| Universities | 5 | University of Washington | 76 | 2% | 468 (13%) | 1% |
| Universities | 6 | University of Oxford | 71 | 2% | 538 (15%) | 1% |
| Universities | 7 | University of Toronto | 63 | 2% | 598 (16%) | 1% |
| Universities | 8 | Harvard University | 61 | 2% | 652 (18%) | 1% |
| Universities | 9 | Cornell University | 56 | 2% | 708 (19%) | 1% |
| Universities | 10 | Georgia Institute of Technology | 50 | 1% | 754 (21%) | 1% |
Between 1979 and 2025, the combined share of company researchers at the top three companies (Google, Microsoft and Meta) moved by 14 percentage points, from 45% (41 of 91) to 31% (559 of 1,782). Over the same 46 years, companies ranked 4th through 10th moved from 22% (20 people) to 22% (388 people), with Google, Microsoft, Meta, NVIDIA, OpenAI, Amazon, Anthropic, DeepSeek, Alibaba and Apple (321 people, 127 people, 119 people, 94 people, 79 people, 57 people, 50 people, 44 people, 35 people and 33 people) entering the corporate top 10. Across the full corporate sector, the number of companies employing at least 10 tracked people was 2 in 1979 and 19 in 2025, and the share of company researchers working outside the top 10 moved from 36% (33 people across 34 companies) to 49% (871 people across 713 companies).
In universities and in research institutes and labs, the top 10 hold 21% and 36% of their sectors' people in 2025. Across 1,130 active universities in 2025, the top 10 (Massachusetts Institute of Technology at 120, Carnegie Mellon University at 116, Stanford University at 105, and University of California, Berkeley, University of Washington, University of Oxford, University of Toronto, Harvard University, Cornell University and Georgia Institute of Technology) employ 754 of 3,635 academic researchers (21%), and 103 universities employ at least 10 people. Across 594 active institutes, nonprofits and government labs in 2025, the top 10 employ 463 of 1,295 people (36%) (or 477 people, 37%, across 11 organizations when including the tie for 10th place between DFKI and Munich Center for Machine Learning at 19 people each), led by Allen Institute for AI (Ai2) (82), CNRS (78) and Chinese Academy of Sciences (68).
Career data comes from the biographies on this site and from OpenAlex author records, as of 22 October 2026. People whose OpenAlex record may be incomplete or mixed are left out of citation figures.