20VC: Mercor CEO on Why Application Layer Companies Have No Defensibility, The Model is the Product | Token Spend Will Exceed Headcount Spend in 5 Years | The True Cost of Hiring AI Researchers in the Valley Today with Brendan Foody
20VC: Mercor CEO on Why Application Layer Companies Have No Defensibility, The Model is the Product | Token Spend Will Exceed Headcount Spend in 5 Years | The True Cost of Hiring AI Researchers in the Valley Today with Brendan Foody
20VC: Mercor CEO on Why Application Layer Companies Have No Defensibility, The Model is the Product | Token Spend Will Exceed Headcount Spend in 5 Years | The True Cost of Hiring AI Researchers in the Valley Today with Brendan Foody
Brendan Foody, CEO of Mercor, discusses the company's rapid growth to $1.5 billion in ARR and a $10 billion valuation, addressing market rumors and his vision for the future of AI and labor. He refutes claims about losing key customers and clarifies the nature of Mercor's revenue, emphasizing the company's end-to-end service model for AI data tasks.
Foody argues that fears of AI-driven job displacement are historically unfounded, predicting the creation of new job categories like training agents. He believes value will accrue to the infrastructure layer over the application layer due to stronger defensibility from network effects and data advantages. He notes that token spend within organizations is increasing, driven by Jevons paradox, and predicts enterprises will commoditize the model layer through evaluations. Foody also discusses the challenge of competing for AI talent against major labs and confirms plans for an eventual IPO. He proposes eliminating income tax for lower-income Americans, funded by higher capital gains tax, to address inequality.
00:00
00:00
Token spend now exceeds headcount spend.
12:56
12:56
I would not sell the company for $30 billion.
14:31
14:31
250 years of productivity increases created more jobs
16:03
16:03
Economy creating new job categories faster than past revolutions
17:01
17:01
Token spend will triple or quadruple in 12 months
19:53
19:53
Horizontal aggregation and economies of scale are more important
25:33
25:33
High-quality data drives model improvement
27:59
27:59
Covering the full distribution of tasks and contexts in Google Workspace
32:11
32:11
The Series B round at 100x revenue felt most uncomfortable.
33:02
33:02
Infrastructure has moats, applications do not
34:50
34:50
Focus on long-term market dominance, not short-term optimization.
35:52
35:52
The model itself is becoming the product
37:24
37:24
Models will clone Slack end-to-end by 2026
38:47
38:47
AI-enabled services can be a goldmine
42:03
42:03
Spending more on tokens than on employee headcount
46:47
46:47
Enterprise compute spend will exceed headcount spend in 5 years.
59:39
59:39
Europe should accept losing the model race.
1:01:37
1:01:37
Hiring AI researchers is the hardest role.
1:08:51
1:08:51
Foundation model labs will be the most valuable companies.
