Data is Back: MongoDB, Databricks, Snowflake
Sourcery
1 DAYS AGO
Data is Back: MongoDB, Databricks, Snowflake
Data is Back: MongoDB, Databricks, Snowflake
Sourcery
1 DAYS AGO
MongoDB CEO CJ Desai discusses the evolving AI landscape, highlighting how data has become the critical foundation for AI applications. He explains that hyperscaler capacity shortages are driving a surprising return to on-premise infrastructure, and outlines the distinct needs of frontier labs, AI-native companies, and enterprises building agentic workloads.
Desai identifies three customer classes: frontier labs, AI-native startups like ElevenLabs (running 50M+ agents on MongoDB), and Global 2000 enterprises. He notes that hyperscalers are running out of capacity, forcing even top-50 customers to keep workloads on-prem for data sovereignty and AI workloads. MongoDB differentiates from Snowflake and Databricks as a real-time operational database (OLTP) versus analytical (OLAP). Frontier labs use MongoDB as a memory layer for AI inference. Desai emphasizes that enterprise AI lacks standardization, with customers using a mix of open and closed-source models. Agentic architectures are far more complex than prototypes, requiring deterministic outcomes, auditability, and human-in-the-loop for compliance. He sees data as the unsung hero of the AI cycle, with MongoDB positioned as the foundational data layer for agentic applications.
00:00
00:00
Hyperscalers are running out of capacity.
01:12
01:12
Data layer is crucial for agentic applications
03:24
03:24
AI and agents generate massive data, positioning MongoDB as a downstream winner.
07:25
07:25
Enterprises choose MongoDB for multi-cloud resilience
09:12
09:12
Hyperscalers running out of capacity.
11:19
11:19
MongoDB is a real-time operational database for OLTP.
13:18
13:18
ElevenLabs runs 50M+ agents on MongoDB.
15:23
15:23
Data labeling companies are complementary rather than competitive.
16:25
16:25
Customer focus and team building are key
19:32
19:32
Unprecedented innovation pace in AI.
27:04
27:04
AI companies cannot afford expensive DBAs.
28:31
28:31
Agentic architecture requires deterministic outcomes and human-in-the-loop
31:22
31:22
Customers use both open and closed source models
34:39
34:39
Non-zero chance of success if resources are focused
36:55
36:55
Mixed pride and concern about competitors hiring MongoDB people.
40:32
40:32
MongoDB is the foundational data layer for AI and agentic applications.
41:38
41:38
Data is the unsung hero in AI
