The Analytics Engineering Podcast The data jobs to be done (w/ Erik Bernhardsson)
44 snips
Nov 3, 2024 
In this discussion, Erik Bernhardsson, CEO and co-founder of Modal Labs, dives into the intricacies of cloud computing, emphasizing the competitive landscape with giants like Snowflake and AWS. He shares insights on the GPU shortage and its implications for AI startups. Erik boldly critiques the future role of data engineers, questioning its necessity while spotlighting the rise of analytics engineers. The conversation also touches on the impact of generative AI and the democratization of software development, showcasing a landscape ripe with innovation.
AI Snips
Chapters
Transcript
Episode notes
GPU Economics and Availability
- The high cost of GPUs necessitates high utilization, impacting availability in single zones.
- Multi-AZ and multi-cloud architectures help overcome poor availability but need resource pooling.
GPU vs. CPU Economics
- GPU economics differ from CPUs; GPU costs mainly go to NVIDIA, requiring high utilization.
- CPU costs are tied to power consumption, making utilization less critical for providers like AWS.
GenAI and Egress Fees
- Generative AI workloads are less sensitive to egress fees due to their compute-intensive nature.
- They send small amounts of data relative to computation, making multi-region architectures feasible.