scripod.com

Neil Movva - Making AI 10x Cheaper - [Invest Like the Best, EP.488]

Shownote

My guest today is Neil Movva, founder of Sail. Sail is building what Neil calls a token factory, an inference company designed for a specific kind of future, one where AI agents run in the background for hours or days at a time rather than answering a huma...

Highlights

This episode examines how AI infrastructure may evolve as systems shift from instant responses to sustained, autonomous work. Neil Movva explains the technical and economic challenges involved in making that future practical.
00:00
AI inference can become dramatically cheaper
02:22
Cheaper inference could unlock longer-running AI agents
03:22
Sail aims to make AI inference dramatically cheaper
07:44
Background workloads could dominate token consumption
11:23
Cheap inference makes long-horizon AI practical
17:30
Interactive AI demands speed, not just throughput
22:28
Interconnects can outweigh raw compute efficiency
29:49
Long-context AI turns KV cache into a memory bottleneck.
35:58
Overfitting reveals patterns that enable generalization
39:12
Software efficiency may not remain a durable moat
44:06
No chip is bad at the right price
50:31
Smaller sites make large-scale AI more practical.
53:40
AI agents can trade uptime for lower costs
59:07
AI inference still leaves enormous efficiency gains untapped
1:06:03
Technological leads last only while they remain difficult to copy
1:09:22
AI must become millions of times cheaper to scale
1:10:37
Model architecture will shape AI infrastructure
1:12:50
NVIDIA’s biggest vulnerability may be limited HBM supply
1:14:39
Understanding the full stack creates lasting advantage

Chapters

Welcome to Invest Like The Best
00:00
Neil Movva
02:20
Building a Token Factory
03:22
The Rise of Long-Running Agents
05:32
Deep Research and Cybersecurity
08:47
The Full Stack of Intelligence
15:12
Throughput Versus Latency
20:03
The Future of AI Chips
24:58
Why Transformers Work
33:19
The Future of Data
36:43
The Market for AI Chips
44:05
Is the AI Boom Different?
47:56
Reinventing the Data Center
51:08
Scavenging Power
56:43
Where Compute Is Most Inefficient
1:01:04
Open Versus Closed Models
1:07:02
A Trillion Tokens a Day
1:10:37
The Contrarian Case on NVIDIA
1:12:42
Advice for AI Hardware Founders
1:14:38

Transcript

Speaker 4: Ramp is the only platform built to make your finance team leaner, faster, and better, saving businesses 5% annually on average. So you can stay focused on growth. Ramp customers grow revenue 3.2 times faster than the average American business. V...