AI Summary
5 min readAnjney Midha's Plan to Radically Lower the Price of Compute
When Anjney Midha wrote the first check for Anthropic in early 2021, he and the founders tried to raise $500 million. They couldn't. Instead they scraped together about $100 million from "a bunch of cats and dogs" who believed in the mission. "It was a bit of a wake-up call," Midha recalls. That experience—watching a world-class team struggle to get the compute they needed—eventually led him to start AMP PBC, a public benefit corporation building what he calls "a grid for compute."
The Jagged Frontier
Midha pushes back hard on the popular narrative that only three labs—OpenAI, Anthropic, and Google DeepMind—are competing at the frontier. "Which frontier and which three models?" he asks. "From where I'm sitting, there's like seventeen different frontiers right now, and there's four different players in each one."
The frontier is not a single destination. Midha draws an analogy to the American West: "There was a frontier of gold and there was a frontier of jeans. Levi's turned out to be a new modern behemoth." In AI, Anthropic leads on software engineering, OpenAI on consumer chat, and Black Forest Labs on video generation. The models are not at parity—they are remarkably different in meaningful ways, reflecting what each team focuses on day after day.
Continue reading the full summary in the app — free to try.
Read Full Summary →Free • No credit card required
Never miss an episode of Odd Lots
Get every new episode summarized in your inbox — free, ~5 minutes to read.
No spam. Unsubscribe anytime.
What you'll learn
- 1 (00:00) **Advertisements and Sponsor Messages** - Multiple ad reads for VanEck, IBM, Audi, Public, Chase, and Odoo.
- 2 (01:39) **Episode Introduction: Physical Constraints on AI** - Tracy and Joe frame the discussion around the physical resources (power, GPUs, real estate) required for AI, comparing it to the "paperclip maximizer" thought experiment where all resources are consumed for a single goal.
- 3 (05:08) **Guest Introduction: Anjney Midha & AMP PBC** - Midha is introduced as a former a16z GP, Stanford lecturer, early Anthropic investor, and founder of AMP PBC (a public benefit corporation aiming to lower compute costs).
- 4 (06:12) **The Anthropic Origin Story: Early Compute Struggles** - Midha recounts writing the first check for Anthropic in late 2020, the difficulty of raising capital, and how the experience taught him that compute infrastructure is the key bottleneck for AI labs.
- 5 (11:02) **Definition: The "Jagged Frontier" of AI Models** - Midha argues there is not just one frontier (AGI), but many (software engineering, chat, video). He disagrees with the idea that all frontier models are at parity, noting they are "remarkably different" based on their focus.
- 6 (14:42) **Advertisement Break** - Ad reads for VanEck, IBM, and Public.
- 7 (16:55) **Mechanism: Verifiable Feedback vs. Subjective Feedback** - Midha explains the critical distinction between subjective feedback (user opinions) and verifiable feedback (factual, testable results).
+ Full timestamped outline available in the app
Guests on this episode
Show Notes
Anjney Midha wrote the first check to Anthropic. He teaches a viral course at Stanford on how AI works. And he was, until recently, a partner at a16z. In other words, he is AI-industry royalty. Midha's new project is AMP PBC, a company that believes it can radically lower the price of compute. To accomplish that, he is working on building a compute grid that turns GPUs into a standardized utility. But right now, compute is too fragmented. It's too heterogeneous. And given the way contracts are structured, he says that labs are being forced to spend money on capacity that often goes unused. In other words, small labs are forced to pay up for big, long-term contracts, even though their own demand (particularly during model training) may be very spiky. On this episode, Midha explains how the market for compute currently works and why he believes there's a software solution that could significantly improve compute utilization. He also tells us why he does not anticipate one company will emerge as the dominate player and that instead we'll have a wide range of models, each optimally used in specific applications.
Read more:
Amazon Says Its Data Centers Use 2.5 Billion Gallons of Water
Oracle Falls Most in Six Months on Mounting Data Center Costs
Only http://Bloomberg.com subscribers can get the Odd Lots newsletter in their inbox each week, plus unlimited access to the site and app. Subscribe at bloomberg.com/subscriptions/oddlots