Business Pill 44 · Why labs keep building bigger
A scaling law is a pattern: more size, more data and more computing power give predictably better models. It explains the spending. It does not promise where it ends.
A short explainer video, under a minute. The robot and the chart are an illustration, and the video gives no unit for its cost labels.
The Short Answer
Picture a lab that builds a robot, and it is not very smart. So the lab builds one ten times bigger, with ten times more to read. It gets better. They do it again, and it is better again.
When they plot the results, the points fall on a steady line.
What a Scaling Law Is
The steady line lets the lab predict how good the next robot is before building it.
That pattern is called a scaling law: more size, more data and more computing power give predictably better models.
The Bottom of the Chart
Then the video points at the bottom of the chart, titled what each step costs. Every step to the right costs ten times more.
The labels read one million, then ten million, then a hundred million. The video gives no currency or unit for them, so read them as an illustration of the shape. This publication’s own arithmetic: ten million is ten times one million, and a hundred million is a hundred times one million.

Nobody Knows How Long the Line Holds
The line describes the steps already taken. Nobody knows how long it holds. It may bend.
Why It Matters
The video turns this into one question: does our plan depend on the line holding?
Scaling laws explain the spending. They do not promise where it ends.
The Question to Ask
- Does our plan depend on the line holding?
See It in the News
Fractile’s Case That FLOPs Are the Wrong Metric. A news piece on how far compute has scaled against memory bandwidth.
How This Concept Connects
Layer 2 of 8 · Models & Training, on the Business Engineer Concept Map
- GPUs constrains Scaling Laws. Each scaling step needs ten times more computing, and the GPU lesson says those chips are scarce, costly and hungry for power.
- Scaling Laws builds on Transformers and Attention. Reading every word at once let work spread across thousands of chips, which made today’s huge models and their scaling possible.
- Scaling Laws builds on Parameters. Scaling laws tie predictable gains to model size, and size is counted in the parameters the dials lesson describes.
- Scaling Laws drives the cost of Capex and Depreciation. Each scaling step costs ten times more, and the capex lesson shows how such chip spending is spread over assumed lifetimes.
- Scaling Laws contrasts with Distillation. Scaling laws explain why labs build bigger; distillation shows capability getting cheaper as small models learn from large ones.
- Context, Not Capability contrasts with Scaling Laws. Scaling laws explain building bigger models; this lesson says a very capable model still answers poorly without your context.
AI Quality Plateau
Whether scaling gains hold is the quality-plateau question.Business Engineer framework
The AI Capex Map
Each scaling step drives the capex build-out.
Get the full Concept Map as a free poster.
More Business Pills
- The Memory Wall: Why a Faster AI Chip Is Not Faster AI
- Tokens per Watt: What an AI Data Centre Actually Produces
- Why AI Can’t Be Both Instant and Cheap: Latency vs Throughput
- The Model and the Harness: Why Same-Model Products Differ
- Context, Not Capability: Why a Smart AI Model Gives Poor Answers
- Discardable Software: When Code Is Cheap Enough to Throw Away
- Human in the Loop vs Human on the Loop: Supervising AI Agents
- The AI Audit Problem: When Making Work Is Cheaper Than Checking It
- AI Evaluation as Acceptance Test: How to Know It’s Good Enough
- Extensibility Is Control: Who Holds the Power in an AI Product
- RLHF Explained: How an AI Model Learns What People Prefer
- Pretraining Explained: How an AI Model Learns Before Anyone Teaches It
- Fine-Tuning Explained: How to Adapt a General AI Model to One Job
- Tokens Explained: The Unit AI Reads, Writes and Bills In
- Embeddings Explained: How a Machine Compares Meaning
- RAG Explained: How a Model Answers From Your Documents
- AI Hallucination Explained: Why Models State False Things
- Distillation Explained: How a Small Model Learns From a Large One
- Reasoning Models Explained: What Changes When AI Thinks First
- Tool Use Explained: How an AI Model Goes From Text to Action
- Capex and Depreciation Explained: Why Chip Lifetime Drives AI Profits
- Run-Rate Revenue Explained: What an AI Company’s Number Means
- Backlog Explained: Revenue That Is Signed but Not Yet Earned
- Switching Costs Explained: Why It Is Hard to Leave an AI Supplier
- Temperature Explained: The Dial That Sets How Predictable an AI Model Is
- Guardrails Explained: Rules Enforced Around an AI Model, Not Inside It
- Benchmarks Explained: Why a Test Score Is Not Your Own Result
- Agent Memory Explained: How an AI Assistant Remembers You Between Chats
- MCP Explained: The Shared Plug That Connects AI Models to Tools
- Quantization Explained: How a Large AI Model Is Shrunk to Fit
- Parameters Explained: Where an AI Model’s Knowledge Lives
- Synthetic Data Explained: When a Model Writes Its Own Training Examples
- Transformers and Attention Explained: How AI Reads Every Word at Once









