AI Safety

Anthropic Says Its AI Now Trains 30,000x More Efficiently

Anthropic Says Its AI Now Trains 30,000x More Efficiently

⚡This could mean cheaper AI for you and huge profits for Anthropic.

Deep Dive

Anthropic, the company behind the Claude AI chatbot, is preparing to go public. In its investor roadshow video, it showed a graph that seems to claim its AI models have become dramatically more efficient to train. The graph has no numbers on the vertical axis, but one analyst estimates each line represents a 10x jump, suggesting the latest model needs only 1/30,000th of the computing power that a 2024 model needed for the same quality. That's like going from a supercomputer to a laptop for the same job.

Why does this matter? If true, it means Anthropic can train advanced AI models for a fraction of the cost, which could lead to cheaper AI services for everyone. The company's CFO said that over the last six quarters, revenue grew more than 45 times while computing costs grew only 10 times. That's a 4x improvement in revenue per dollar of compute. For you, that could mean more powerful AI tools at lower prices, or new features that were previously too expensive to offer.

But there's a catch. The graph is vague and the 30,000x figure is highly uncertain. The analyst who wrote the post says they are 'pretty unsure' and the real number could be much higher or lower. Also, the efficiency gain applies only to pretraining—the initial phase of teaching an AI—not to the entire process of making it useful. And the graph includes an arrowhead for an unreleased model, which might be a future projection rather than a proven result.

Finally, Anthropic claims these efficiency gains come from 'compute multipliers'—discoveries that make training more efficient. They say these multipliers stack and could accelerate as Claude itself helps find new ones, potentially leading to recursive self-improvement. That's a bold claim, but if it pans out, it could reshape the economics of AI and put pressure on competitors to keep up.

Key Points
  • Anthropic's graph suggests its AI models can now be trained with 30,000 times less computing power than in 2024, but the exact number is uncertain.
  • The company's revenue grew 45x while compute costs grew only 10x, showing real-world efficiency gains.
  • These savings apply only to pretraining, not the whole AI development process, and may not translate directly to user prices.

Why It Matters

Cheaper AI training could lead to lower prices and more powerful AI tools for everyday users.

🔒 FREE VIP ACCESS

Read the full breakdown & get daily AI executive intelligence

Join like-minded builders, researchers, and AI professionals who read The Neural Feed daily. 100% free. No spam.

🔔 BREAKING RELEASE RADAR INSTANT ALERTS • 1-CLICK PAUSE

Never miss the next Anthropic & Claude drop

We matched Anthropic & Claude from this briefing. Pick your AI radar tracks below for instant breaking release alerts + our free daily morning briefing.

Includes full article unlock • Only major releases • Unsubscribe anytime
🎁 FREE EXECUTIVE DOWNLOAD ISSUE #21

Download This Week's 8-Page Executive AI Briefing (PDF)

Our curated executive intelligence packet: in-depth breakdowns of frontier models, market movements, plus full-color AI comics.

  • ✓ 8-page formatted executive PDF briefing
  • ✓ Frontier AI models & technical breakthroughs
  • ✓ Full-color satirical AI comic strips
Instant 1-click download • 100% free • Unsubscribe anytime
📬 Get the top 10 AI stories daily