AI Fact or Fiction: The Fable Ban, Tokenmaxxing, Saaspocolypse — With Ara Kharazian
TL;DR
Anthropic is now ahead of OpenAI in Ramp's business data: Kharazian says 41% of US firms on Ramp use Anthropic versus 39.5% using OpenAI, with Anthropic still growing while OpenAI is relatively flat.
Government pressure may have boosted Anthropic's brand: After the DoD labeled Anthropic a supply chain risk earlier in the spring, adoption still accelerated, partly because businesses did not take the warning seriously and partly because exceptions undercut the government's credibility.
AI spend is exploding, but most companies are still spending very little: Token spend per business is up 15x since January 2025, yet the median firm spends only $11 per employee per month on AI, while the top 10% spend $611 and the top 1% spend $7,449.
'Tokenmaxxing' is overstated because firms want control, not just lower prices: Businesses are still increasing AI spend month over month, up 14% last month, but the bigger complaint is that OpenAI and Anthropic give admins too little visibility and control over runaway token usage.
DeepSeek and open source are growing, but still from a tiny base: Only about 5% of firms on Ramp use open-source platforms, up from 1% last year, and DeepSeek's recent spike is notable but likely fragile if OpenAI and Anthropic respond with more price cuts.
The SaaS apocalypse is not in the data yet: Traditional software is still mostly bought on seat-based contracts, metered usage is under 5%, and even AI-native threats are more likely to come from new vertical products like Attio than from companies simply replacing software with OpenAI or Anthropic.
The Breakdown
Anthropic just passed OpenAI in business adoption, with 41% of US firms on Ramp using Anthropic versus 39.5% using OpenAI, and Eric Kharazian argues the White House and DoD clashes may have strengthened Anthropic more than hurt it. He also throws cold water on two loud AI narratives: companies are not wildly overspending on tokens yet, and the "SaaS apocalypse" still is not showing up in actual purchasing data.
Was This Useful?
Share
Keep Reading
Make Alcreon Yours
Tune your feedFive quick questions, and the feed ranks what matters to you first.Or just get notified
The weekly Echo. Signal worth keeping in your inbox.
Every new piece, announced on X.
Read Next
See all
Playbook
Cheap Models, Hard Tasks
Most agent workflows route every step to the frontier model by default. The bill scales with how chatty the agent gets, even when most steps don't need that brain.

Playbook
Tasteful Skills
“Tasteful Skills” argues that the best agent skills are not documentation or best-practice lists.

Playbook
The Art of Tasteful Prompting
Learn how tasteful prompting helps you move beyond generic AI output by shaping context, style, and judgment from the start.