Bet on Sam Altman

Essay: Anthropic is having a killer 2026, but secondary markets aren't reflecting OpenAI's reality.

I need to vent for a second.

So, yes, Anthropic is winning.

Secondary retail markets are bidding up the company’s valuation to $1.2 trillion, 73% of new enterprises choose Claude Code, they’re adding a million new users per day. No, genuinely it’s crazy — Anthropic is fielding bids up to $800B from institutional investors (more than double its February raise) while hedge funds are trying to dump OpenAI shares and can’t find buyers.

Don’t get me wrong, I’m a huge fan of Claude & Co., but the market is underestimating Sam Altman and mispricing OpenAI as a result.

The last six months have hosted OpenAI’s worst moment since the Altman board drama — users unsubscribing, Sora’s death, multiple exec exits, a coding model that got leapfrogged…

But I believe something has changed more recently — OpenAI has refocused, is shipping fast — even faster than Anthropic (?) — and is now AI’s most mispriced opportunity.

Looking at the Wrong Charts

Anthropic’s last reported valuation was its $380B round it announced back in February. Business Insider reported last week that institutional investors are now pitching the company on deals pricing it at up to $800B (110% markup in 8 weeks lol).

OpenAI’s last round was $852B in March. The secondary market is pricing it below that. In fact, shares are so illiquid that Bloomberg reported institutional investors are struggling to unload hundreds of millions of dollars in shares, and interested parties are coming in at 10% below its current stated valuation.

When the secondary market diverges this hard from primary pricing, one of two things is true: the primary is a bubble or the secondary is a stampede…

But given that OpenAI just posted $24B ARR with 900M+ weekly ChatGPT users and a monopoly on consumer AI, I’m going to take the stampede bet here.

Sam Cut the Fat

Remember when OpenAI was going to do everything?

Videos. Agents. Browsers. Social Networks. Adult-mode. The full consumer stack. It was a company that had decided frontier research wasn’t enough, it needed to also be Apple, Google, and Meta at the same time.

That OpenAI is dead — in the last 60 days, Sam has shut down Sora, fired three of his top operating execs plus his CMO, and refocused the entire company onto three things: building the best general model, the best coding model, and the best image model.

The word inside the company is “consolidation.” Every team that isn’t directly feeding those three things has been consolidated or cut.

Compare that to Anthropic (which just leaked its entire Claude Code source code), which is getting openly roasted on social media for rate-limit restrictions and is reportedly struggling to scale up compute to even serve the models to users.

Anthropic, branded as the “mature” lab is having the toughest security and user-trust period of any AI lab at the moment. I am a huge fan of Anthropic, but we have to look at both companies and realize OpenAI has closed the gap.

The Gap is 2 Weeks, Not 2 Years

Six months ago OpenAI was 2 quarters behind Anthropic on coding. Today, that gap is measured in weeks (if that).

• Codex CLI now beats Claude Code on SWE-bench Pro (57.7% vs 46%) and Terminal-Bench 2.0 (75.1% vs 69.4%).

• While Claude still wins SWE-bench Verified (87.6% vs ~80%) the feedback from real engineers using Codex every day is clear: it’s better.

Then there’s shipping cadence:

• ChatGPT Image 2 shipped yesterday with frontier image quality. Claude doesn’t even have native image-gen capabilities.

• GPT-5.4 shipped March 5 and has native Computer Use built in

• Codex macOS desktop app shipped February

• Codex CLI now running on non-Nvidia hardware

• Agent Builder shipped today

Does all of this sound familiar? It should.

Every major Anthropic feature is being matched within 2-4 weeks. OpenAI is obviously using AI to build features and it’s clear they have the superior compute to keep winning. Sam has publicly said OpenAI is using its own models to write 100% of its code.

Yes, There Are Risks

I’m not going to pretend this take is bulletproof…

Three things could invalidate it:

1. Anthropic’s enterprise moat is real. It has 30.6% enterprise market share, and climbing. If Fortune 500s keep doubling down on Claude, then Codex is clearly in trouble.

2. ChatGPT’s cancellations continue. OpenAI’s ARR growth curve is visibly flattening ($20B -> $24B over 6 months) while Anthropic went from $9B to $30B in the same window (although their accounting seems a bit more questionable).

3. Consumer AI shifts. Apple is about to turn on personalized intelligence for swaths of devices. Google is burning $160B on Gemini. OpenAI’s consumer lead could die in a blink.

These are all real risks but I just don’t think they outweigh the refocus + shipping velocity + secondary-market mispricing setup.

In my humble opinion, people are sleeping on Sam.

---

Thanks for joining us for our latest issue. Now go listen to our podcast :)