Anthropic and OpenAI Were Racing Each Other. And Maybe Now They're Not

EDITOR’S NOTE

Dear Nanobits readers,

I have been sitting with this one for a few weeks now, honestly.

Every time I open LinkedIn or scroll through my feeds, someone is either declaring that Anthropic has "won" the AI race, or that OpenAI is in chaos, or that the whole thing is a distraction from whatever the next headline is. The discourse is loud and moves fast, and most of it, I would argue, is built on a false premise that these two companies are in a head-to-head race for the same finish line.

I don't think that's true anymore. And I think understanding why matters not just as a spectator, but because it shapes which tools you build with, which platforms you trust your workflows to, and where the actual opportunities are quietly sitting while everyone argues about benchmarks.

This week, I want to break down what's actually happening between Anthropic and OpenAI. Not the version the breathless tech headlines give you ("OpenAI is CRASHING OUT!" / "Anthropic is taking OVER!") but the more honest, more interesting version.

Let's get into it.

THE SPLIT THAT NOBODY CALLED

Cast your mind back to 2022. OpenAI launched ChatGPT, broke every adoption record in history, and created the entire modern AI category essentially overnight. Anthropic, founded just a year earlier by ex-OpenAI researchers, was the safety-focused upstart that most people had barely heard of.

Fast forward to today, and the gap is still enormous in some ways. OpenAI's ChatGPT has over 900 million weekly active users. Claude's web traffic is about 70x smaller. By consumer numbers, it's not even a competition.

But here's the thing: revenue is not the same as users. And enterprise is not the same as consumer.

By 2026, the difference between the two has become this: OpenAI is a consumer company trying to make enterprise products. Anthropic is an enterprise company that also has a consumer product. That's not spin. That's what the data is showing.

Claude has an estimated 18.9 million monthly active users, roughly 5% of ChatGPT's user base. Yet this smaller, more focused audience generates 40% of OpenAI's revenue relative to scale. Anthropic has a fraction of OpenAI's user base but generates revenue at a dramatically higher rate per user because nearly 80% of its revenue comes from enterprise contracts, not consumer subscriptions.

That is not a story about who's "winning." That's a story about two completely different games being played on the same court.

ANTHROPIC'S REMARKABLE RUN

Anthropic just had one of the most remarkable quarters any AI company has ever had. OpenAI just shut down its most hyped product. These two things are connected

Anthropic was at roughly $1 billion in annual revenue in December 2024. It hit $4 billion by mid-2025. It crossed $9 billion by end of 2025. And then $14 billion in February 2026: that's $1B → $14B in about 14 months.

As Meritech's Alex Clayton noted: "We've looked at the IPOs of over 200 public software companies, and this growth rate has never happened."

The engine behind it? Claude Code. Claude Code's run-rate revenue grew to over $2.5 billion, more than doubling since the beginning of 2026. The number of weekly active Claude Code users also doubled since January 1. A recent analysis estimated that 4% of all GitHub public commits worldwide are being authored by Claude Code.

And in late March 2026, Anthropic shipped computer use: the ability for Claude to actually control your desktop, navigate apps, fill forms, move files, run terminal commands, all through natural language. Launched on March 23, 2026, the feature works across both Cowork (Anthropic's tool for non-developers) and Claude Code, letting you assign tasks from your phone via Dispatch and come back to finished work on your computer. Think of it as giving Claude hands, not just a voice.

A December 2025 report from Menlo Ventures found that Anthropic captures 40% of enterprise LLM spend, up from 24%, while OpenAI's share fell to 27%, down from 50%.

The bet Anthropic made, coding as the gateway into enterprise, safety as a differentiator, reliability over spectacle, is paying off in a way that would have seemed implausible two years ago.

WHAT HAPPENED TO OPENAI?

This is the part of the story that requires some nuance, because the "OpenAI crashing out" narrative is both partially true and quite misleading.

Let's start with the honest version.

Sora, OpenAI's AI video generation tool, was burning through roughly $1 million every day in compute costs not because people loved it but because video generation is extraordinarily expensive to run. While a whole team inside OpenAI was focused on making Sora work, Anthropic was quietly winning over the software engineers and enterprises that actually drive revenue. Sora's estimated $15 million daily inference costs at peak dwarfed its $2.1 million total lifetime revenue. Downloads had dropped 66% from their November 2025 peak.

Disney, which had committed a billion dollars and a three-year character licensing deal featuring 200+ Marvel, Pixar, and Star Wars characters, reportedly found out about the shutdown less than an hour before the public announcement. That's not a graceful product sunset. That's an emergency stop.

At an all-hands meeting on March 16, OpenAI's Applications CEO, Fidji Simo, bluntly stated that Anthropic was a "wake-up call" and that the company was "spreading its energy across too many applications and technology stacks."

But here's where I want to push back on the "crashing out" framing: OpenAI is generating $2 billion in revenue per month and made $13.1 billion in revenue last year. That is not a failing company. That is a company that got distracted and is now course-correcting.

And on March 31, 2026, just days ago, OpenAI officially closed the largest private funding round in Silicon Valley's history: $122 billion in committed capital at a post-money valuation of $852 billion. The round was led by SoftBank, a16z, D.E. Shaw Ventures, MGX, and TPG, with participation from Amazon ($50B), Nvidia ($30B), and Microsoft. Enterprise customers now make up over 40% of revenue, and OpenAI says they're on track to reach parity with consumer by end of 2026.

This is not a company in panic mode. This is a company that was spread too thin, made a very public correction, and now has an eight-hundred-billion-dollar war chest to execute on its refocused strategy. The Sora shutdown wasn't a white flag. It was an IPO-readiness move.

THE REAL BATTLEGROUND: NOT BENCHMARKS, BUT MOATS

Here's what I think the AI discourse gets wrong constantly: it treats this like a benchmark race. Who scored higher on MMLU? Who's better at math? Who has the longer context window?

That's not the competition anymore.

Choosing Claude has become a signal of professional identity for engineers and researchers who value precision and transparency over sheer scale. Data from a16z's Top 100 Gen AI Consumer Apps report confirms that these two platforms are no longer trying to occupy the same market niche.

Anthropic has a cultural moat. Their Constitutional AI approach, their safety-first positioning, their refusal to let certain government uses go forward, regardless of what you think about the politics of it, has made them the preferred choice for a specific kind of enterprise buyer. The kind that cares about auditability and not being in the headlines for the wrong reasons. In early 2026, Ramp's spending data showed Anthropic capturing over 73% of all spending among companies buying AI tools for the first time.

OpenAI has a distribution moat. A billion users is not a small thing. The network effects of ChatGPT being the verb for AI, the way Google became the verb for search, is genuinely defensible. Kids start there. Companies start there. Enterprise customers now accounting for 40% of OpenAI's $25B ARR means the enterprise play isn't dead, it's just getting serious.

The question for the next 24 months isn't "who wins." It's whether these moats hold as Google (which already has your Calendar, your Gmail, your Docs) figures out how to deploy its agent features at scale.

WHY THIS MOMENT MATTERS FOR BUILDERS

Why I am covering this particular story today: because I have been building with these tools, and the gap between watching this happen and making things happen is smaller than it's ever been.

In the past 3 weekends, I built using Claude Code 1/ a real-time flight detection system, the Flight Whisperer, that sends WhatsApp notifications using Twilio and the OpenSky API; 2/ a custom Alexa skill called Calendar Buddy that pulls from Google Calendar and announces as well as books your calendar; and 3/ a system, The Message you Leave Behind, using Google Apps, Gmail and Google Drive to send letters to your loved ones even if you are not there. All of it in a fraction of the time it would have taken even 18 months ago.

These weren't prototype demos. They were working tools I use. And the pattern I keep seeing is that Claude Code has fundamentally changed what "building something" means, you're no longer fighting the environment to get to the idea.

Four engineers built Cowork in 10 days, with most of the code written by Claude Code itself. That's the company's own internal product, built by the tool it's selling. That tells you something about where we actually are.

The people winning right now aren't necessarily the ones with the biggest teams or the most funding. They are the ones who picked a problem worth solving and got moving.

WHERE DOES THIS GO?

A few things I am actively watching:

The consumer monetization question is genuinely unresolved. 79% of companies paying for Anthropic are already paying for OpenAI too. Enterprises aren't picking sides, they are running both in parallel. Which means the real question is which model gets chosen for which workflow. That's going to be won by UX and reliability, not press cycles.

Google is the sleeping giant here. They already have access to your email, your calendar, your documents. CIOs project that by 2026, OpenAI will hold 53% market share, while Anthropic and Google are each expected to account for 18%. The Google agent, when it properly ships, inherits trust that no new entrant can buy.

The SaaS disruption is real, not hype. The Cowork launch triggered a massive selloff in global SaaS stocks. The software sector lost roughly $2 trillion in market cap as investors recognized that agentic AI tools could disrupt traditional enterprise software models. Worth watching closely.

END NOTE

The honest take: Anthropic and OpenAI are not competing for the same thing anymore, and trying to crown a winner misses the more interesting question: which is what happens when the infrastructure layer of AI stabilizes and the application layer explodes.

That's where the next set of opportunities live. Not in the model race, but in the problem layer on top of it.

If you are building something, anything, this is probably the best moment in history to do it. The tools have never been this good, this accessible, or this willing to do the boring parts for you.

See you next week!!