ChatGPT hit 1 million users in 5 days. Here's the growth data in context.

One million users in five days.

That number has been bouncing around Twitter since OpenAI confirmed it, and I've seen a lot of people say "fastest ever!" without showing the comparison data. So let me show the comparison data.

The leaderboard

Time to reach 1 million users for major consumer technology products, compiled from Statista, SimilarWeb, data.ai, company announcements, and press reporting:

| Product | Year launched | Time to 1M users | |---------|-------------|-------------------| | ChatGPT | 2022 | 5 days | | Instagram | 2010 | 2.5 months | | Spotify | 2008 | 5 months | | Dropbox | 2008 | 7 months | | Facebook | 2004 | 10 months | | Foursquare | 2009 | 13 months | | Twitter | 2006 | 24 months | | Airbnb | 2008 | 2.5 years | | Netflix (streaming) | 2007 | 3.5 years | | Kickstarter | 2009 | 2.5 years |

Five days vs. 2.5 months for Instagram (the previous speed record among major consumer products). ChatGPT is roughly 15x faster than Instagram's trajectory.

And this isn't even a fair comparison. Instagram was mobile-first, launched into the App Store where discovery is built in. ChatGPT is a website you have to navigate to, with no app store listing, minimal marketing, and a product that requires you to understand what a "language model" is.

But wait. TikTok.

I should address the TikTok comparison because people keep bringing it up. TikTok's growth story is complicated. Depending on whether you count Musical.ly's existing user base (TikTok absorbed it), the time to 1M varies:

| Version | Time to 1M | |---------|-----------| | Musical.ly (pre-merger) | ~4 months (2014) | | TikTok China (Douyin) | ~1 month (2016) | | TikTok International | ~2 months (2018) | | TikTok post-Musical.ly merger | Instant (inherited users) |

Douyin in China hit 1M in about a month, which is faster than Instagram. But even Douyin took 6x longer than ChatGPT.

The growth rate in context

Let me put ChatGPT's daily growth rate into perspective:

| Product | Day 1 est. users | Day 5 est. users | Implied daily growth rate | |---------|-----------------|-----------------|--------------------------| | ChatGPT | ~100K | ~1,000K | ~58% daily | | Instagram (first week) | ~25K | ~40K | ~10% daily | | Threads (Meta, 2023)* | N/A | N/A | N/A | | Spotify (first week) | ~5K | ~8K | ~10% daily |

*Threads hasn't launched yet, but I'm leaving a row for it because I suspect it'll be the next contender for this list.

A 58% daily growth rate is absurd. At that rate (which obviously won't sustain), you'd hit 100 million in 17 days. ChatGPT won't maintain this pace because the pool of people who are both tech-curious and aware of ChatGPT's existence is finite. But even a sharp deceleration still implies tens of millions of users within months.

What's driving it

I've been reading every mention of ChatGPT I can find on Twitter, Reddit, Hacker News, and news sites. The viral loop is different from typical consumer apps.

Most viral products spread through social sharing: "Look at this photo I took" (Instagram) or "Watch this video" (TikTok). ChatGPT is spreading through output sharing. People screenshot ChatGPT conversations, post them on social media, and others go try it themselves.

The pattern is:

1. Someone asks ChatGPT to write a poem/essay/code/joke
2. They screenshot the result
3. They post it on Twitter/Reddit with "I asked ChatGPT to..."
4. Viewers think "I want to try that" and go to chat.openai.com
5. Repeat

It's product-as-content. Every ChatGPT output is a potential ad for ChatGPT.

The free tier strategy

One thing that's easy to overlook: ChatGPT is free. No credit card. No waitlist (at least not initially). No freemium upsell on the landing page. Just a login and you're in.

For comparison:

| Product | Pricing at launch | Friction to start | |---------|------------------|-------------------| | ChatGPT | Free | Email signup | | DALL-E 2 | Free credits, then $15/115 credits | Waitlist (initially) | | GitHub Copilot | $10/month (after free preview) | Payment required | | GPT-3 API | Pay per token | Developer account + payment | | Midjourney | $10-60/month | Discord account |

OpenAI's decision to launch ChatGPT as a free, no-waitlist product removed every friction point. The GPT-3 API had been available for over two years, but it required developer knowledge and a payment method. ChatGPT made the same underlying technology (well, GPT-3.5) accessible to anyone with a browser.

How this compares to GPT-3's launch

The contrast with GPT-3's API launch in June 2020 is striking:

| Metric | GPT-3 API (2020) | ChatGPT (2022) | |--------|------------------|----------------| | Users after 1 week | ~3,000 (waitlisted) | ~1,000,000 | | Access model | Waitlist, developer-only | Open, anyone | | Interface | API (code required) | Chat (browser) | | Cost | Per token | Free | | Public awareness | ML/tech community | Global mainstream |

Same company, same model family, roughly 330x more users in the first week. The product packaging matters more than the technology. GPT-3 was arguably more capable (you had full parameter control), but ChatGPT is usable by non-technical people. That's the difference between thousands and millions.

What happens next

I'm watching two numbers: total registered users and daily active users. Registered users will keep climbing as awareness spreads. DAU is the harder metric because novelty wears off.

My guess, based on how other viral products have behaved after explosive launches: ChatGPT will hit 10 million registered users within a month, then DAU will settle at some fraction (maybe 20-30%) as people who tried it once don't come back daily.

But even 20% of 10 million is 2 million daily active users for a product that launched two weeks ago. That's a real product, not a demo.

The data on this one is being written in real time. I'll update as the numbers come in.

---

If you found this interesting, you might also like:

• 5 charts that explain why GPU prices went insane in 2021
• The training cost curve is doing something weird
• AI research papers published in 2021: a mid-year count
• My 2021 AI data roundup: the 10 numbers that mattered most
• I tracked AI image generation quality over 6 months. The improvement rate is scary.

-- dataku