futuresearch.ai | Report 3Li1

1
Sources of OpenAI's reported $3.4B ARR on June 12 2024

$57 %$

$(50% CI: [54, 60])$

Percent of ChatGPT Plus subscribers from the US at the start of April 2024

Key Parameter	Value	Source
Weight of estimate from mobile subscribers $(W_{1})$	$0.5$	Estimate
Weight of estimate from comparable apps $(W_{2})$	$0.5$	Estimate
Weight of data point from Oct 2023 $(W_{P 1})$	$0.4$	Estimate
Weight of data point from May 2024 $(W_{P 2})$	$0.6$	Estimate
Percent mobile subscribers in US, Oct 2023 $(P_{mobille, Oct 23})$	$63%$	1
Percent mobile subscribers in US, May 2024 $(P_{mobille, May 24})$	$43%$	2
Adjustment for catchup in global users in last data point $(A_{t})$	$0.05 (50% CI: [0.03, 0.065])$	Estimate
Difference in ratio of US-global subscribers on mobile vs. desktop $(A_{M})$	$1 (50% CI: [0.8, 1.2])$	Estimate
Ratio of US-global subscribers of comparable apps $(P_{similar})$	$0.59 (50% CI: [0.49, 0.7])$	Estimate

We have two different ways of estimating the share of US subscribers out of all ChatGPT Plus subscribers:

Drawing from known past datapoints of the ratio of mobile subscribers in the US
Inferring from comparable paid consumer apps for which this ratio is known

For estimate (1), we have two datapoints from October 2023[1] and one from May 2024[2]. The data point from May 2024 is closer to our target date, but it might be more noisy, given that GPT4o was released in May, spurring great attention and an increase in new users. To account for these factors, we give a weight of 0.4 ( $W_{P 1}$ ) to the October data point and a weight of 0.6 ( $W_{P 2}$ ) to the more recent one from May 2024.

We then estimate our two adjustments, for mobile - desktop, and for time since the last datapoint.

We don't really have a reason to believe mobile subscribers to ChatGPT Plus are different from those that use Desktop, so we leave the mobile adjustment at 1 ( $A_{M}$ ).
We should likely see the share of US users to go down over time, given that the US is usually an early adopter. Since the last datapoint we have is after the target date, we'll apply an adjustment of 0.05 ( $A_{T}$ ) – fairly substantial, since our estimate is strongly influenced this last datapoint.

For estimate (2), we compiled this dataset of comparable paid consumer products with known values.

Data for Other Products

Product	% Paying Users from the US	Source
Squarespace Pro	73%	backlinko
DocuSign	73%	docusign
Slack	63%	enterpriseappstoday
Adobe Creative Cloud	43%	enlyft
Codecademy Pro	60%	WSJ
Grammarly	58%	enlyft
Asana	61%	enlyft
Figma	39%	enlyft
Notion	60%	enlyft
Average	59%

\frac{N _{US}}{N _{Global}} Estimate_{mobile} Estimate_{comparable} \frac{N _{US}}{N _{Global}} = W_{mobile} \cdot Estimate_{mobile} + W_{comparable} \cdot Estimate_{comparable} = A_{M} \cdot (W_{P1} \cdot P_{mobile, Oct 23} + W_{P2} \cdot P_{mobile, May 24} + A_{t}) = 1.0 \cdot (0.4 \cdot 0.59 + 0.6 \cdot 0.43 + 0.05) = 0.544 = P_{similar} = 0.59 = 0.5 \cdot 0.544 + 0.5 \cdot 0.59 \approx 57%

Supporting Evidence

[1]
According to Appfigures data, the United States accounts for 59% of ChatGPT's net revenue, followed by Germany, Japan, Great Britain, France, Canada, Brazil, Australia, and India, which account for 2% to 5% of ChatGPT's net revenue.
synthedia.substack.com/p/chatgpt-mobile-app-tops-50m-downloads
[2]
In the week after the launch of OpenAI's latest language model, GPT-4o, the ChatGPT Plus mobile app generated $4.2 million in total revenue. The United States was the top market, accounting for $1.8 million of that revenue, which represents approximately 42.86% of the total.
techcrunch.com/2024/05/20/chatgpts-mobile-app-revenue-saw-biggest-spike-yet-following-gpt-4o-launch/