| Both sides previous revision
Previous revision
Next revision
|
Previous revision
|
wiki:ai_timelines:ai_inputs:recent_trends_in_ai_investment [2023/04/03 18:35]
rickkorzekwa |
wiki:ai_timelines:ai_inputs:recent_trends_in_ai_investment [2023/07/13 22:59] (current)
rickkorzekwa Fixed dollar signs |
| ====== Recent trends in AI funding ====== | ====== March 2023 Recent trends in funding for AI companies ====== |
| |
| //Published 29 March, 2023; Last updated 31 March, 2023// | //Published 29 March, 2023; Last major updated 3 April, 2023// |
| |
| Funding for companies using or developing AI averaged \$36B/year from January 2018 through August 2020, rose sharply to average \$87B in 2021, then dropped to average \$37B from July 2022 to March 2023, during which time funding for OpenAI and Anthropic accounted for more than a third of all funding for AI companies. | //13 July, 2023 update: Since the last major update for this page, several AI companies have received large investments, including \$1.3B for Inflection AI, \$500M for SandboxAQ, \$450M for Anthropic, \$350M for Adept AI, \$300M for OpenAI, \$270M for Cohere, and \$150M for character.ai.// |
| | |
| | //For broader, but less up to date analysis, see [[ai_timelines:ai_inputs:funding_of_ai_research|Funding of AI Research]].// |
| | |
| | According to Crunchbase, funding for AI companies averaged \$36B/year from January 2018 through August 2020, rose sharply to average \$87B in 2021, then dropped to average \$37B from July 2022 to March 2023, during which time funding for OpenAI and Anthropic accounted for more than a third of all funding. |
| |
| ===== Details ===== | ===== Details ===== |
| Companies were identified based on their industry categorization on Crunchbase (either "Artificial Intelligence" or "Machine Learning"). We did not make a strong effort to discern how much these companies' efforts go toward AI-related activities, but our impression from examining a small sample is that, while most of the companies included make some claim to using AI, many are using relatively unsophisticated technology or do not use AI as part of their core business model. | Companies were identified based on their industry categorization on Crunchbase (either "Artificial Intelligence" or "Machine Learning"). We did not make a strong effort to discern how much these companies' efforts go toward AI-related activities, but our impression from examining a small sample is that, while most of the companies included make some claim to using AI, many are using relatively unsophisticated technology or do not use AI as part of their core business model. |
| |
| While collecting data, we found several cases of missing data from Crunchbase. This suggests that our dataset may be incomplete, possibly in a misleading way. We have not tried to fill in or correct for missing data or systematically check for bias missing data. | All of the values in this article have been adjusted for inflation, unless otherwise indicated. This does not make a large difference--most of the aggregate statistics are changed by about 10% or less.((For comparison, the standard deviation for total annual funding was 40% of the average, and the 75th percentile funding round raised 8x as much as the 25th percentile round.)) |
| |
| All of the values in this article have been adjusted for inflation. This does not make a large difference--most of the aggregate statistics are changed by about 10% or less.((For comparison, the standard deviation for total annual funding was 40% of the average, and the 75th percentile funding round raised 8x as much as the 25th percentile round.)) | This data does not reflect the overall amount of investment in AI. For example, it does not include spending by companies with their own funding (e.g. Google spending on AI R&D), government contracts, or any private investment not captured by Crunchbase. While collecting data, we found several cases of missing data from Crunchbase. This suggests that our dataset may be incomplete, possibly in a misleading way. We have not tried to fill in or correct for missing data or systematically check for bias missing data. |
| |
| Our methods are explained in more detail in the [[wiki:ai_timelines:ai_inputs:recent_trends_in_ai_investment#Methods|Methods]] section. | Some of our data can be found in [[https://docs.google.com/spreadsheets/d/1mHiSl3lRcqwcEZgPLt3muVM-5ZYoD4_W8lzBJ-KrGh0/edit?usp=sharing|this spreadsheet]].((Due to limitations in our Crunchbase license, we are unable to share our full dataset. If you have questions that you are unable to answer using the data we are sharing, please let us know.)) Our methods are explained in more detail in the [[wiki:ai_timelines:ai_inputs:recent_trends_in_ai_investment#Methods|Methods]] section. |
| |
| ==== Results ==== | ==== Results ==== |
| |
| Between January 1, 2018 and March 28, 2023, approximately 9,000 AI companies raised \$267 billion. Of that funding, \$263B went to 4,483 companies across 9,958 funding rounds. This represents about 19% of funding for all software companies during that time, and 4.6% of funding for all companies from any industry during that time. Monthly funding for AI and software is shown in Figure 1, and cumulative funding is shown in Figure 2. The large spike in AI funding in January 2018 is a $10 billion investment in OpenAI by Microsoft, which was the largest ever investment in an AI company, and was tied for the third largest investment ever in any software company, according to Crunchbase. | Between January 1, 2018 and March 28, 2023, approximately 9,000 AI companies raised \$267 billion. Of that funding, \$263B went to 4,483 companies across 9,958 funding rounds. This represents about 19% of funding for all software companies during that time, and 4.6% of funding for all companies from any industry during that time. Monthly funding for AI and software is shown in Figure 1, and cumulative funding is shown in Figure 2. The large spike in AI funding in January 2023 is a $10 billion investment in OpenAI by Microsoft, which was the largest ever investment in an AI company, and was tied for the third largest investment ever in any software company, according to Crunchbase. |
| |
| [{{:wiki:ai_timelines:ai_inputs:monthly_funding_for_ai_and_software.png?direct&600|Figure 1: Money raised by AI companies for each month (blue bars) and by all software companies (red line) from January 2018 to March 2023. The data for all software companies is averaged over each quarter and scaled down by a factor of 5 to make comparisons easier. The large amount of funding in January 2023 is due primarily to a $10B investment in OpenAI by Microsoft. Source: Crunchbase}}] | [{{:wiki:ai_timelines:ai_inputs:monthly_funding_for_ai_and_software.png?direct&600|Figure 1: Money raised by AI companies for each month (blue bars) and by all software companies (red line) from January 2018 to March 2023. The data for all software companies is averaged over each quarter and scaled down by a factor of 5 to make comparisons easier. The large amount of funding in January 2023 is due primarily to a $10B investment in OpenAI by Microsoft. Source: Crunchbase}}] |
| Funding for AI companies over the past five years has been a departure from the longer-term trend, which showed rapid, sustained growth, as shown in figures 3 and 4. | Funding for AI companies over the past five years has been a departure from the longer-term trend, which showed rapid, sustained growth, as shown in figures 3 and 4. |
| |
| [{{:wiki:ai_timelines:ai_inputs:annual_money_raised_since_1990.png?600|Figure 3: Annual funding for AI companies (blue bars) and all software companies (red line) since 1990. The vertical axis is on a logarithmic scale. Source: Crunchbase}}] | [{{:wiki:ai_timelines:ai_inputs:annual_money_raised_since_1990.png?600|Figure 3: Annual funding for AI companies (blue bars) and all software companies (red line) since 1990. The vertical axis is on a logarithmic scale. For 2023, the values are scaled based on the rate of funding so far, so they show what funding for the full year will be if it continues to accumulate at the same rate. Source: Crunchbase}}] |
| |
| [{{:wiki:ai_timelines:ai_inputs:total_money_raised_since_1990.png?600|Figure 4: Cumulative funding for AI companies (blue bars) and all software companies (red line) since 1990. The vertical axis is on a logarithmic scale. Source: Crunchbase}}] | [{{:wiki:ai_timelines:ai_inputs:total_money_raised_since_1990.png?600|Figure 4: Cumulative funding for AI companies (blue bars) and all software companies (red line) since 1990. The vertical axis is on a logarithmic scale. Source: Crunchbase}}] |
| |
| - For companies included in the [[https://gcrinstitute.org/papers/055_agi-2020.pdf|2020 Survey of Artificial General Intelligence Projects for Ethics, Risk, and Policy]] from GCRI((McKenna Fitzgerald, Aaron Boddy, and Seth D. Baum, 2020. 2020 Survey of Artificial General Intelligence Projects for Ethics, Risk, and Policy. Global Catastrophic Risk Institute Technical Report 20-1.)), we checked Crunchbase for recent funding rounds. Of the 72 projects included in the survey, 24 were associated with private companies and 11 of those had received funding since 2018. Three companies (Optimizing Mind, ORBAI, and FutureAI) had not been included in the original Crunchbase query, because they had not met the $3M cutoff. | - For companies included in the [[https://gcrinstitute.org/papers/055_agi-2020.pdf|2020 Survey of Artificial General Intelligence Projects for Ethics, Risk, and Policy]] from GCRI((McKenna Fitzgerald, Aaron Boddy, and Seth D. Baum, 2020. 2020 Survey of Artificial General Intelligence Projects for Ethics, Risk, and Policy. Global Catastrophic Risk Institute Technical Report 20-1.)), we checked Crunchbase for recent funding rounds. Of the 72 projects included in the survey, 24 were associated with private companies and 11 of those had received funding since 2018. Three companies (Optimizing Mind, ORBAI, and FutureAI) had not been included in the original Crunchbase query, because they had not met the $3M cutoff. |
| - To identify influential AI labs, we searched the [[https://paperswithcode.com/methods|Methods page]] on Papers with Code for machine learning methods that had been used in at least 400 publications. If one of the authors for the publication introducing the method had a private company as an institutional affiliation, that company was added to a list. This resulted in 15 companies or organizations within companies, which was then filtered to exclude any that had not received private funding since 2018, leaving only one (SenseTime). | - To identify influential AI labs, we searched the [[https://paperswithcode.com/methods|Methods page]] on Papers with Code for machine learning methods that had been used in at least 400 publications. If one of the authors for the publication introducing the method had a private company as an institutional affiliation, that company was added to a list. This resulted in 15 companies or organizations within companies, which was then filtered to exclude any that had not received private funding since 2018, leaving only one (SenseTime) that was not already included in our list of notable labs. |
| - Finally, we added companies that are known to us to be major players in AI development, or that commonly appear in lists of companies advancing the state of the art in AI. Only three such companies had recent funding data on Crunchbase, and had not already been included for other reasons (Cruise, Darktrace, and Stability AI). | - Finally, we added companies that are known to us to be major players in AI development, or that commonly appear in lists of companies advancing the state of the art in AI. Only three such companies had recent funding data on Crunchbase, and had not already been included for other reasons (Cruise, Darktrace, and Stability AI). |
| |
| === Inflation correction === | === Inflation correction === |
| |
| We obtained Consumer Price Index data from the Bureau of Labor Statistics((https://www.bls.gov/cpi/data.htm)). The dataset we used was for all prices for all urban consumers, seasonally adjusted. | We obtained Consumer Price Index data from the Bureau of Labor Statistics((https://www.bls.gov/cpi/data.htm)). The dataset we used was for all prices for all urban consumers, seasonally adjusted. The inflation data we used can be found in [[https://docs.google.com/spreadsheets/d/1mHiSl3lRcqwcEZgPLt3muVM-5ZYoD4_W8lzBJ-KrGh0/edit?usp=sharing|this spreadsheet]] under the tab "inflation_data". |
| |
| === Random sample of AI companies === | === Random sample of AI companies === |
| Plotting histogram of funding rounds by money raised on a log-log plot shows the data is fit reasonably well by a parabola, suggesting it is log-normally distributed. | Plotting histogram of funding rounds by money raised on a log-log plot shows the data is fit reasonably well by a parabola, suggesting it is log-normally distributed. |
| |
| {{:wiki:ai_timelines:ai_inputs:log_number_of_rounds_vs._log_money_raised_.png?600|}} | {{:wiki:ai_timelines:ai_inputs:histogram_of_funding_rounds_by_money_raised.png?600|}} |
| |
| //Primary Author: Rick Korzekwa// | //Primary Author: Rick Korzekwa// |
