A 340 page huge report on AI trends - released by @bondcap Some...
Rohan Paul@rohanpaul_ai
108 views
Jun 01, 2025
~5 min read
Advertisement
1
A 340 page huge report on AI trends - released by @bondcap
Some wild findings from this report.
🧵1/n
Some wild findings from this report.
🧵1/n
2
🧵2/n
Meta’s Llama Downloads Exploded 3.4× in Eight Months.
an unprecedented developer adoption curve for any open-source LLM.
bondcap. com/reports/tai
Meta’s Llama Downloads Exploded 3.4× in Eight Months.
an unprecedented developer adoption curve for any open-source LLM.
bondcap. com/reports/tai
3
🧵3/n
AI Chatbots Now Mistaken as Human 73 Percent of the Time
In Q1 2025, testers mistook AI responses for human replies 73 percent of the time in Turing-style experiments. That’s up from roughly 50 percent only six months earlier—showing how quickly models have learned to mimic human conversational nuance
AI Chatbots Now Mistaken as Human 73 Percent of the Time
In Q1 2025, testers mistook AI responses for human replies 73 percent of the time in Turing-style experiments. That’s up from roughly 50 percent only six months earlier—showing how quickly models have learned to mimic human conversational nuance
4
🧵4/n
Rohan Paul@rohanpaul_ai
ChatGPT reached 365 Bn annual searches in just two years after its public debut (November 2022),
vs Google took 11 years (1998–2009) to hit that same milestone.
i.e. ChatGPT’s search volume ramped up about 5.5x faster than Google’s did.
-------
Source - bondcap research report
vs Google took 11 years (1998–2009) to hit that same milestone.
i.e. ChatGPT’s search volume ramped up about 5.5x faster than Google’s did.
-------
Source - bondcap research report
5
🧵5/n
Rohan Paul@rohanpaul_ai
NVIDIA GPUs Deliver 225× More AI FLOPS While Using 43% Less Power
Between 2016 and 2024, NVIDIA’s GPU line improved AI inference throughput by 225× and simultaneously cut data-center power draw by 43 percent.
That combination led to a staggering >30,000× increase in theoretical yearly token capacity per $1 billion data center
@NVIDIAAIDev
-------
Source: bondcap. com/reports/tai
Between 2016 and 2024, NVIDIA’s GPU line improved AI inference throughput by 225× and simultaneously cut data-center power draw by 43 percent.
That combination led to a staggering >30,000× increase in theoretical yearly token capacity per $1 billion data center
@NVIDIAAIDev
-------
Source: bondcap. com/reports/tai
6
🧵6/n
Rohan Paul@rohanpaul_ai
DeepSeek Secured 34% of China’s Mobile AI Users in 4 Months
DeepSeek went from zero to 54 million monthly active users in China between January and April 2025, capturing over a third of China’s total mobile AI market share in just four months.
Source: bondcap. com/reports/tai
DeepSeek went from zero to 54 million monthly active users in China between January and April 2025, capturing over a third of China’s total mobile AI market share in just four months.
Source: bondcap. com/reports/tai
7
🧵7/n
Annual AI Inference Token Revenue Potential Jumped from $240K to $7 Billion in 8 Years
In 2016, a $1 billion-scale data center could handle about 5 trillion inference tokens per year (~$24 million token revenue).
By 2024, that same $1 billion facility could process ~1,375 trillion tokens (theoretical revenue ~$7 billion)—a 30,000× shift in earnings opportunity
Annual AI Inference Token Revenue Potential Jumped from $240K to $7 Billion in 8 Years
In 2016, a $1 billion-scale data center could handle about 5 trillion inference tokens per year (~$24 million token revenue).
By 2024, that same $1 billion facility could process ~1,375 trillion tokens (theoretical revenue ~$7 billion)—a 30,000× shift in earnings opportunity
8
🧵8/n
inference costs per million tokens plunging nearly 99.7% from late 2022 to early 2025.
In September 2022, GPT-3.5 cost over $10 per million tokens. By mid-2023, those fees dropped to around $1
inference costs per million tokens plunging nearly 99.7% from late 2022 to early 2025.
In September 2022, GPT-3.5 cost over $10 per million tokens. By mid-2023, those fees dropped to around $1
9
🧵9/n
Cost declines of several technologies over time.
ChatGPT’s price for a 75-word response dropped nearly to zero within its first year.
Computer memory costs fell to almost zero in under two decades.
Electric power costs declined more gradually, reaching roughly 2–3% of their initial price after 60–70 years.
By contrast, the light bulb’s cost remained nearly flat over the same span
Cost declines of several technologies over time.
ChatGPT’s price for a 75-word response dropped nearly to zero within its first year.
Computer memory costs fell to almost zero in under two decades.
Electric power costs declined more gradually, reaching roughly 2–3% of their initial price after 60–70 years.
By contrast, the light bulb’s cost remained nearly flat over the same span
10
🧵10/n
IT consumer-price index (red line) from 1989–2024.
It shows compute requirements rising roughly 360% per year since around 2010—reaching 10²⁶ FLOP by 2024—while IT costs fell from an index of 100 to below 10.
In other words, training power has surged as hardware has become dramatically cheaper.
IT consumer-price index (red line) from 1989–2024.
It shows compute requirements rising roughly 360% per year since around 2010—reaching 10²⁶ FLOP by 2024—while IT costs fell from an index of 100 to below 10.
In other words, training power has surged as hardware has become dramatically cheaper.
11
🧵11/n
Performance of Top AI Models on LMSYS Chatbot Arena – 1/24-2/25, per Stanford HAI
Performance of Top AI Models on LMSYS Chatbot Arena – 1/24-2/25, per Stanford HAI
12
🧵12/n
Share of Developers Currently Using AI in Development Processes – 2023-2024,
per Stack Overflow
Share of Developers Currently Using AI in Development Processes – 2023-2024,
per Stack Overflow
13
🧵13/n
Cumulative AI-related GitHub repositories from 2015 through March 2024.
Major milestones—TensorFlow (2015), the Transformers paper (2017), Hugging Face’s “hf/transformers” (2018), GPT-3 (2020), Stable Diffusion (2022), and ChatGPT launch (2022).
From late 2022 to early 2024, repo count jumped roughly 175%, driven by expanding categories like infrastructure, model development, model repos, AI engineering, and applications.
Cumulative AI-related GitHub repositories from 2015 through March 2024.
Major milestones—TensorFlow (2015), the Transformers paper (2017), Hugging Face’s “hf/transformers” (2018), GPT-3 (2020), Stable Diffusion (2022), and ChatGPT launch (2022).
From late 2022 to early 2024, repo count jumped roughly 175%, driven by expanding categories like infrastructure, model development, model repos, AI engineering, and applications.
14
🧵14/n
Google Monthly Tokens Processed (T) – 5/24-5/25, per Google
Google Monthly Tokens Processed (T) – 5/24-5/25, per Google
15
🧵15/n
Microsoft Azure AI Foundry Quarterly Tokens Processed (T) – Q1:24-Q1:25, per Microsoft
Microsoft Azure AI Foundry Quarterly Tokens Processed (T) – Q1:24-Q1:25, per Microsoft
16
🧵16/n
NVIDIA Quarterly Revenue by Business Line ($B) – 1/19-1/25, per NVIDIA
NVIDIA Quarterly Revenue by Business Line ($B) – 1/19-1/25, per NVIDIA
17
🧵17/n
Google TPU (Tensor Processing Unit) Estimated Sales – 2021-2024, per Morgan Stanley
Google TPU (Tensor Processing Unit) Estimated Sales – 2021-2024, per Morgan Stanley
19
🧵18/n
Cloud Computing Revenue growth
Cloud Computing Revenue growth
20
🧵19/n
Foundation Model Estimated Revenue & Capital Raised – 5/13/25,
per Pitchbook, The Information, Bloomberg, The Wall Street Journal & CNBC
Foundation Model Estimated Revenue & Capital Raised – 5/13/25,
per Pitchbook, The Information, Bloomberg, The Wall Street Journal & CNBC
21
🧵20/n
Foundation Model Estimated Revenue Multiple – 5/13/25,
per Pitchbook, The Information, Bloomberg, The Wall Street Journal & CNBC
Foundation Model Estimated Revenue Multiple – 5/13/25,
per Pitchbook, The Information, Bloomberg, The Wall Street Journal & CNBC
22
🧵21/n
Estimated Enterprise Value / Next 12 Months Revenue Multiple – 5/25,
per Capital IQ & Bloomberg
OpenAI tops the list at roughly 24× revenue, followed by Duolingo at about 22×. Meta sits near 8×, Spotify at 7×, while Alphabet and Pinterest trade around 5× and 4.5× respectively.
A dotted line marks the median multiple of 6.9×, showing how OpenAI and Duolingo command significantly higher valuations relative to their projected revenues.
Estimated Enterprise Value / Next 12 Months Revenue Multiple – 5/25,
per Capital IQ & Bloomberg
OpenAI tops the list at roughly 24× revenue, followed by Duolingo at about 22×. Meta sits near 8×, Spotify at 7×, while Alphabet and Pinterest trade around 5× and 4.5× respectively.
A dotted line marks the median multiple of 6.9×, showing how OpenAI and Duolingo command significantly higher valuations relative to their projected revenues.
23
🧵22/n
2024 annual revenue per user .
OpenAI sits around $15 per user, below the $23 median. Alphabet leads at roughly $70, followed by Meta near $40 and Spotify at about $23. Pinterest and Duolingo trail at approximately $8 and $6, respectively.
So looks like OpenAI commands a high valuation despite only mid‐pack revenue per user.
2024 annual revenue per user .
OpenAI sits around $15 per user, below the $23 median. Alphabet leads at roughly $70, followed by Meta near $40 and Spotify at about $23. Pinterest and Duolingo trail at approximately $8 and $6, respectively.
So looks like OpenAI commands a high valuation despite only mid‐pack revenue per user.
24
🧵23/n
Specializations of Ten Leading AI Companies – 4/25, per The Wall Street Journal
Specializations of Ten Leading AI Companies – 4/25, per The Wall Street Journal
25
🧵24/n
ChatGPT’s paid user base soaring from zero in October 2022 to roughly 20 million by April 2025—a compound annual growth rate of about 153%.
On the right, OpenAI’s revenue jumps from near zero in 2022 to nearly $3.7 billion in 2024, a staggering 1,050% year-over-year increase.
ChatGPT’s paid user base soaring from zero in October 2022 to roughly 20 million by April 2025—a compound annual growth rate of about 153%.
On the right, OpenAI’s revenue jumps from near zero in 2022 to nearly $3.7 billion in 2024, a staggering 1,050% year-over-year increase.
26
🧵25/n
Anthropic’s rapid revenue surge: annualized income grew from near zero in September 2023 to $2 billion by March 2025—about a 6.4× increase per year and over 20× in eighteen months.
Anthropic’s rapid revenue surge: annualized income grew from near zero in September 2023 to $2 billion by March 2025—about a 6.4× increase per year and over 20× in eighteen months.
27
🧵26/n
Perplexity’s annualized revenue climbing from near $16 million in March 2024 to $120 million by May 2025—a roughly 7.6× increase in fourteen months.
Perplexity’s annualized revenue climbing from near $16 million in March 2024 to $120 million by May 2025—a roughly 7.6× increase in fourteen months.
28
🧵27/n
The median time it took top AI startups (blue) versus top SaaS startups (red) to reach $5 million in annualized revenue.
AI companies hit that $5 million ARR mark in about 24 months, whereas traditional SaaS peers took around 37 months—a 35% faster ramp for AI businesses.
The median time it took top AI startups (blue) versus top SaaS startups (red) to reach $5 million in annualized revenue.
AI companies hit that $5 million ARR mark in about 24 months, whereas traditional SaaS peers took around 37 months—a 35% faster ramp for AI businesses.
29
🧵27/n
Grok’s desktop visits from December 2024 through April 2025. From near zero in January 2025, traffic surged to about 30 million visits in February—after Grok 3’s February 17 release—and then jumped to roughly 140 million by March and 150 million by April.
Grok’s desktop visits from December 2024 through April 2025. From near zero in January 2025, traffic surged to about 30 million visits in February—after Grok 3’s February 17 release—and then jumped to roughly 140 million by March and 150 million by April.
30
🧵28/n
Google embedding “AI Overviews” directly into search results (1.5 billion monthly users as of April 2025)
Gemini chatbot MAUs climbing from 350 million in March 2025 to 400 million by May 2025.
Google embedding “AI Overviews” directly into search results (1.5 billion monthly users as of April 2025)
Gemini chatbot MAUs climbing from 350 million in March 2025 to 400 million by May 2025.
31
🧵29/n
How often “AI” was mentioned on earnings calls from Q1 2020 to Q1 2024.
That spike reflects how AI became a central topic for tech earnings discussions.
How often “AI” was mentioned on earnings calls from Q1 2020 to Q1 2024.
That spike reflects how AI became a central topic for tech earnings discussions.
32
🧵30/n
Zoom’s AI Companion adoption from September 2023 to December 2024.
Active accounts rose from near zero in late 2023 to about to over 4 million by December 2024
Zoom’s AI Companion adoption from September 2023 to December 2024.
Active accounts rose from near zero in late 2023 to about to over 4 million by December 2024
33
🧵31/n
ChatGPT Enterprise’s rapid uptake, 2 million by February 2025.
over 80% of Fortune 500 teams adopted it within nine months
ChatGPT Enterprise’s rapid uptake, 2 million by February 2025.
over 80% of Fortune 500 teams adopted it within nine months
34
🧵32/n
The Internet took 23 years to reach 90% outside North America, whereas ChatGPT hit 90% by its third year.
The Internet took 23 years to reach 90% outside North America, whereas ChatGPT hit 90% by its third year.
35
🧵32/n - Enterprise SaaS Incumbent AI Rollouts
In three months since launch, Salesforce Agentforce secured 3,000 paid deals (up from 1,000 at 12/24), powering $900M Data Cloud and AI ARR (up 120% YoY).
Data Cloud now holds 50 trillion records, fueling productivity, efficiency, and cost savings.
In three months since launch, Salesforce Agentforce secured 3,000 paid deals (up from 1,000 at 12/24), powering $900M Data Cloud and AI ARR (up 120% YoY).
Data Cloud now holds 50 trillion records, fueling productivity, efficiency, and cost savings.
36
🧵33/n
In 25 months since launch in March 2023, AnySphere’s Cursor AI ARR soared from $1 million to $300 million by April 2025
In 25 months since launch in March 2023, AnySphere’s Cursor AI ARR soared from $1 million to $300 million by April 2025
37