The AI revolution is 'unprecedented' in the scale of human history, new report finds

The maturation of artificial intelligence (AI) successful our lives is officially unprecedented, according to nan female dubbed nan "Queen of nan Internet". 

Mary Meeker, a task capitalist known for her net inclination insights, released a 340-page study titled 'Trends - Artificial Intelligence,' wherever copious investigation and charts show what she characterises arsenic nan “unprecedented” gait of alteration successful which AI is being invested in, developed, adopted, and used. 

Meeker ran superior patient Kleiner Perkins’ maturation believe during nan 2010s, wherever she invested successful early tech giants for illustration Facebook, Spotify, and Canva. 

A floor plan successful Forbes credits her pinch predicting nan emergence of Apple, Google, and nan integer system pinch her erstwhile inclination reports. 

“To opportunity nan world is changing astatine unprecedented rates is an understatement,” Meeker, now nan co-founder of task superior money Bonds, writes successful nan caller report’s introduction. 

ChatGPT is history's 'biggest overnight success'

One of nan staggering charts of nan study shows that ChatGPT, OpenAI’s AI chatbot, reached 800 cardinal users by April 2025, a mates of years aft its first motorboat successful October 2022. 

The company’s gross besides skyrocketed successful a akin fashion; from zero successful 2022 to conscionable nether $4 cardinal (€3.5 billion) by 2025.

Based connected personification data, nan emergence of ChatGPT, successful particular, is history’s biggest “overnight success,” 9 years aft nan institution was founded successful precocious 2015, Meeker wrote.

"And, dissimilar nan Internet 1.0 revolution, wherever exertion started successful nan USA and steadily diffused globally - ChatGPT deed nan world shape each astatine erstwhile increasing successful astir world regions simultaneously," according to nan report. 

The study shows that ChatGPT reached 100 cardinal world users successful little than 2 months aft its launch, nan fastest exertion to do so. 

In comparison, it took Facebook 4.5 years to scope nan aforesaid number of users aft its motorboat successful nan early 2000s. 

Meeker estimates successful nan study that it will return 3 years for a mostly of households to adopt AI technology, down from nan 12 years it took for households to commencement utilizing desktop net regularly. 

Companies being 'extremely aggressive' pinch AI development

There are various factors astatine play for nan “rapid and transformative” emergence of AI, Meeker writes successful nan report. 

There’s wide buy-in from caller AI institution founders and much accepted companies for AI adoption, Meeker contends, demonstrated by rate flows being "increasingly directed" towards AI "in efforts to thrust maturation and fend disconnected investors". 

The study notes that technology’s biggest players, including NVIDIA, Google, Meta, Microsoft, and China’s Baidu person each accrued nan mentions of AI successful their firm net reports to shareholders since 2022. 

That, paired pinch a caller activity of AI institution founders that are "extremely aggressive" pinch each stages of nan AI merchandise improvement - from innovation, to merchandise releases, acquisitions, rate burn, and superior raises - intends that AI “user and usage trending is ramping materially faster” than before. 

The study besides noted that nan number of caller AI models has gone up 167 percent twelvemonth complete twelvemonth since 2020, and nan size of nan information sets they are utilizing is up 260 percent successful nan aforesaid period. 

Meeker notes that nan alteration successful costs to create caller models is besides unprecedented. 

Citing Stanford research, nan study shows conclusion costs for those that usage nan tech person dropped 99 per cent complete 2 years, moreover though nan costs of training a exemplary is up to $1 cardinal dollars (€850 million). 

The gait astatine which competitors tin lucifer each different connected nan marketplace is unparalleled. For instance, Meeker notes that NVIDIA’s 2024 Blackwell GPU chip, which helps train AIs to do what users expect, has 105,000 times little power per token than nan 2014 Kepler model. 

"It’s a staggering leap, not conscionable of costs reduction, but of architectural and materials invention that is reshaping what’s imaginable astatine nan hardware level," she writes.