It’s not your imagination: AI is speeding up the pace of change

If AI acceptance feels different than any technical revolution that you have experienced before – mobile, social, cloud computing – it is actually.

Venture Capitalist Mary Meeker has just fallen A 340 -page slide report – who used the word ‘unprecedented’ on 51 of those pages – to describe the speed with which AI is developed, adopted, spent on and used, supported with the graph after the graph.

“The pace and scope of change with regard to the evolution of artificial intelligence technology is indeed unprecedented, as supported by the data,” she writes in the report, “trends – artificial intelligence”.

There is a certain poetic history for this person who writes this kind of report. Meeker is the founder and general partner at VC Firm Bond and was once Known as queen of the internet For its earlier annual reports from internet trends. Before she founded Bond, she led the growth fraction of Kleiner Perkins, from 2010-2019, where she supported companies such as Facebook, Spotify, Ring and Block (then square).

She has not released a trend report since 2019. But she has dusted her skills, in laser detail, to document how AI adoption has surpassed any other technology in human history.

Chatgpt reach 800 million users in 17 months: unprecedented. The number of companies and the rate with which so many others achieve high annual recurring entry homes: also unprecedented.

The speed at which the operating costs fall: unprecedented. Although the costs of training a model (also unprecedented) are a maximum of $ 1 billion, the conclusion costs – for example, those who pay to use the technology – have already fallen 99% for two years, when calculating the costs per 1 million tokens, she writes, which quotes, with reference to research Van Stanford.

The pace at which competitors match each other’s characteristics, at a fraction of the costs, including open source options, in particular Chinese models: unprecedented. For example, she points out that Nvidia’s 2024 Blackwell GPU uses 105,000x less energy per token than the 2014 Kepler GPU predecessor 2014.

In the meantime, Chips from Google, just like its TPU (Tensor Processing Unit) and the Trainium of Amazon, are developed on a scale for their clouds – that also moves quickly. “These are not ancillary projects – they are fundamental bets,” she writes.

The only area where AI is not exceeded any other technical revolution is in financial returns. While VC’s money on the AI ​​fire as soon as possible, AI companies and cloud service providers also burn through cash. AI requires huge investments in infrastructure.

That is good for consumers and companies, the beneficiaries of rapid improvements, while the competition lowers the costs, Meeker notes. But the jury is still out of which of the current harvest of companies in the long term, profitable, subsequent generation of technical giants. “Only the time will tell which side of the money comparison will land the current AI aspirants,” she writes.

Regarding the rest of us: just hold your hats.