Cost of key technologies relative to launch year

Hey there!

Welcome back to The Pulse, where we dive into interesting AI stories and trends backed by data, all presented through simple visuals.

> 99.7% fall in inference costs over last 2 years: ChatGPT's cost per 75-word response ~0 within its first year

> enabled due to

• hardware advances mainly: NVIDIA GPUs 105,000x more efficient over last 10 years

• algorithmic improvements such as distillation, quantization, etc.

> a16z also reported 10x dec in inference costs every year, for LLMs of equal performance

> AI cost reductions at least 20-70x faster than previous transformative tech

> Perplexity to raise $500M soon; on the lower range of expected $500M - $1B raise

> valuation to be $14B vs earlier expected $18B

> gets over $100M in ARR as of March, up 6.3% YoY as per CEO

> currently 15M weekly active users, up 50% from 10M at year beginning

> by 2050:

• humanoids market to cross $5T

• 1B humanoids expected → 90% for industrial & commercial

• much less for households overall, at 10% (US) & 3% (China)

• China to have most at 302.3M, followed by the US at 77.7M

• cost of humanoid in China to be much lower at $15,000

> adoption to accelerate in late 2030s