Mobile computing and AI
NVIDIA, Apple and OpenAI: The AI-computing era
From parallel computing and efficient silicon to widely used conversational AI.
In this lesson
The main idea
The current wave comes from hardware-software coevolution: parallel processors, programming tools, data centers, large models and accessible interfaces.
Companies and milestones
- 2006 — NVIDIA CUDA: opened GPU parallelism to general computing through a programming model and tools. Type: hardware + software.
- 2020 — Apple M1/Apple silicon for Mac: integrated CPU, GPU, memory and a neural engine with an emphasis on performance per watt. Type: hardware + software integration.
- 2022 — OpenAI ChatGPT: made interaction with a language model through conversation accessible to a broad audience. Type: software/cloud AI.
- 2024 — NVIDIA Blackwell: a GPU and interconnect platform aimed at training and serving large-scale AI models. Type: hardware + software ecosystem.
- Through 2026 — the race continues across chips, models, cloud and devices; no single leap erases earlier layers. Type: both.
Why was this a qualitative leap?
The AI surge is neither software-only nor hardware-only; CUDA, chips, cloud, models and conversational interfaces are interdependent.
Check your understanding
Classify CUDA, M1, ChatGPT and Blackwell, then explain why the categories overlap.
Answer
CUDA is software tied to GPUs; M1 and Blackwell are hardware with software ecosystems; ChatGPT is software dependent on large-scale hardware and cloud. The label marks the center of the shift, not every dependency.