Marc Andreessen on why AI can spawn a new generation of big, important technology companies
Recent breakthroughs in artificial intelligence and machine learning are enabling computers to understand the world and respond intelligently to it. Google is already embracing these technologies for Android …
Context & Ripple Effects
Writing in 2016, Marc Andreessen laid out the bull case that machine learning breakthroughs would do what prior computing waves did: spawn an entirely new generation of major technology companies rather than merely upgrade the old ones. The intervening coverage tests that claim from both directions — Ars Technica traced the enabling stack back through CUDA, convolutional networks, and transformers, while a 2019 report flagged that AI research increasingly demands datacenter-scale computation, concentrating capability in a handful of incumbents.
First-order effects
- Google is the immediate proof point on the incumbent side: it folded AI and machine learning directly into Android, using its distribution advantage to absorb the technology into an existing product rather than ceding ground to newcomers.
- Foundation model makers like OpenAI and Anthropic are the strongest candidates for Andreessen's 'new generation' — companies whose core product is the intelligence itself, not a delivery vehicle for it.
Second-order effects
- Compute concentration forces a dependency structure: startups building on AI must rent datacenter capacity from the same hyperscalers they hope to displace, which is why Stratechery frames the real contest as centralized models versus open source.
- If AI is a sustaining innovation for Big Tech, as Google's AI-heavy I/O suggested, then rivals like Apple, Meta, Microsoft, and Amazon must each re-platform their own franchises around models — turning every incumbent into both a customer and a competitor of the model makers.
Third-order effects
- The unresolved question the corpus sets up is whether the next cohort of giant companies will be the model builders or whether incumbents' distribution and capital advantages re-absorb the value — the 2025 landscape review treats Apple, Google, Meta, Microsoft, Amazon, OpenAI, and Anthropic as one interlocking field precisely because that question is still live.
- If the pattern holds, AI joins the PC, internet, and mobile as a platform shift where the winners are decided by who controls distribution and compute — a structure that also raises the policy stakes Gates flagged around global access and inequity.
The trend: Andreessen's 2016 thesis that AI would mint a new generation of tech giants is being adjudicated now, with compute concentration pulling toward incumbents and independent model makers pulling toward his original prediction.