True. It's a massive shift of power, all being centralized.
As you mentioned, they know they need good data though, so they might actually try to find some equilibrium.
If not, it's possible that the creation of new valuable content, to feed the LLMs, will be produced in-house by the AI labs. Sounds insane, but Netflix also makes their own content.
I think the AI labs will become so big that they'll take on more roles than just offering LLM inference. I think they'll become as or more powerful than many current nation state governments.
> True. It's a massive shift of power, all being centralized.
It may seem that way in the short term. But in the long term, the tendency in technical development is for the infrastructure and capital requirements for new technology to start off very high, but then shrink over time, such that use cases that required massive amounts of upfront investment in the early stages become incrementally more viable at smaller and smaller scales.
People were saying the same things about computers in general in the 1960s as they are about GenAI now. That was an era when computer technology itself had developed to a point when it was economically impactful, but still only affordable to large institutions. People making predictions that increasing use of computers would lead to massive centralization of economic and cultural power didn't predict that merely twenty years later, computing power equivalent to contemporary mainframes would be available in a convenient desktop box available at a local shop to any individual or main-street business that cared to buy one.
The widespread availability of computing technology from the '80s to present actually had the opposite effect, and led to quite a bit of decentralization, as enterprising individuals and new startups started applying that technology to do at small scales what only large enterprises could do before. In fact, a lot of the reaction to AI in its current stage may actually be because it's disrupting the expectation of decentralization and autonomy over our technology that the personal computing revolution established in the first place.
Like most new technologies, GenAI in its initial stages has required massive infrastructure investments that have led to the early iterations being offered by centralized institutions, but that might not last. Open-source AI models are approaching the capabilities that the big players' frontier models had arrived at only a couple of years ago.
In 2026, we're already at the point where local inference is economically viable for commercial use cases -- at my own company, while we do use our Claude Enterprise account for a variety of use cases, it turned out to make much more sense in terms of both cost and risk exposure to instead process certain datasets (e.g. a large volume of phone recordings containing PII) with local models running on commodity GPUs. That proved to be entirely effective, and the one-time hardware investment (which created a bookable asset for the company) turned out to be less than the cost of running the same task on Claude (which would have been pure OpEx).
Your positive view makes sense to me and is refreshing. Let's see how things play out. So the pattern will be: that which can be done with smaller models will be decentralized first; gradually the more advanced stuff will become within reach.
I already do use google search's AI Mode (probably a 300B model) for many quick questions. Local models would be great for things like checking my email and many other things (sensitive + continuously running = not suitable for Opus). My 64GB DDR4 laptop can already run Qwen 47B at .7 tok/s, that's already usable for some usecases (overnight stuff mostly).
As you mentioned, they know they need good data though, so they might actually try to find some equilibrium.
If not, it's possible that the creation of new valuable content, to feed the LLMs, will be produced in-house by the AI labs. Sounds insane, but Netflix also makes their own content.
I think the AI labs will become so big that they'll take on more roles than just offering LLM inference. I think they'll become as or more powerful than many current nation state governments.