Interesting research from MIT on the LLM inference market (~$35-60B in 2025) reveals surprising dynamics that challenge both the “AI is a commodity” and “winner-take-all” narratives. Approximately $25B in annual spending — roughly 70% of the market — represents potential inefficiency.
I think [1] is a good definition of the architectural pattern (outside of networking applications). It’s the one we use within the crossplane community [2]
Good to know, at least that warm me up to crossplane further.
The messaging might need update, including within the docs.
I mean - crossplane is the OAM implement - coupled with OAM sprinkled all over docs, gave me very different impression.
This aside, I think crossplane work is interesting.
[disclosure: I started the crossplane project and ceo/founder at upbound]
If you look at the governance of Crossplane [1], you'll see the three original maintainers of Crossplane (all Upbound employees) will be bootstrapping the steering committee and have a 2 year term, 2 others are currently being added without upbound affiliation. We believe this bootstrapping period is important to ensure that the project remains true to its vision, and to build trust and rapport with new folks joining the project. After the bootstrap period ends, the community will vote new members in. This is the same way Kubernetes and the CNCF Technical Oversight Committee were bootstrapped.
Upbound will have a commercial solution around Crossplane, but our commercial goals are not at odds with creating a strong open source community with strong governance. For Crossplane to be successful it's going to take a village, this is why we're donating it to the CNCF.
I made a similar argument [1][2] about tipping the cloud computing market from vertical to horizontal integration. One important aspect of this transformation is to maintain the feel of an integrated cloud provider, and not let the customer/end-user deal with the cost of heterogeneity [3].
+1 we really wanted to leverage the ecosystem and have something that is immediately familiar when we decided to use kube-apiserver (and etcd) for crossplane. I think the K8S resource model [1] goes well beyond container orchestration
Yes, indeed. There appears to be an overlap in general principles (even in use-case example). From what it appears, KEP sig was "published/released" (not sure what is the right term here) on the GitHub 3 days ago, we did not see this. We will definitely reach out for further discussion.
