A reflection on 2025: the chair's year-end message
Hello NoLimit community,
A warm welcome to every new member, and to everyone who has supported the community so far. NoLimit is an ambitious project with a lofty and worthy goal: the best Ethernet-based high-performance network for AI and HPC workloads.
As 2025 winds down, here is what we accomplished in a very short time and where we are heading in 2026.
We started in October with a charter, some aggressive goals and timelines, and a lot of passion. Along the way we began to organize ourselves and find our footing:
- We established a Technical Advisory Committee whose job is to keep every project in the foundation coherent. The committee took that seriously and immediately set about herding ideas into a vision of what the first architecture release should look like.
- We began forming alliances with key industry associations. We already have partnership or liaison agreements with OCP and IEEE 802.3, and we are in discussions with organizations focused on management, storage, software, networking, and compliance and testing.
- We chartered eight working groups (we started with four) to deliver a truly workload-tuned Ethernet protocol built on a solid technical foundation.
- Roughly 40 companies joined in under two months, with about 30 more requesting the charter documentation. We went from around 50 individual participants to more than 300 in less than a month. By any measure we are one of the fastest-growing efforts of this kind.
- Our progress is being followed and reported by the trade and technical press. We have struck a nerve, and we are filling a very real need.
None of this would have happened without the working group chairs, draft authors, and contributing volunteers. That level of dedication deserves to be called out specifically, because it has been impressive.
What about 2026?
As chair I have been working from one principle: we will develop a large-scale, highly performant network that rivals and surpasses anything available today. I also work from a simple philosophy. If you are not going to do it right, when are you going to do it over? We cannot simply say we will ship a 1.0 and then race to a finish line while parts fall off the car. A first release that does not work, or is so full of holes you could float a cruise ship through it, helps nobody.
With that in mind, my objectives for 2026:
- Ratify at least a minimal set of drafts across all working groups that can be implemented in practical, real-world scenarios as the NoLimit 1.0 architecture.
- Establish a compliance and testing program where vendors can validate products and designs against the architecture.
- Develop a certification program so end users can implement NoLimit solutions with confidence.
- Nurture broad ecosystem adoption alongside sister standards and specification organizations.
- Maintain a collegial, approachable, and friendly working environment for every contributor.
The culture of this foundation is meant to encourage cross-pollination of skills and talents across the Ethernet ecosystem, and that should show in how we treat each other too. I challenge everyone to think about how we reach these goals by the end of 2026.
One last thing
To those who have been involved since the beginning: my sincere gratitude. This year was not easy and it was not simple. Yet we have built, and are still building, something we can be proud of, and I intend to push it further with your help. Thank you for the hours, the hard work, and the dedication to this experiment turned experience.
To every member, novice and veteran, thank you for supporting NoLimit and working to fulfill its promise as the premier AI and HPC networking standard. On behalf of the steering committee, we look forward to even greater things in 2026.
Best,
Everett Caldwell
Chair, NoLimit Foundation