What is not done yet
No chip has been taped out and no chip exists. This is the plain list of what BramHA has not done, with a sentence or two on what each item means.
· 4 min read
IN SHORT
BramHA is in development. No chip has been taped out and no chip exists. The first design's specification is not published, and the process and the foundry have not been announced. There are no performance, power, or price figures. No safety certification is claimed. Nothing can be ordered, and there is no partner intake.
BramHA is Element 31's silicon for robots: the chip on the machine's side of the link, built on a mature process node, designed by AI Employees, with a named engineer signing every gate.
It is in development. The rest of this article is the list of what that phrase covers. Each item is stated as plainly as it can be, with a short note on what it means for a reader.
The part itself
No chip has been taped out, and no chip exists. Tape-out is the release of a finished layout to a foundry for fabrication. It has not happened. There is no fabricated part, no prototype, and no test chip. Everything the site says about what the silicon does is a statement of what it is being designed to do.
The first design's specification is not published. The first design is deliberately small: a real-time controller with a root of trust. It is not a vision processor and not an AI accelerator. Beyond that description, the specification is not public. There is no list of features, interfaces, or guarantees for an engineer to evaluate, so a robot maker cannot yet judge whether the part would fit a machine.
The process and the foundry have not been announced. The program intends to build on mature process nodes, because the jobs on the machine's side need certainty, not density. Which process, and which foundry, has not been announced.
What cannot be claimed
There are no performance, power, or price figures. Figures come from measuring a part, and there is no part. The site publishes no speeds, no power numbers, no timings, and no prices, and none should be inferred from the description of the jobs.
What can be said in general is a trade-off: mature nodes cost area and power per unit of compute. What mature nodes cost you covers that.
No safety or functional-safety certification is claimed. None is claimed, and none should be implied. The chip is being designed so that it could run a robot maker's control loop. That does not make it a safety system.
Safety, stops, and motion stay on the machine as its vendor built them. The safety case for a machine is the vendor's design and the vendor's responsibility. Where safety lives sets that out.
What cannot be ordered
Nothing can be ordered, and there is no partner intake. There is no part, no price, and no order process. There is also no program for robot makers to join. The silicon is intended, eventually, for the vendors and integrators who build machines, as a part they build in. Today the site only describes it. How robot makers would eventually obtain the part is one of the matters covered in what has not been decided.
What the program does have
A list of absences can mislead in its own way, so here is the other side: a thesis, a scope, and an intended way of working.
The thesis is that a robot's own silicon does jobs that need certainty, not density, and that mature process nodes are enough for them. The heavy AI work is served from inside the building by a Czar Grid, and the machine keeps the work that cannot wait and cannot fail.
The scope is four jobs: control, motion, sensing, and identity. The first design is limited to a real-time controller with a root of trust. A small first design does little, and that is intended.
The way of working is AI Employees doing the engineering, with a named engineer signing the specification, the design freeze, and the tape-out. The design work is being set up to run on Czar Forge, so that the design stays on hardware Element 31 controls.
None of those is a product. They are the plan that the list above is measured against. The page at /program keeps the current state of both.
Questions this raises
- Is there a prototype or a test chip?
- No. No chip has been taped out, and no chip exists in any form.
- When will any of this change?
- The program publishes one target: first silicon in November 2027. It is a target, not a promise, and it is the only date on the site. Each item changes when the thing it describes has actually happened, and the site is intended to say so then and not before.
- Why publish a list of what is missing?
- Because the site describes a program in development, and a reader deciding how much weight to give it should be able to see its state in one place.
KEEP READING