Defer ARM64 Pro Hardened build until Core42 commits — group decision Peter endorsed

April 27, 2026 at 10:12 PMstrategyhigh

Situation

In Apr 26 Sovereign AI response review meetings, the team — with Peter participating — decided the response language to Core42 will acknowledge that Pro Hardened on ARM64 (and FIPS-143 ARM certification) is contingent on a client commitment, not unilateral CIQ investment. ARM64 build estimated weeks not months once committed; FIPS-143 ARM is ~$200k / 4-6 months and gates on a deal commitment. Peter explicitly told the room: "We are going to need Nathan to say when. I am not going to be able to say on this call."

Reasoning

Don't oversell capability we don't have, don't burn engineering capacity to manufacture it on spec. ARM64 Pro Hardened is genuinely on the product roadmap as a CIQ-driven priority, but a single-customer custom build economics is different — Core42's commitment is the right unlock. FIPS-143 at $200k+ is the bigger version of same calculus; that's a big-order trigger. Peter framed responses as "weeks not months" and "acknowledge gaps directly — they already suspect optimism." Bjorn drove the FIPS-143 cost call ("yeah it's fine... if covered in deal"); Peter deferred ARM build estimate to Nathan.

Additional Context

Sovereign Response Review (noon Apr 26, recorded by Brian Dawson) drafted strategy; Sovereign AI Final Response review (7pm Apr 26, recorded by Peter) finalized language. Submission target Apr 27. Connects to Apr 17 reflection arc — same "reshape ambiguous customer asks into defensible structure" play applied to Core42 BIS Export License compliance matrix.

Observed Evidence

Two Fathom recordings on Apr 26. Peter participated in both, recorded the second. Direct Peter quote on ARM build: deferred estimate to Nathan. Direct Peter quote on framing gaps: "It's fine to say we have a gap. They expect that we have some gaps." Bjorn explicitly approved $200k FIPS-143 cost contingent on deal coverage. Group consensus on conditional language.

Matching Patterns

60%
Decouple to Protect Momentum(separate winning the deal from building deal-required custom work)
45%
Protect Engineering Capacity(don't burn ~1 month eng on a maybe, gate engineering investment on commercial signal)

Confidence Breakdown

25/35
Evidence
28/30
Pattern
15/20
Source
12/15
Corroboration

Reasoning Depth Analysis

Org Signal:Engineering investment in custom builds gates on commercial commitment, not on hope
Who Affected:Nathan (action: investigate ARM64 scope), Brian Dawson (action: ARM64 to product priorities parking lot), Bjorn (joint owner of response), Core42 (gets honest answer)
Precedent:Same play as Apr 17 — make trades explicit, reshape asks into defensible structure
Consequences:Real: response goes out conditional. Saves ~1 month of engineering if deal does not close. Costs nothing if deal does close (work just starts then).
Timing:Compressed — Sunday-night review for Monday submission

Source

reflection

AI Confidence

78%

Related Context

🎥
Soverign Response Review (Apr 26 noon)

fathom

We are going to need Nathan to say when. We can have that. I am not going to be able to say on this call.

🎥
Sovereign AI: Final Response review (Apr 26 7pm)

fathom

Decision: The current response is sufficient to advance. The engineering work will only begin after the client commits to moving forward.

Follow-up Todos

Suggest follow-up todo

Prioritize team running RLCH on GB200 hardware this week

@Nathan

From: Sovereign Response Review (Fathom 141129838) - Peter committed to prioritization in response to Brian: "Yes we can prioritize this... can we start? Yes. Can we finish? I don't know." Need to coordinate with Nathan on team capacity; Brian Dawson confirmed access already provisioned via Jeff. NOTE: Fathom auto-assigned a different action item to Peter ("Draft GB200 ARM Pro Hardened build/test plan") — that was a misattribution of the prioritization commitment.

Why: Peter explicitly said yes to prioritizing this on the call. Without action, the GB200 access sits idle and the Core42 ARM scope estimate Nathan owes the response remains blocked. This is gating evidence we can use to reinforce "weeks not months" claim if Core42 pushes back on timeline.

Outcome

No outcome recorded yet.

Decision ID: 02c53976-1fe9-4a02-9b12-8f3940308535