Strategy / IR / industrial capital / IT distribution

Listed-company and industrial collaboration

What you want is a position you can verify before you decide to make it your own P&L: a gated joint test of about ten weeks (G1–G4), no financing at that stage, stop-loss if a gate is missed. What passes can become contract and disclosure: downstream token-supply / offtake may be co-signed by you and Xinyuanyi as joint suppliers; a formal contract has not been signed. We offer no services tied to secondary-market performance, and make no such statements.

First confirm the owning team (strategy / IR / industry / supply chain) and whether a 2026 plan exists. Materials on request, no strings.

Value to you

For strategy and IR, the value is a cheap falsification: no financing first. You run G1 delivery, G2 single-node baseline, G3 main gate and G4 stability on your own criteria, and you stop if a gate fails.
The position need not stop at a financial stake. After a deal, offtake may be co-signed as joint suppliers, so you obtain recognisable revenue, a customer relationship, and a disclosable industrial-collaboration or tech-finance case. A formal contract has not been signed; this is a prospective condition.
The technical hook is checkable: signed throughput +29–40%, TTFT −26–32% versus local NVMe (reports R2/R3), on our own hardware. If supply and finance sit in one group, one introduction is enough.
Continuity: 278 chassis / 2,224 cards are purchased inventory; about 600 chassis are a prospect pipeline (intent, not held); mid-term plan 5,000-card class. That supports adding, not stacking into one stock figure.
The red line is stated first: we offer no services tied to secondary-market performance, and make no such statements.

Agenda for a first formal meeting

Owning team: strategy, IR, industry development, or supply-chain / IT distribution; whether a 2026 plan or budget line exists.
Acceptance: whether you add your own metrics on top of the G3 main gate.
Form after a pass: joint supply, joint operations, or equipment / storage purchase (a formal contract has not been signed).
Disclosure: how far an industrial-collaboration or tech-finance case may be written; nothing about the secondary market.
Whether group leasing or SCF joins in the same introduction.

What this means in a term sheet

278 chassis / 2,224 cards are purchased inventory; about 600 chassis are a prospect pipeline (intent, not held) — not stacked into one stock figure.
The partnership can become your own line of business: downstream token-supply / offtake may be co-signed by you and Xinyuanyi as joint suppliers; a formal contract has not been signed.
This can become a disclosable industrial-collaboration or tech-finance case. We offer no services tied to secondary-market performance, and make no such statements.
A gated joint test of about ten weeks (G1–G4) is a low-cost position; stop-loss if gates are missed. No financing is required at that stage.

Why we can deliver this

Five reasons this is not another deck. Figures carry a denominator or a report ID. Economic numbers are engineering-model estimates, delivered as runnable Python after NDA.

Purchased kit, countable
278 eight-GPU chassis / 2,224 cards on the books; value on tax documents. About 600 chassis are a prospect pipeline (intent, not held) and are not added into stock.
Acceleration is our line
FX arrays are in production; KV tiering is source-adapted by us, not a resale. Signed throughput +29–40%, TTFT −26–32% (R2/R3 vs local NVMe).
Pre-production acceptance is done
8×MI308X, 1T-parameter INT4 MoE, 3,210 samples across four benches, official delta inside the 2% red line.
Downstream progress is not overstated
Talks and joint tests with Zhipu, Kimi and MiniMax; MiniMax is in token-supply signing, and a formal contract has not been signed. Communication is not collection.
Verify first, stop-loss built in
Joint test first, decisions second: gate-based acceptance with built-in stop-loss. G1 Week 2 / G2 Week 4 / G3 Week 7 / G4 Week 10. Full model after NDA.

On this track specifically

There is kit and there are reports: 278 chassis purchased, value on tax documents; signed R1–R9 / R2/R3 reports in the evidence library — not a consultant’s sector note.
There is a method: gate weeks and pass/fail rules are written (weeks 2 / 4 / 7 / 10); your team can sit in the test design.
There is a product, not a resale contract: FX arrays plus source-level KV work; exclusivity inside the collaboration can be discussed.
Downstream progress is real and not overstated: MiniMax is in commercial signing, and a formal contract has not been signed.
The pipeline is not mixed with stock: 2,224 cards purchased; about 600 chassis intent, not held; 5,000-card-class plan.

Next actions and dates

D0 (today)
Name the owning team and any 2026 plan; say which pack you want (one-pager / report index / joint-test note).
48 hours after NDA
Diligence index, gate note and scripts on request — no investment condition attached.
Week 1
Align acceptance criteria and a visit; restate the disclosure red line.
Week 2–Week 4
G1 Delivery acceptance → G2 Single-node baseline.
Week 7 G3
Main joint-test gate: KV tiering acceleration: TTFT reduction ≥25% and throughput gain within the measured +29–40% band; 8 nodes sharing one array approach the NIC ceiling on read-back.
Week 10 G4
Stability & acceptance. Structure and joint supply only after a pass; stop-loss if not. No financing at this stage.

Verifiable progress

Purchased: 278 × 8-GPU AMD Instinct MI308X chassis (2,224 cards); equipment value to be shown on tax documents; mid-term plan 5,000-card class.
Pre-production acceptance on a single 8×MI308X node for a 1T-parameter INT4 MoE model: 3,210 samples across four accuracy benches all passed (official delta −0.8% to −1.5%, inside the 2% red line).
Deployment adaptation and performance tests with MiniMax are complete; token supply is in the commercial-signing process; a formal contract has not been signed.
Signed storage-acceleration result: throughput +29–40%, TTFT −26–32% (Qwen3-Coder-480B-FP8 long-context cold recovery vs local NVMe; reports R2/R3).

Evidence library →

Send an inquiry All five tracks

Contact Wang Limin · +86 13161818898 · mingxin@agentmail.to

Other counterparties

This page is a public description of one-to-one B2B cooperation. It does not constitute an offer, an investment offer, or any return promise. Figures are the 2026-09-15 external baseline; economic numbers are engineering-model estimates, not signed purchase or sales prices.