Cold Start Strategy for the First

Cold Start Strategy for the First Defense Customer KAI.COM CORE STRATEGIC ROADMAP

This problem represents the absolute core challenge for all platform-based monopoly companies—the classic chicken-or-egg dilemma. Defense customers cannot entrust critical computing power to an unproven system, yet the system can never scale up to prove itself without defense customers. The breakthrough point lies not in the “technical solution,” but in the incremental building of trust.

Analysis of the First Defense Customer Cold Start Prerequisite: Defense customers will never hand over core operational nodes on day one The US Department of Defense (DoD) procurement path is notoriously conservative —averaging 3-5 years from Proof of Concept (PoC) to production deployment. You cannot expect the Pentagon to hand over nuclear submarine acoustic signature identification on day one. Your path must follow: edge entry, value validation, and step-by-step upgrades. KAI.com Strategic Confidential

Step 1: Do Not Call It “Compute Scheduling”; Call It “Cost Optimization PoC” Target Customers: The Defense Innovation Unit (DIU) or the Chief Digital and AI Office (CDAO) under the US Department of Defense. Entry Point Point-of-Attack: CORE PITCH SCRIPT: “We do not schedule your sensitive missions. We only take over and optimize your commercial compute procurement across AWS, Azure, and GCP. You run a massive volume of unclassified training workloads on commercial clouds (logistics models, personnel forecasting, large language model fine-tuning, etc.). We help you reduce these compute costs by 20-30%. No connection to sensitive networks required, no data leaves the designated boundary. Pure cost optimization.” KAI.com Actual Operations: The Pentagon’s unclassified training tasks account for 40% of its total compute spend. These workloads are latency-insensitive and have low data security requirements. They can be dynamically routed to idle overnight compute capacity in locations like Iceland or Quebec. KAI.com’s initial “liquidity pool” consists of: Azure Reserved Instances + AWS Spot Instances

Google Committed Use Discounts. The Form of the First Deal: Not “nuclear weapons simulation scheduling,” but a “Pentagon Budget Optimization Tool.” Sign a 12- month SaaS contract guaranteeing savings ≥15% , with a money-back differential guarantee. Why This Works: Between 2016 and 2020, DIU validated countless projects this way—starting small, saving money with commercial tech, establishing trust, and upgrading. JEDI and JWCC followed exactly this trajectory. • • • • • • KAI.com Strategic Confidential

（CUI）"

Step 2: Transition from “Unclassified” to “Controlled Unclassified Information (CUI)” Scenario Assumption: The first PoC operates for 6 full months, successfully verifying a 22% reduction in compute expenditure. Upgrade Path & Expansion PHASE 2 PITCH SCRIPT: “A portion of the training workloads you run on Azure/AWS contains Controlled Unclassified Information (CUI)—such as troop readiness metrics and deployment logistics. These shouldn’t run on public clouds, but your on- premise data centers lack space. We have constructed a ‘Trusted Compute Pool’: workloads only route to Pentagon-pre-approved data centers; all routing records are encrypted on- chain, rendering them fully auditable and immutable; strict physical isolation: KAI.com’s orchestration layer never touches data, only metadata.” Trust Progression Mechanism: Having cleared Phase 1, the Pentagon recognizes KAI.com’s technical stability. The focus now shifts to safety protocols. Engage the NSA or NIST to spearhead SOC 2 Type II + FedRAMP certification, dedicating 6 months to rigid compliance audits. Critical Milestone: Securing FedRAMP Authorization—the definitive gateway for US Federal government cloud service procurement. While certification averages 12-18 months and costs $1M-$3M, once achieved, every federal agency in the US can purchase directly. KAI.com Strategic Confidential

Step 3: Introduce the “Strategic Spare Parts” Concept Strategic Transformation: Evolve from a “cost- saving tool” to a core “insurance mechanism.” PIVOT IN PITCH STRATEGY: “We have already saved you $20 million. Now, let’s discuss something far more critical— compute elasticity insurance. Your mission- critical tasks (missile early warning, battlefield situational awareness) run on sovereign data centers. However, on-premise infrastructure only covers 40% of peak demand. The remaining 60% is currently handled via fragmented commercial clouds and delayed scheduling queues. We propose the KAI.com Strategic Spare Parts Program: Pay a 5% monthly readiness fee, and we guarantee the reservation of equivalent compute pools globally. In emergencies, call upon them instantly at market rates. Normally, these pools run commercial workloads, but you retain a 48-hour exclusive reclamation right.” The Nuclear Submarine Logistics Analogy: A navy does not build 100 nuclear submarines to sit idle in shipyards waiting for war; instead, 30 are deployed, 30 are in maintenance, and 30 serve as strategic reserves. Reserves remain offline normally but activate instantly during conflict. KAI.com’s Strategic Spare Pool = The “Defense Mobilization Mechanism” for Compute. Why This Resonates: During the 2022 Ukraine Conflict, multiple AI defense firms saw their compute capacity squeezed out by commercial clients, delaying critical training queues by weeks. Every Ministry of Defense intimately understands the pain point of “compute elasticity insurance.” KAI.com Strategic Confidential

Step 4: Scale from the US to the “Allied Compute Sharing Agreement” Vision Upgrade: One country is insufficient; the ultimate goal is alliance-scale network effects. INTERNATIONAL ALLIANCE PITCH: “You signed the AUKUS pact with the UK, Australia, Canada, and New Zealand, which explicitly includes AI and quantum technology sharing clauses. But how is compute shared? Currently, everyone builds their own silos. KAI.com can serve as the AUKUS Compute Orchestration Layer: Defense AI tasks across these four nations can be cross-scheduled on our platform. Idle inference capacity in the US → Australian UUV acoustic analysis; UK overnight idleness → Canadian Arctic surveillance inference. All data flows securely encrypted strictly within allied boundaries, avoiding any non-AUKUS data centers.” Crucial Geopolitical Leverage: AUKUS is already breaking sovereignty barriers regarding nuclear submarine tech sharing; compute sharing is the next logical step. Once AUKUS compute sharing goes live, the progression from Five Eyes → NATO → Global Ally Expansion becomes an inevitable domino effect.

Complete Cold Start Roadmap for the First Defense Customer KAI.com Strategic Confidential

Timeline Strategic Milestones & Execution Details T+0 Months DIU PoC Launch: Cost optimization tool managing unclassified commercial cloud procurement. Contract value: $5M/year, guaranteed savings ≥15%. T+6 Months Success verified (22% saved) → Initiate FedRAMP certification pipeline → Introduce second unclassified task (logistics optimization). T+12 Months Achieve FedRAMP Moderate Authorization → Formally scale up to Controlled Unclassified Information (CUI) workloads. T+18 Months Roll out “Strategic Spare Parts” initiative → Backup orchestration for the first core task (defensive) → Secure logistics contract with US USTRANSCOM. T+24 Months Five Eyes Compute Sharing Accord → AUKUS compute layer signing → Evolve from “saving tool” to an official “Defense Infrastructure Component”. T+36 Months Compute sharing under NATO framework → Expansion to major global allies (JP, KR, IL, UAE, KSA) → Annual contract value exceeds $500M.

Evolution of KAI.com’s Market- Making Capabilities KAI.com Strategic Confidential

T+0

AWS/ Azure/ Google公

T+6

T+12

FedRAMP

T+18

T+24

T+36

Phase Market-Maker Capabilities Liquidity Source T+0 Commercial cloud arbitrage (Spot/RI bidding) AWS/ Azure/ Google public markets T+6 Unclassified task bid-ask matching Idle capacity across Five Eyes data centers T+12 CUI hierarchical pricing pools FedRAMP- certified data center alliance T+18 Strategic spare parts futures contracts Long-term lockup + commercial subsidy model T+24 Cross-border allied compute trading Five Eyes + AUKUS shared network pools T+36 Global defense compute liquidity market NATO + Asian Allies

Middle East nodes

The Critical Ice-Breaking Tool: Compute Price Spread Visibility KAI.com Strategic Confidential

Price Index）：

GCP us-central1 A100 80GB $3.47/ hr AWS us-east-1 H100 80GB $4.12/ hr Azure westeurope ND40rs $3.89/ hr 冰岛 Borealis DC (过

A100 / Equivalent $1.20/ hr ←

A100 / Equivalent $0.95/ hr ←

A100 80GB $2.10/ hr 新加坡 Equinix SG1

H100 80GB $4.50/ hr

The first version of the product delivered to DIU is not an orchestration engine, but a highly visible Global Compute Price Index Dashboard: Compute Node / Provider Specification Real- time Price GCP us-central1 A100 80GB $3.47/ hr AWS us-east-1 H100 80GB $4.12/ hr Azure westeurope ND40rs $3.89/ hr Iceland Borealis DC (Surplus) A100 / Equivalent $1.20/ hr ← 68% lower Quebec Hydro DC (Night rate) A100 / Equivalent $0.95/ hr ← 77% lower Tokyo Equinix TY2 (Off-peak) A100 80GB $2.10/ hr Singapore Equinix SG1 (Peak) H100 80GB $4.50/ hr This index serves fundamentally as a market- making apparatus: When the Pentagon visually tracks a 60%+ gap between Iceland and Tokyo, they instinctively query, “Can you route our unclassified training tasks to Iceland?” Demand spawns organically from the client side; KAI.com bypasses standard hard-selling. Compute Price Spread Visibility = Primary Driver of Liquidity Demand. It replicates the Bloomberg Terminal mechanism—Bloomberg forces no trades, but the moment you see a 2% arbitrage spread, you naturally seek out the market maker.

Core Strategy Summary • • • • • • KAI.com Strategic Confidential

The first defense customer is never swayed by abstract “solutions”; they are driven by cold financial metrics. Begin with cost reduction, initiate via unclassified scopes, upgrade from standard PoCs to strict regulatory compliance, scale from a solitary entity to cross-border allied frameworks, and transform cost-saving features into mission-critical infrastructure— each progression builds an ironclad layer of trust, and each layer unlocks exponentially greater compute liquidity pools. This trajectory is explicitly not tailored for impatient Venture Capital timelines. Yet, for an organization truly architecting civilizational digital infrastructure, it stands as the singular, viable closed-loop path through monopoly barriers. KAI.com Strategic Confidential