Calisa Acquisition Corp · NASDAQ:ALIS
Calisa Acquisition — SPAC Merger (October 2026)
25 slides from Calisa Acquisition's SPAC merger presentation, Oct 6, 2026
Slide 1: Building Smart Compute Grid for AI Inference Globally
Q GoodVision A! Building Smart Compute Grid for Al Inference Globally
Slide 2: Disclaimer
Disclaimer @ coodvision i ‘This preventation|the Presentation’ Thas been prepared in connection with the proposed business combinatin|the Business Combination'/between Calica Acquistion Corp. ("Cal==")and GoadVicion Aline. (Goodison. This Presentation doesnot purport ta containall ofthe information
Slide 3: GoodVision AI The Complete AI Inference Solution
GoodVision Al The Complete Al Inference Solution "7-layer Al Cake” Built Around Token Flows — and GoodVision Al is differentiated to own the compute that supplies the flow, plus the distribution and optimization. AI DATA CENTERS e1 & Cloud Service From AWS and Google Cloud to Microsoft Azure and oth
Slide 4: Investment Highlights
Investment Highlights e1 @2 @3 04 @5 Inference-First Go-to-Market Faster, Lower-Cost Capacity Deployment Capex-Light Business Model Significant Market Validation Leadership - Proven Track Record @ GoodVision A! Inference-native platform delivering quality Al results at structurally lower cost across
Slide 5: The Team Behind GoodVision AI
The Team Behind GoodVision Al intel. Former Senior Expert Former Senior Executive David Wang / CEO 25+ years building and scaling cloud and compute systems globally. He has brought large-scale infrastructures into production across markets — from launch to significant commercial scale. aws —7 Former
Slide 6: GoodVision AI at a Glance
GoodVision Al at a Glance Q cooavision a\ >960 crus Infrastructure 200+ 40+ Per Container Pro6000 -1MW. =200m? < 1 - 2 & Technology MW reserve Strategic sites >512 crus Footprint / MW PUE efficiency Per Container B300 -1.5MW FY26E FY28E Financial $24M — -20%000 > ( sasomcnsr Einanels $38.9M $418.6M
Slide 7: Competitive Advantage: Full Inference Chain in One Platform
Competitive Advantage Full Inference Chain in One Platform = Q ccoavision «i GOODVISION Al LanBoa FIREWORKS Al ‘OPENROUTER ‘own Compute Capacity v v A Optimize Workioads v v AssetModel v A v v Edge Compute v v Intelligent Scheduling System v v v Valuation Logie v A V Full capability A Partial Not of
Slide 8: Validated by Customers and Partners
Validated by Customers and Partners Measured impact from enterprise Al workloads Reduced by CUSTOMER CATEGORIES BA Gaming Reduced by Ls Education PARTNERS ACROSS THE AI STACK HARDWARE intel. CLOUD & MODELS aws 7 SINvIDIA. ® Google Cloud Note: Impact figures reflect management estimates based 2 B Bio
Slide 9: Enterprise Deployments on GoodVision AI Platform
Enterprise Deployments on GoodVision Al Platform Q soodvision i 62 GAMING BIOTECHNOLOGY Creating an in-game Al agent system Automating customer data onboarding ~$1M annualized platform spend $1.5-2M annualized platform spend CHALLENGE CHALLENGE Anin-game Al agent concept with no infrastructure to ru
Slide 10: AI Wave with Inference as its Growth Engine
Al Wave with Inference as its Growth Engine D cooavision 4) € “The golden age of Al is not training, it's inference. Every application and every edge device requires inference, which will be the biggest Al compute market in the future. ” — Jensen Huang, NVIDIA Source: Bloomberg Intelligence, Generat
Slide 11: AI Is Harder to Use Than It Looks
Al ls Harder to Use Than It Looks THE PROBLEM MODEL FRAGMENTATION + Dozens of models available + Users don't know which one to choose + Nounified orchestration layer SYSTEM COMPLEXITY + Dependencies across cloud, hardware, and models + Hard to optimize cost, latency, and performance simultaneously H
Slide 12: Smart Routing Engine: The Right Model, Every Time
Smart Routing Engine The Right Model, Every Time @ cooavision ai Real-time decision layer, NVIDIA-accelerated. Routes each request to Public Cloud (frontier models) or Al Factory (locally deployed models) based on workload requirements. INPUT User Request Complex: simple - personalized - sensitive b
Slide 13: Cloud Services & AI Native Infrastructure
Cloud Services & Al Native Infrastructure @ coodvision ai Model-agnostic by design — whichever model wins, we monetize the token flow itself. Enterprise Cloud Across Every Major Platform Al Native Infrastructure, Powering Al Agents with Intelligent Token Routing En end cloud consulting, migration, o
Slide 14: AI Factory: Built for AI, Available Faster
Al Factory Built for Al, Available Faster @ cooavision ai Al Factories are purpose-built compute centers designed for fast deployment and high-performance inference. 5 So COOLING SYSTEM DEPLOYMENT COMPUTE ARCHITECTURE ~3D ays toiive >9GCO crus container 912 cpus container <1.2 5 Operational in appro
Slide 15: Convert Crypto Mining to AI Compute
Convert Crypto Mining to Al Compute D cooavision ai Repurposing energized crypto-mining sites into Al inference capacity — live in weeks, not years. CONVERT Live power & grid interconnect Al Compute Asset ~30 days 30-40% Day 1 TIME TO LIVE CAPACITY REPURPOSED GRID ENERGIZED ‘in
Slide 16: Repeatable Model
REPEATABLE MODE Transform from EXPAND AI FACTORY FOOTPRINT + WHERE WE ARE CURRENTLY BUILDING OUT 3 SITES IN BUILD - FIRST BATCH LIVE SEPT-OCT 2026 TEXAS, US AN KOREA Miner conve F onve Factory conversion h power and grid already in place Modular, containerized deployment that replicates fast @ GoodV
Slide 17: Our AI Factory Model: Light Asset, Rapid Scale
Our Al Factory Model: Light asset, Rapid Scale @ coodvision ai Minority capital, full operating control, 30-50% of SPV returns — plus token profit on the same compute. Al Factory SPV Model Synergistic Effect Ry Factory Retrofit ~10% seksi 30-50% 100% . ——<—_—_ ' investment & Full control economic ©
Slide 18: Secular Driver
SECULAR DRIVER Enterprises moving Al from pilot — production Inference (token) volume is increasing — the tide that lifts every lever below TARGET VERTICALS Gaming - video - cross-border e-commerce - education - health - biotech Why: late: e, Al-hungry, hi GEOGRAPHIES U.S., Japan, Korea, EU, and bro
Slide 19: Financial Summary & Projection
Financial Summary & Projection FY2024A FY20250 Revenue 3.6 W 24.0 Cloud & Token Service - - - AlApplication Service - - - Al Factory Computing Service - - - % Growth - 112.7% - % of repeat customers” 100% 100% 100% Gross Profit 0.05 0.16 2.00 % Margin 1.4% 2.1% 8.3% () General & Admin 0.02 0.07 2.68
Slide 20: Revenue Buildout
Revenue Buildout ANNUAL RECURRING REVENUE (ARR) $60M mA\ Factory Computing Service Al Application Service (Smart Routing) Cloud & Token Service $224.0M 1$25.1M $64.0M $5.1M $169.5M $2.4M $107.5M | ssc.sm | FY2026 FY2027 FY2028 @ GoodVision A! SMART ROUTING High-margin section, differentiator from ot
Slide 21: Valuation vs. Comparable Platforms
Valuation vs. Comparable Platforms Entering at a fraction of the comparable set — a ~5x price-to-sales multiple in line with peers, at a fraction of their absolute scale. Valuation $10B $168 $138 $180M ‘$isom~ $180Me Revenue Forecast SIB» $3I5M $50M $37Mem S138Hem $284Mem Price to Sales 10.0« 47.8x
Slide 22: Transaction Summary
Transaction Summary Transaction Structure + Calisa's wholly-owned subsidiary will merge with and into GoodVision Al, with GoodVision Al surviving the merger and becoming a wholly-owned subsidiary of Calise. + Existing GoodVision Al shareholders will receive an aggregate of 18 million Calisa ordinary
Slide 23: Appendix
@ GoodVision A! Appendix
Slide 24: Management and Board of Directors
Management and Board of Directors Shelly Wu / cro + Oil& Gas M&A, Digital Asset Fund Management, Investment & Incubation MSc in Applied Mathematics, Imperial College London @ OO Huang / ADVISOR Ex-Mi(Xiaomi) Ecosystem — Market & Partner Cooperation Led multiple hardware product lines to 10M+ units i
Slide 25: Thank You
Thank You Building the Global Compute Grid for Al Inference © GoopvisroN.az