TCS × OpenAI & AMD: Dual AI Partnerships, Data-Center Scale-ups, and Revenue Upside for Long-Term Investors

AI • DATA CENTERS • REVENUE

TCS × OpenAI & AMD — Dual AI Deals and a Decade-Long Revenue Runway 🚀

100 MW → 1 GW HyperVault build-out, a 200 MW “Helios” blueprint, and Enterprise ChatGPT services could reshape India’s AI-revenue stack for investors.

QUICK TAKE 👓

• ⚡ Capacity: 100 MW pilot scalable to 1 GW (OpenAI JV) • 200 MW Helios blueprint (AMD)

• 💰 Revenue rails: recurring GPU-colocation fees + high-margin Enterprise ChatGPT customisation

• 📈 Upside: JV could exceed US $8 B annual sales by FY-36

1 | TCS + OpenAI — Enterprise ChatGPT 🤝

• 🔑 Roll-out to “several thousand” Tata staff to boost productivity

• 🏢 Domain teams co-build agentic-AI apps for BFSI, retail & manufacturing

• ⚡ HyperVault phase 1: 100 MW, expandable to 1 GW across multiple campuses

• 🌱 Pledge to upskill 1 M Indian youths on generative-AI tools

2 | TCS + AMD — “Helios” Rack-Scale Blueprint 🔩

• 🖥️ Stack: Instinct MI455X GPUs • EPYC “Venice” CPUs • Pensando Vulcano NICs • open ROCm

• 📐 Reference design for AI-ready data centres up to 200 MW for sovereign-AI factories

• 🚀 Open-rack architecture cuts build time for hyperscalers & regulated clouds

• 🌍 Aligns with India’s National AI Mission by localising high-density compute
  
  

  Growth Outlook — Sales Ramp From the TCS × OpenAI + AMD JV

  

  
  

  Key take-aways 

  • Hyper-growth window: First three years drive ~43 % compound growth as the initial 100 MW fills up and higher-margin AI services attach.

  • Scale phase: 300 MW online by FY-31 still delivers ~27 % CAGR through FY-36—even after the “easy wins” are in the base.

  • Mix shift matters: Infrastructure revenue grows faster (51 % → 29 % CAGR) than AI services (29 % → 24 % CAGR) because every new megawatt brings recurring colocation fees.

  • Even if Tata Consultancy Services keeps its core business flat at US $30 B, the JV alone could lift group topline contribution from 2.6 % in FY-28 to 26 % by FY-36.

  Assumptions behind the math

  • Rack-scale AI colocation priced at ~US $7 M per MW per year (liquid-cooled, managed GPU clusters).

  • Utilisation ramps from 70 % (FY-28) → 85 % (steady state).

  • AI-service revenue equals 35-40 % of infra sales (ChatGPT customisation, MLOps, governance).

  • FX held constant; no core-business growth baked in.

  Risks to Monitor

  1. Execution cadence: 100 MW → 1 GW without cost overruns.

  2. Power sourcing: 24×7 renewable PPAs, grid reliability.

  3. Pricing pressure: faster GPU efficiency = lower rack rates post FY-33.

  4. Supply gluts: multiple players chasing the same demand curve.

  Illustrative only — not company guidance. Do your own due diligence before acting.

  
  

  Note: Educational content only. No buy/sell advice. Markets carry risk. Do your own research.