OPINION | Davos Diaries: AI, and the battle for cognitive infrastructure

Davos presented a clearer picture of how the AI era is being shaped by forces outside the lab.

I’m writing this at 30,000 feet—returning from Davos with sore feet (too many conversations on the Promenade), hot chocolates in icy air, too little sleep, and sharply dividing Trumponomics—but also with something more valuable: a clearer view of how the AI era is being shaped by forces outside the lab.

The global conversation has moved beyond “who has the best model” to a more consequential question: who controls the cognitive infrastructure that societies and economies will increasingly depend on—something Ashwini Vaishnaw underscored in a sharp WEF conversation.

From Model Wars to Cognitive Infrastructure

That shift was visible in two anchor moments for India’s first AI-focused delegation of founders and investors at India House. On January 20, we (via the Global Futures Foundation) convened a dialogue on AI for the Global South. On January 23, we held a focused discussion on acceleration with capital.

Between these two moments, Trump’s Davos appearance reinforced the reality of a volatile, multipolar world—where policy, tariffs, energy security and alliances can reprice technology risk in real time.

Trust as the Real Scaling Constraint

As a founder, I’ve learned to distrust hype and focus on what scales. In AI, the scaling constraint is becoming clear: trust—not as a slogan, but as an engineered property. Provenance, auditability, accountability and governance for agentic systems are rapidly becoming table stakes.

India has a rare advantage here. We have built population-scale Digital Public Infrastructure because it was interoperable, low-friction and incentive-aligned. The next leap is what Abhishek Singh (CEO, IndiaAI Mission) calls DPI for AI—Cognitive Digital Goods: open, trust-by-design rails for verified AI authority, consent and accountability, provenance, safety evaluation and governable agents.

AI for the Global South: Deployment Over Power

The January 20 session was a reminder that the Global South’s challenge isn’t powerful AI—it’s deployable AI under real constraints: patchy infrastructure, language diversity, limited fiscal headroom and high sensitivity to misinformation.

Without open standards, the Global South risks becoming a passive market for imported cognition—paying “digital rent” while locking critical services into opaque dependencies.

Capital, Compute and the Missing Moat

On January 23, the capital conversation made the complement clear. Sophisticated investors are increasingly separating compute-only narratives from capability narratives. Compute is necessary, but not sufficient; the moat is the stack—governance, evaluation, deployment and trusted rails.

India will need $100–150 billion over 10–15 years. To mobilise this, we should create private, investable structures—aligned with IndiaAI’s Fund of Funds and the recently unveiled $12 billion RDIF—potentially housed at GIFT to attract sovereign and pension capital. 

(Umakant Soni, Chairman of AI Foundry; Cofounder of ARTPARK.)

Views are personal and do not represent the stand of this publication.