As AI adoption accelerates across industries, companies and investors are assessing how infrastructure, data strategy and regulatory expectations should inform their business and investing decisions.
At the recent Morgan Stanley Sustainability Leadership Summit, Sridhar Ramaswamy, the CEO of Snowflake—which provides data and AI solutions—offered his perspectives on what enterprises and governments are prioritizing in their AI roadmaps.
In a fireside chat with Jeff McMillian, Morgan Stanley’s Head of Firmwide Artificial Intelligence, Ramaswamy spoke about how Snowflake and its collaborative partnerships with other big tech companies are aiming to meet AI demands from enterprises for data interoperability and from governments for sovereign cloud infrastructure, and how model efficiency may help reduce energy consumption.
AI Value in Unified Data for Enterprises
For many enterprises, one of the biggest promises of AI is its potential to break down silos between structured data—such as transactions and customer records—and unstructured data—such as documents, audio and video, according to Ramaswamy.
Advances in large language models and machine learning have made it possible to extract insights from unstructured data at scale. When enterprises can connect this with structured data, they may gain a more complete understanding of their business priorities, improve operational efficiency and deepen their competitive advantage. For investors, it signals a shift in how data assets can be monetized.
“One lens that all of you should have in your thinking about AI models is as a bridge between things like unstructured data and structured data,” Ramaswamy said. “That’s a fancy way of saying, AI is really good at figuring out from your spoken words what concepts, what numbers you’re looking for. It can also do that from documents. It’s these kinds of applications that end up creating the most value in enterprises.”
Another enterprise need that AI can help fulfill is interoperability—software platforms’ ability to communicate and share data and information. To facilitate this integration, Snowflake embraces partnerships with other companies, such as cloud providers and large software platforms, Ramaswamy said.
Rising Demand for Sovereign Clouds
As geopolitical tensions rise and the global order becomes more multipolar, data sovereignty has moved to the forefront of technology strategy for governments.
“We’re headed into a fracturing world in terms of how different areas of the world are thinking about things like software and data and where it should sit,” Ramaswamy said. “More and more, nations are demanding that when it comes to the personal information of their citizens especially, that it needs to be stored in buildings and run by people that are under their jurisdiction.”
Snowflake is approaching the opportunity in building sovereign clouds through partnerships with cloud providers, software platforms and other tech companies such as Google, OpenAI, Anthropic, Salesforce and ServiceNow, Ramaswamy said. “Many of the sovereign clouds are being built by the hyperscalers. We don’t own the data centers, we don’t own the software that are run on the machines that are in data centers, so we run on top of the hyperscalers.”
In the next decade and beyond, local on-the-ground companies will crop up to capitalize on opportunities to gain market share and partner with cloud providers and software platforms: “Local companies see this as opportunity by using this push from many governments for there to be local sovereignty over data, software and the controlling infrastructure that goes with it—it will play out over the next 10-20 years.”
More Efficient Models, Lower Energy Demands
Training and running large-scale models, particularly those used in generative AI, has historically required massive computational resources and energy consumption. However, the trajectory could be shifting, Ramaswamy said. Advances in model architecture, training and compute optimization could achieve similar or better performance outcomes while using fewer resources and power.
“You don’t need the biggest models to do a lot of things,” Ramaswamy said. “This is what engineers love doing: Let’s make this 100 times more efficient. There’s hope in terms of us not just building bigger and bigger data centers that consume all the world’s energy.”
For enterprises and governments alike, compute-efficient AI at scale could be more cost-effective and environmentally viable—critical for resilience and long-term value creation.
