Mohsen Moazami, an AI infrastructure investor, argues India has become an emerged market critical for U.S. tech growth. Despite massive GPU availability, utilization sits at only 40-50% because inference costs are too high and enterprises are cutting budgets. Real AI value comes from data, not just hardware, and governments need to treat AI as national infrastructure.
Chapters
The level of investment in AI infrastructure is unprecedented. I've never, never in my 37-year career experienced anything of this magnitude. The appetite for AI infrastructure and compute and everything that goes with it is unending.
What you're seeing is real and what you're seeing is unprecedented.
Well, you know, I'm right now in Menlo Park. These are all venture capital firms and private equity firms. So I see the level of investment. I don't see any slowdown in the investment. And therefore, you know, of course, everything will ultimately plateau or slow down. But over the long run, I'm extremely optimistic. This is going to— it's a transformational movement that's going to impact every layer and every aspect of our life.
And while the gold rush is all seeming right now for the models, I think it's important for us to understand what's happening with the infrastructure piece.
There's infinite appetite for compute, but ultimately what generates profits and revenue is what people do with their data. Correct. So DDN is the data engine for the AI economy.
Do you see a clear opportunity for players like DDN in India?
We, companies like DDN and/or any American technology company, regards India as a critical growth market. It's no longer an emerging market, it is an emerged market. So there's this sort of, I don't know, common denominator, natural bonding between Silicon Valley and India. But nevertheless, a country of your size and size of population, the emerging businesses and global conglomerates We cannot afford to ignore India.
Is meeting demand itself a challenge right now?
And the markets for memory is very squeezed. There are a very handful of providers in the world, predominantly in Asia and in South Korea, and their supplies are asked, you know, called for. But the challenge is this, you know, right now with all the focus on the computer deployment, there's more and more GPUs available in the world. But the problem is They remain idle for the large portion of the time. You know, on average, these GPUs that are in production today are utilized at a rate of 40 to 50%. So 50% of the GPU capacity is idle.
The enterprise adoption is not playing out as aggressively as what one would have imagined. And the argument for that is that the return on investment is not feeling worth its while.
Profitability matters. And right now we're beginning to hear noises in the market that the cost of inferencing is so high that people are beginning to notice and trying to slow down and/or control the budgets because otherwise expenditure goes on through the roof. So making that infrastructure more productive and profitable is the key to the game. Now, as the data governance, data provenance, data management, data sovereignty issues are more and more understood and addressed, absolutely you will see an exponential increase in enterprise adoption.
What's your advice going to be to governments around the world right now that are grappling with some of these issues?
Well, in one word, sovereignty. We have historically spent money on national infrastructure such as railways and roads. AI is the new infrastructure, and nations have to prioritize because it's It's a key element of national sovereignty and competitiveness. So if you want to compete in this world, we cannot do it without having your own AI.
