Towards DPGI 2.0: A case for deep-tech public goods

'If data is the new oil, large scale high performance computing is the new oilfield'

As Prime Minister Nare­n­dra Modi is set to host the global community of leaders from within the G20 nations in New Delhi, a significant area of focus has been the discussion around digital public goods and related infrastr­ucture (DPGI). India’s success in scaling digital public goods at the population level to serve hundreds of millions of citi­zens through billions of online transactions has in recent times attracted the attention of global leaders from Japan to Germany.

Elaborating on this mission of digital inclusion in my recent book, Collective Spi­rit, Concrete Action, as elucidated by Prime Minister Modi during his Mann Ki Baat, it was highlighted how social and ec­o­nomic barriers have been bre­ached to empower the lowest strata of society through digital public goods. From Aadhaar to UPI, government e-marketplace (GeM) to the account aggregator framework, India’s digital public goods are driving financial inclusion, expanding the credit base, democratising e-commerce marketplaces and formalising large swathes of the economy through digital transactions. With several nat­ions looking to adopt the India Stack approach to create digital public goods, the DPGI model is emerging as an alternative to the big-tech platform model that has largely dominated the global digital economy over the past two decades.
 

There is, however, a significant gap between the two mod­els, which is widening at an ac­celerated pace. While DPGIs are based on open APIs and rely on a decentralised network of apps and services to participate in concert with a centrali­sed public platform, the underlying technology building blocks are anything but open or public. From cloud computing to the silicon that powers the data centres, the underlying foundation for DPGI continues to be dependent on deep-tech capabilities from the big-tech majors.

In recent tim­es, this foundation has seen ex­p­onential advancements in complexity and scale with the advances in artificial intellig­ence (AI). Today, not only is infrast­ructure for AI models based on sophisticated and custom-designed, high-perfor­mance computing chips, but AI algori­t­hms are also being used to learn and develop even more sophisticated chips. This accelerated loop of innovation is driving advancements that are widening the gap between the big-tech majors and startups and public bodies. While whether this gap can be bridged is an open question, there is a case for developing a new class of digital public goods focused on key deep-tech capabilities — DPGI 2.0.
 

Two illustrative examples can help understand better the case for deep-tech public goods envisioned to encom­p­ass DPGI 2.0. The first example has to do with high performa­nce computi­ng. Writing in the SemiAnalysis recently, Dylan Patel and Daniel Nishball have highlighted how access to high-performance computing cap­acity is highly re­s­tricted to a few large play­ers. While a handful of large tech platforms such as Google, Meta, X and OpenAI have amassed massive amounts of computing power necessary to drive AI research and development, there are a large number of startups and individual researchers that have limited or no access to such large-scale, high-performance computing infrastructure.

In their assessment, these smaller efforts will remain uncompetitive owing to their inability to access large-scale computing, which requires huge investments to the tune of billions of dollars. Giving the example of Nvidia’s DGX cloud service, they highlight how large-scale, high-performance computing is now the new equivalent of an Ama­zon Web Service for AI models and application services. If data is the new oil, as is often argued in the context of digital public goods, large scale high performance computing is the new oilfield, and those who control the new oilfields will end up exercising disproportionate influence on how the new oil economy conducts itself. Hence. the case for large-scale, high-performance computing as a deep-tech public good that is available to start-ups and independent innovators to level the playing field with big-tech platforms and the Silicon Valley majors becomes essential as we envision DPGI 2.0.
 

A natural extension of the above example is the need to open up AI models and datasets to empower startups and researchers to innovate on AI on par with the big-tech platforms. There is a real risk that increasing the sophistication of AI models of the big-tech platforms results in a scenario where digital public goods end up depending on these models to a large extent, resulting in further skewing the digital ec­onomy. While it may be argued today that the India Stack mo­del of DPGI is an alternative to the Big-Tech Plat­form approach, in a future dominated by AI and high-performance computing, the underlying deep-tech for the stack would be heavily dependent on those very same platforms and tech majors.
 

India needs to spur and catalyse a deep-tech ecosystem all the way from fabless chip designs to semiconductor fou­ndries to large-scale, high-performance computing cloud ser­vices if the level playing field created by the current gen­eration of digital public goods is to be sustained furt­her. Uniquely, Indian innovations such as direct-to-mobile broadcasting, D2M, offer the opportunity to spur such an ecosystem, creating a new generation of digital public goods rooted in deep-tech developed indigenously. From D2M to large language models (LLMs) and a public high performance cloud service, it is important to envision the DPGI 2.0 to sustain the digital public goods momentum while securing India’s technology leadership in this area.


The writer is former CEO, Prasar Bharti