Much of Google’s success or failure in generative artificial intelligence (AI) will hinge on how its India market fares in the immediate future. This was a common refrain among developers, consultants and analysts alike, even as Google CEO Sundar Pichai spent two hours on Tuesday convincing the world that the search giant was not passé in the race for generative GenAI supremacy.
Why? Well, for one, Google is using its dominant market share in mobile operating systems (Android), in browsers (Chrome), and in email agents (Gmail) to place AI applications based on its Gemini large language model (LLM) in front of all its customers. On Android, it’s doing this with live transcriptions, summaries, photo and video editing, and more; on Chrome and Gmail, it is doing this by placing Gemini as directly accessible pop-up sidebars—closely integrated into each user experience.
Both Gmail and YouTube have over 1.8 billion users worldwide, while Android has 4 billion users. In India, nearly 97% of its 800 million active smartphones run on Android. While Gmail numbers are not absolute, market data suggests 83% of India's email usage occurs through Gmail. India is also YouTube’s biggest geography with over 460 million monthly active users. All these make India a strategically crucial geography for Google to market its AI products and, in turn, take the fight to its key industry rivals.
Read more: How Fractal, India’s first AI unicorn, is prepping for the next race
The second reason for Google’s India focus would lie on the latter’s developer base. On GitHub, the world’s largest developer community, India had over 13 million developers as of 31 December—second only to the US, which had close to 20 million. Naturally, India’s developers, be it of Indian origin or those based in India, would play a key role in adopting foundational AI models and building tools that keep Big Tech at the centre of both consumer and enterprise tech industries.
To be sure, a foundational AI model is the basis upon which AI sub-models and applications are built. Foundational models include OpenAI’s GPT (generative pre-trained transformer), Meta’s Llama and Google’s Gemini, among others. Developer and application support is crucial for these models to be successfully deployed, especially given the steep cost of compute and cloud infrastructure needed to train and run them.
A key example is how ChatGPT, built initially on GPT-3, created mainstream recognition for OpenAI and brought generative AI to the masses—in turn sparking enterprise interest. In India, demand for generative AI is gradually building, with K. Krithivasan, chief executive of India’s largest IT services firm, Tata Consultancy Services (TCS), stating on 12 April that the company had received $900 million in generative AI contracts through FY24. For Google, this opportunity could be key.
Read more: Airtel ties up with Google Cloud to offer cloud and gen-AI solutions to Indian businesses
Yet, industry stakeholders feel Tuesday’s announcements may not be enough. “The perception that OpenAI’s and Meta’s models are superior to Google’s, among developers, is accurate. While Meta should have faced the same challenge as Google to take on OpenAI, its decision to take its LLM open-source has helped boost developer confidence. There were no large-scale open-source LLMs from Big Tech—Meta cashed in on this," said Aakrit Vaish, co-founder and chief executive of conversational AI startup, Jio Haptik.
Vaish added that Haptik will continue to use OpenAI’s latest models as its foundational tech owing to GPT-4 (now GPT-4o) being the “most flexible" in terms of fine-tuning its output with custom data. He also noted that Google did not offer any “compelling" feature for developers already working with OpenAI to switch. “With GPT-4o being natively multimodal, there’s nothing that Gemini does that stands out entirely," he said.
Others agree. Jayanth Kolla, co-founder and partner of technology consultant Convergence Catalyst, grouped LLMs such as Claude and Mistral, built by US-based Anthropic and France-headquartered Mistral AI, with Gemini—even after the announcements. “It’s not about how good or bad the LLMs are in terms of response times, accuracy and hallucinations. When it comes to working with developers, the quality of API integration is what makes a difference. Today, OpenAI’s GPT models are the best—making it the LLM of choice for most developers," Kolla said.
However, as stated above, Google does have numbers on its side. “This therefore becomes a question of quality versus distribution—in the past, we’ve always seen distribution win. This is why it is crucial for Google to place Gemini in every platform it dominates today, to make sure that as a company that innovates, Google continues to retain its relevance," said Kashyap Kompella, analyst and founder of tech advisory firm, RPA2AI Research.
Read more: India Inc, not just IT, dangles big bucks for AI specialists. But is it all hype?
Kolla concurred, highlighting Microsoft’s resilient Office suite as a key example of how distribution helps tech firms. “If you look at Microsoft Office, it was already dominant when Google entered the market with G Suite. Today, Microsoft’s office services remain dominant by enterprise revenue, even after years of free apps distribution and continued enterprise push for G Suite from Google. The same can happen with AI as well, but developer interest will be hard to sway until this dominance is established. India, with its volume of both India-born developers globally, and India-based developers, will likely play a significant role due to the sheer volume," he added.
“OpenAI already has a head-start on the competition, and also beat Google to multimodality with its new iteration. While the race is still Google's to lose, the key here would be to rope in developer interest, which is currently swayed to OpenAI's side," Kompella added.
Google is making the effort, too. For most of its latest AI solutions, Indian developers are in the first rung of waitlists and beta rollouts. However, a Google demo confirmed that while its latest AI model, Gemini 1.5 Pro, will be available to Indian subscribers of Google One AI Premium storage and services plan, only US customers get the full list of features that its latest model is capable of.
“Upload file to analyse as a feature isn’t available for India yet—we’re working to add that in the coming weeks," a Google executive said, requesting anonymity.
The author is in Mountain View, California to cover Google I/O 2024, on Google's invitation.