|
Listen to this story
|
It’s now widely recognised that selling AI models is a zero-margin game. The next wave of AI startups must capitalise on LLMs in the application layer to tackle real-world challenges.
“The next billion dollar startups in AI will play on the application layer and not the infrastructure layer,” said AIM Media House chief Bhasker Gupta in a LinkedIn post.
Gupta added that there is a plethora of problems to be solved using AI, and these startups will localise their solutions while maintaining a broad-based approach.
Echoing a similar market sentiment was Nayan Goswami, the founder and CEO of Chop. “The next major wave of AI innovation will focus on the application layer, where startups will build specialised vertical AI software-as-a-service (SaaS) companies for global markets,” he said.
Goswami further elaborated that with robust foundational models like Anthropic, Cohere, and OpenAI, along with infrastructure companies like LangChain and Hugging Face advancing rapidly, we’re poised to witness a surge in application-layer startups targeting specific verticals.
“Think of foundational models as the roadways, and application layers as the vehicles driving on them,” he explained.
Finding the Right Application to Build is Key
Andrew Ng, the founder of DeepLearning.AI believes AI’s usefulness in a wide variety of applications creates many opportunities for entrepreneurship. However, he advised budding entrepreneurs to be extremely specific about their ideas for integrating AI.
For instance, he explained that building AI for livestock is vague, but if you propose using facial recognition to identify individual cows and monitor their movement on a farm, it’s specific enough.
A skilled engineer can then quickly decide on the right tools, such as which algorithm to use first or what camera resolution to pick.
In a recent interview, Ng explained that the cost of developing a foundation model could be $100 million or more. However, the applications layer, which receives less media coverage, is likely to be even more valuable in terms of revenue generation than the foundation layer.
He also said that unlike foundation models, the ROI on the application layer is higher. “For the application layer, it’s very clear. I think it’s totally worth it, partly because it’s so capital efficient—it doesn’t cost much to build valuable applications. And I’m seeing revenues pick up. So at the application layer, I’m not worried,” he said.
Perplexity AI serves as a strong example by integrating search with LLMs. Rather than building its own foundational models, the startup leverages state-of-the-art models from across the industry, focusing on delivering optimal performance. The company is planning to run ads as well from the next quarter onwards.
However, not everyone is going to make the cut; some startups are going to fail. Statistically speaking, around 90% of startups don’t survive long enough to see the light at the end of the tunnel.
Ashish Kacholia, the founder and managing director of Lucky Investment Managers, said, “AI is the future but key is how the applications shape up to capitalise on the technology.”
India is the Use Case Capital of AI
“India is going to be a use case capital of AI. We’ll be very big users of AI, and we believe that AI can significantly help in the expansion of the ONDC Network,” said Manoj Gupta, the founder of Plotch.ai, in an exclusive interview with AIM.
Similar thoughts were shared by Nandan Nilekani when he said that India is not in the arms race to build LLMs, and should instead focus on building use cases of AI to reach every citizen. He added that “Adbhut” India will be the AI use case capital of the world.
“The Indian path in AI is different. We are not in the arms race to build the next LLM, let people with capital, let people who want to pedal chips do all that stuff… We are here to make a difference and our aim is to put this technology in the hands of people,” said Nilekani.
Krutrim founder Bhavish Aggarwal believes that India can build its own AI applications. Agreeing with him, former Tech Mahindra chief CP Gurnani said, “It’s time to stop ‘adopting’ and ‘adapting’ to AI applications created for the Western world.”
Gurnani said that the time is ripe for us to build AI models and apps based on Indian data, for Indian use cases, and store them on India-built hardware, software and cloud systems. “That will make us true leaders in the business of tech,” he added.
Notably, Gurnani recently launched his own AI startup AIonOS.
Startups Offering More Than LLMs
Lately, several AI startups in India have been building services using generative AI. For example, Unscript, a Bengaluru-based AI startup, is helping enterprises create videos with generative AI. Another video generation startup, InVideo, is estimated to generate $30 million in revenue in 2024.
Recently, Sarvam AI launched Sarvam Agents. While the startup, backed by Lightspeed, Peak XV, and Khosla Ventures, is not the only company building AI agents, it stands out for its pricing. The cost of these agents starts at just one rupee per minute.
According to co-founder Vivek Raghavan, enterprises can integrate these agents into their workflow without much hassle.
These agents can be integrated into contact centres and various applications across multiple industries, including insurance, food and grocery delivery, e-commerce, ride-hailing services, and even banking and payment apps.
Similarly, Krutrim AI is making AI shopping co-pilot for ONDC. Khosla Ventures-backed upliance.ai is building kitchen appliances integrating generative AI.
Meanwhile, Ema, an enterprise AI company founded by Meesho board member Surojit Chatterjee, recently raised an additional $36 million in Series A funding.
The company is building a universal AI agent adaptable to a wide array of industries, including healthcare, retail, travel, hospitality, finance, manufacturing, e-commerce, and technology.
Enterprises use Ema for customer support, legal, sales, compliance, HR, and IT functions.
Lately, we have observed that Y Combinator is bullish on Indian AI startups, many of which are focused on building AI applications.
For example, the creator of AI software engineer Devika, Mufeed VH, who founded Stition.AI, is now part of YC S24 batch. His startup works around AI cybersecurity for fixing security vulnerabilities in codebases, and is now renamed to Asterisk.
On the agentic front, Indian co-founders Sudipta Biswas and Sarthak Shrivastava are building AI employees through their startup FloWorks.
Examples are aplenty, with India poised to boast 100 AI unicorns in the next decade.
In a conversation with AIM, Prayank Swaroop, partner at Accel India, said that the 27 AI startups his firm has invested in over the past few years are expected to be worth at least ‘five to ten billion dollars’ in the future, including those focused on wrapper-based technologies.
There are a host of categories, such as education, healthcare, manufacturing, entertainment, and finance, to explore with generative AI, and this is just the beginning.
