Union Budget 2024-25: AI Startups Seek Infra Boost, Govt Collaboration Opportunities

Until two years ago, AI was not a prominent topic of discussion at the government level. However, with increasing conversations about generative AI and machine learning, it is now gaining prominence alongside sectors like semiconductor manufacturing, spacetech, and other emerging industries. As Finance Minister Nirmala Sitharaman prepares to present the full Union Budget 2024 on Tuesday (July 23), industry stakeholders expect more clarity from the government.

A month after the last interim budget, the cabinet announced the India AI Mission, a fund of funds. Developments since then have indicated that a large share of this allocation will focus on infrastructure such as GPUs.

For context, in March, the Union Cabinet approved an outlay of INR 10,372 Cr for five years for the India AI Mission, aiming to foster innovation in the sector through public-private partnerships (PPPs). The Centre has earmarked INR 4,500 Cr for AI computing infrastructure, equipped with GPUs.

“Now, of course, that was a high-level announcement, but something more concrete is expected. I don’t think it will come in the budget speech itself, but at least in the budget details. When they come out with the fine print, we need to see what this means for the segment and when it will be executed,”Ashwin Raguraman, cofounder of Bharat Innovation Fund, said.

Industry stakeholders are expecting a follow-up conversation or an update on the implementation and the next steps, he added. ”Is there anything else they are planning, such as the subsidisation of hardware infrastructure for startups? These are all areas that could be discussed further,” he added.

This sentiment was echoed across the board when Inc42 spoke to AI startup founders and investors. As infrastructure remains the most challenging hurdle for the AI industry in India, all of them are seeking more concrete announcements and policies in this regard.

Cutting Costs And Tackling Infrastructure Hurdles

The biggest challenge in a startup is the substantial upfront cost required for training and fine-tuning machine learning models. These models need significant computation power—CPU, GPU, and other resources.

If the government could offer support, such as subsidising these costs for entrepreneurs or founders for the first three years, it would make a huge difference, Aman Goel, cofounder and CEO of GreyLabs AI, told us.

For example, if import duties on GPUs or related hardware were waived or reduced, or if GST components were lowered, it would significantly decrease the cost of entry. This would encourage more entrepreneurs to enter the space, speed up innovation, and make customers more willing to experiment with new technologies, the GreyLabs CEO added.

“The government is already implementing some public infrastructure initiatives, like AI models for India. If they could offer these resources at an affordable rate, it would be incredibly beneficial for startups like ours,” he added.

India’s ambitious plan to build 10,000 GPUs might seem impressive, but in the context of the global AI race, it’s not a large figure, a policymaker said. When OpenAI developed GPT-4: it took approximately 25,000 GPUs running for three months to train a model with a trillion parameters.

“In a country where startups often struggle with limited access to capital, relying on just 10,000 GPUs may fall short of fostering a thriving indigenous AI ecosystem,” he said. However, he also mentioned that even 10,000 GPUs would be a big step to begin with, especially when very few startups in India are building foundational models.

Beyond Financial Support: Acceleration And Collaboration

Besides looking at a boost for the AI infrastructure, the AI startups also want to engage with the government at a bigger level, and increase collaboration with government agencies.

“AI startups also need customers, especially in the B2B sector. However, reaching out to large companies, government can be challenging. While organisations like NASSCOM assist with this, more collaboration opportunities and support mechanisms could further enhance the connection between startups and large enterprises,” Raga AI’s Agarwal said.

In addition, industry stakeholders believe that increasing the number of incubators could significantly benefit AI startups. To explain this need, IvyCap’s Guta said that there are approximately 30,000 accelerators globally, while India has only about 800 to 815, which accounts around 2.7% of the global total. This is despite India being one of the largest startup ecosystems in the world, with over 1.4 Lakh startups.

Bharat Innovation’s Raguraman also acknowledged the need for more accelerators but added that support must extend beyond the pre-seed stage. While incubators are vital for nurturing seed-stage and pre-seed startups, the government should also focus on policies that advance startups beyond these initial phases.

“It’s important to not only create a robust pipeline of early-stage startups but also to provide substantial support for those that are more mature or on the path to maturity. This could involve creating policies and programs that work in conjunction with incubators, investment funds, and infrastructure providers such as data centers,’ he said.

The government should use existing incubators and create strategies to help startups move from early stages to growth levels. While private venture capital is important, the government also needs to play a strategic role in supporting startups throughout their development, according to him.

Should AI Be Up For Tax Benefits?

“One area where support would be immensely beneficial is tax benefits. Early stage startups often face significant challenges in generating revenue and raising capital in the initial years. By offering tax incentives to AI startups, the government could alleviate some of this financial pressure,” Gaurav Agarwal, founder of Raga AI, said.

On the other hand, Vikram Gupta, founder and managing partner of IvyCap Ventures, thinks both the government and other stakeholders should focus on developing accelerators, angel networks, and seed-stage capital to support early-stage startups. This support can only be effective if accompanied by tax incentives, he added.

For example, in the UK, individual investors can deduct their investments in startups from their taxable income, recognising these investments as contributions to economic growth and job creation. However, India currently lacks such incentives, according to Gupta.

“Additionally, there should be parity in the treatment of long-term capital gains between public and private markets. In India, private investments are treated as a high-risk asset class with long-term capital gains taxed at 40% after two years, compared to just 10% for public markets after one year. This discrepancy discourages investors from putting their money into startups, as they prefer the lower risk and better tax treatment of public markets,” Gupta said.

To attract more investors, there is a need to offer extra incentives for investing in this asset class and remove obstacles such as unclear angel tax regulations, which have been a concern for over five years. Making these changes would make it easier and more appealing for individuals to bet on startups, which will eventually help emerging sectors like AI.

Most importantly, all of this requires a stable AI policy. Industry stakeholders emphasised the urgent need for clear and consistent AI regulations so that startups can align with them and focus on developing new products. However, it is crucial that the government ensures these policies do not stifle innovation, as has occurred in other emerging areas such as crypto in the past.