AI Insights for India’s 2026 Union Budget

As India prepares for the 2026 Union Budget, the dialogue regarding artificial intelligence (AI) is shifting towards practical applications and execution strategies. Arpit Mittal, the CEO of SpeakX.ai, an Edtech startup, communicated via email with Startup Superb, emphasising that instead of merely introducing policies to enhance AI, the government should concentrate on re-evaluating funding methods, offering infrastructure support, and ensuring long-term enablement.

Founded in 2023 by Mittal, the Gurugram-based startup seeks to close the spoken English gap in India by integrating AI with speech technology.

AI Compute Crunch

In October 2025, SpeakX.ai secured $16 million (approximately Rs 142 crore) in financing, led by WestBridge Capital. As a result, the company asserted that access to funding is no longer the main hurdle; rather, the most pressing issue is the compute crunch.

Mittal remarked that startups require access from the outset, rather than through prolonged paperwork procedures. He asserted that upfront compute credits are significantly more beneficial than reimbursement models.

He anticipates that the 2026 Budget will lower import duties on GPUs and AI servers. Mittal expressed that affordable and reliable green power for data centres, alongside expedited approvals, could significantly reduce costs, as AI workloads demand substantial energy.

Additionally, he underscored that the IndiaAI Mission must ensure transparent pricing and guaranteed uptime to assist startups in their growth.

Milestone-Based Funding

Mittal mentioned to Startup Superb that a structured “milestone-based” framework should be established for deep-tech startups. He proposed that rather than offering one-time grants, the government could provide funding linked to specific technical progress.

This approach will enable Indian startups to develop AI models designed to address “real Indian problems.” He pointed to SpeakX.ai as a case in point, emphasising that AI models should effectively operate across “Indian languages, recognising local speech and accents, and facilitating safe tutoring and assessment in education.”

Moreover, he advises that the government ought to adopt India-made AI models early on and incorporate them into public education and skill development programs. This real-world usage would allow startups to test their solutions at scale, fostering trust and credibility, thus nurturing robust, sustainable businesses.

Push for Sovereign AI

During the discussion, Mittal further stressed the importance of data security and stability for startups in India. He revealed that startups face significant delays awaiting GPU access, as they often depend on international providers. These external providers can also hinder long-term growth. Therefore, he highlighted that “accelerating sovereign GPU clusters is essential.”

An India-owned cluster would not only expedite processes but would also ensure stable pricing, local storage, and consistent availability. This is vital for sectors like education and finance, particularly for Tier 2 and Tier 3 cities.

Furthermore, this initiative would “cater to the next 500 million Indians, as startups require extensive, reliable infrastructure within the nation.”

India as a Global AI Hub

Mittal elaborated on the need for the government to implement a “plug-and-play” system in Budget 2026, aimed at making AI computing resources more accessible, dependable, and scalable. He pointed out that when GPU access becomes both affordable and reliable, founders will have the capability to build, train, and deploy models without relocating their teams or intellectual property overseas.

He concluded by stating that this is how India can transition from being merely an AI consumer to emerging as a global leader in AI.