Lessons from Mumbai Tech Week 2025
Scrappy notes on the future of the world, from India
Welcome back to Consortium.
I recently attended Mumbai Tech Week 2025. For the uninitiated, MTW is a collective of tech founders, startup enthusiasts, investors, business professionals, and students coming together to listen and learn from the industry’s best and brightest on the future of technology and its impact on society, the economy, and the world.
This year’s event was centred around the impact of artificial intelligence (more on this from Consortium soon!) in various industries ranging from gaming and entertainment to telecom. Outside of this, key debates that are currently at the centre of the AI revolution were discussed by both technologists who are leading the change as well investors who are fuelling it. To drop a few brands here, I managed to listen to Pratyush Kumar from Sarvam AI, Hemant Mohapatra from Lightspeed Venture Partners, Harshjit Sethi from Peak XV, and Sanjay Nath from Blume Ventures.
Here are some of the notes I picked up from the event on where things are headed.
Happy reading!
On India’s AI Investment Landscape 🇮🇳
A significant investment gap exists between India and other countries. For example:
The U.S has already invested approximately $300B in artificial intelligence
Apple alone has recently announced a staggering $500B investment in AI technology
China has established an $8.2B government investment fund dedicated to AI
Alibaba has earmarked $53 billion specifically for AI initiatives
India is at a meek $1.42B in total AI investments. Why is this the case?
On Research Foundation & Talent Development 🔬
A significant gap in investment is driven by the large difference in India’s AI ecosystem talent compared to the U.S:
The United States breeds a culture of innovation with its universities sprouting research centres where talent emerges from
Major U.S. companies maintain several research groups that drive innovation
India lacks a similar research culture across educational institutions and corporations
Leading tech companies in India have historically prioritised the return of profits to shareholders rather than reinvesting capital into fundamental research
As a side bar to the above point, from the Ken’s article on India’s future in the context of China and AI:
The thing that prevents India from building a Deepseek is how research is viewed, and consequently practised here. Socially and professionally, research projects are neither prioritised, funded, nor lauded. Deepseek’s success is considerable, but it’s probably one of the dozens and dozens of AI research projects that are going on in China. India has chosen to go into use-cases because researchers constantly face the question—“but what’s the point of doing this? What’s the application?”—all the time. So it’s much easier, and tempting professionally, to choose to build something that has some outcome and application than to spend time building something that may not go anywhere.
This research gap fundamentally impacts India’s ability to build bleeding-edge foundational AI models
Despite these challenges, India has a lot of opportunity as visits from major AI leaders to the country and connections between Indian talent and global AI development communities are increasing. The emphasis has to be on stronger linkages between Indian universities and industry to foster AI innovation.
On Foundational Models vs. Applications 🤖
A central debate we keep hearing is whether India should focus on developing its own foundational models or concentrate on just the application layer. Here’s what the current thinking is like:
Majority of Indian companies would like to focus on the application side of things — very much in line with what we see with regard to the attitude towards the economics of doing solid research in India
Although some foundational model work might be done in India, it would form a minority of the AI activity in the country
For data sensitive applications surrounding work with the government, defense, banking, healthcare, the need for sovereign Indian foundational models is real
The market is large enough to accommodate various models and applications can be built across the stack
However, regardless of whether India builds its own models or not, applications will forever be model-agnostic. This means that whichever foundational model that met the use case the best would be used irrespective of whether the model was foreign or Indian. An exception to this would be if applications surrounding national security and other data sensitive sectors were being built. This use case would require model sovereignty.
On Data Privacy, Provenance, and Regulation 🔒
There is a complex balance between rapid AI development and data privacy concerns:
Data provenance1 is essentially the kryptonite to current AI development given how data for AI models are currently scrappy, cheap, messy and untraceable for the most part. Provenance would kill current models if large companies were required to practice this as a standard
Indian regulations require certain data to be stored in data centres that are present on Indian soil
There are two approaches to solving the data privacy challenge:
Technological solutions (edge computing, on-device processes that doesn’t require sending data to the cloud)
Regulatory frameworks that dictate what data foundation models can train on
However, rather than calling for specific regulations, there is a need for single, stable frameworks that aren’t dynamic across states or leadership transitions
On The 10-Year Investment Thesis 🌄
Identifying a single AI investment thesis to commit to for the next decade is difficult but the following are some ideas that investors are bullish on:
Services Enhanced by AI
India has a rich history of being a service based industry (over 50% of the GDP comes from here) and thus transforming these services gives India’s right to win. This would require:
Taking existing knowledge about how service delivery works and reimagining these services with AI capabilities
Creating new service delivery models that are faster, cheaper, and more effective
Combining human expertise with AI to create full-stack solutions for customers
AI as a New Platform Shift
When major platform shifts (such as the shift to the internet or the shift to mobile) take place, the largest outcomes and biggest companies emerge
New opportunities are thus created which leads to the creation of new companies
Mobile enabled the rise of Uber, Instagram and Airbnb. AI will definitely lead to entirely new business models
Global consumer ventures that originate from India have not been common however AI has the capability to facilitate a new wave
Expanding Market Size
Initial SaaS companies sold workflow optimisation, now AI can help these companies sell work itself
This combines software spend and labour spend into a larger TAM
Companies that adapt to this shift could become significantly larger than previous SaaS leaders
On The Future of SaaS in an AI World 🌐
How will software companies fare in an AI-dominated future?:
Three Tiers of Vulnerability
Process/storage products can be easily replaced by AI capabilities
Middle-tier SaaS companies must find a deeper differentiation to survive
Infrastructure companies like Snowflake might be more insulated from disruption
Adaptation vs. Disruption
Multiple SaaS companies are rapidly building AI capabilities into their products rather than resisting the technological transition
“Wrapper” businesses or those that merely add a layer of AI to their existing products, may not thrive for long as opposed to those that are fundamentally driven on AI technology
Thus, the opportunity lies in fundamentally re-architecting products around AI capabilities
Self-Service and Visualisation
Tasks previously outsourced to SaaS vendors might be brought in-house using AI tools leading to more integrated business models
For every $1 spent on SaaS, approximately 1.8x is spent on implementing the SaaS solution
Multiple SaaS tools exist in the market and due to this, several are adopted but only 60-70% of them are utilised despite the significant investment in them
Artificial intelligence could reduce both implementation costs and friction in software development
On India’s Brain Drain Challenge 🧠
India has historically faced a challenge of talent migration to Silicon Valley and other tech hubs, here’s some of the reasons why:
Economic Realities
The U.S. economy is 8-10x larger than India’s which naturally leads to attracting founders seeking the largest possible markets for their products and services
Foreign markets lead to higher compensation which acts as a natural pull factor for top AI talent
As the largest possible markets are in countries like the U.S, many founders prefer to be closer to their customers and to the use cases their products can serve
Models for Talent Retention
“TCS/Infosys model” where HQ is in India but significant portion of the business comes from abroad
“Freshworks model” where the founder and deal team are based abroad but large development centres are in India
“Globally distributed model” where companies take a hybrid route and are not strictly “Indian” or “American” in nature
Creating a Stronger Ecosystem
Corporate partners that offer better compensation to compete with international opportunities
Government initiatives that support AI research and development
Robust partnerships between universities, corporations, and startups
Recognition that talent flow should be bidirectional for learning and growth
Creating an environment where founders can be globally competitive while based in the home market
On The Spectrum of AI Investments 💸
Left Side: Infrastructure Layer
Computational centres, power infrastructure, chip companies, and foundational model developers
Requires significant capital investment but offers a clear win-lose situation
Getting increasingly commoditised despite the higher barriers to entry
Some venture funds (eg. Lightspeed) have investments in power companies across the U.S and Europe, chip companies globally, and foundational model developers
Middle: Application Layer
Middleware companies, horizontal applications, and vertical industry-specific solutions
Represents the most significant portion of venture activity in the AI space, Lightspeed alone has made 75 AI investments in the past year
Crowded space and thus very high uncertainty on who will achieve breakout success
Requires a careful evaluation of team quality and technological differentiation
Right Side: Services Layer
Dominated by large consulting firms eg. PwC has generated ~$1.5B from AI consulting alone
High revenue potential but not venture investable
Difficult for startups to compete with large global consulting firms such as Accenture
On What Investors Look For in AI Founders 🔍
Venture capitalists when investing in new companies are always looking at the strength of the team, here is what they look for when evaluating new AI companies:
Agency and Initiative
Founders who demonstrate “agency” are sought after — those that take initiative with modern tools — Cursor, Replit, Loveable, Claude Code
In line with this, founders who rely less on theorising and more on using the tools above
Teams that rapidly evolve and adapt to the dynamic AI landscape
Younger Teams
Founders that are Gen Z have the advantage of being “AI-first natives”
The advantage here is that these founders are first time learners and do not need to unlearn previous approaches or transition from prior industry practices
This allows for fresh thinking about the various applications of AI technologies
Customer Understanding
Deep understanding of how AI will change customer workflows
Crystal clear insight into why customers would adopt their solution rather than others that are present in the market
Ability to articulate how their product transforms customer experiences
Distinctive Insights
Novel perspectives on the market, customers, and the problem statement
Teams that exploit insights that others will not touch or have not recognised just yet
Compelling forecasts on where the market is potentially heading and why and how their product takes the market there
Team Quality and Adaptability
Ability to rapidly adapt as markets and technology evolves
Capacity to attract other talented individuals to join their mission
On Investment Strategies and Challenges 📝
Different strategies are required for different segments of the AI stack:
High Uncertainty in the Application Layer
Numerous companies are pursuing similar opportunities
Thus it is difficult to predict which companies will achieve breakout success
If there are players that claim to be winners in this space, they are “fooling themselves”
Selective Approach — similar to the 5Ts framework
Wait to see which companies demonstrate real traction before investing large amounts of capital
Making early bets based on founder quality and technological differentiation
Focus primarily on identifying the best founders and the best technologies
Volume Strategy - numerous investments across the AI landscape as picking winners is more than challenging in the current environment
Conclusions
There is a cautious optimism in the air about India’s future in AI given the significant opportunities and their accompanying challenges that lie ahead. The best AI companies are those that achieve rapid revenue growth, faster than previous technology waves.
While India may not lead in foundation models or capital investment, a strategic focus on AI applications, services, and specific sovereign use cases could position the country as a dominant player in the global AI landscape.
If you have reached this far, thank you very much for reading this issue from Consortium. I have loved every bit of research, writing, reading, scrapping, and re-writing. I hope you have enjoyed this as much as I have, and have hopefully learnt something new along the way.
See you soon!
MVP
Data provenance refers to tracking where data comes from, how it was collected, processed, and transformed before being fed into AI models. It’s essentially a full audit trail of the data’s life cycle.