Chapter 2: The Cost of Captivity

It is suitable to ignore the captivity if it brings comfort, however artificial it may be. Until costs start accumulating.

It is too convenient to be always good.

Raji had returned from a vacation only the previous night and was still battling jet lag. The thought of driving through Bengaluru's morning traffic was the last thing she wanted. Fortunately, she had another option available to her, i.e., to work remotely just as seamlessly. Her collaborative development environment was only a click away on the computer, and her colleagues in California were just a Microsoft Teams call away. Within minutes, she was reviewing codes, joining meetings, and contributing as if she were sitting in the office.

Meanwhile, Ramesh glanced at a notification on his phone reminding him that his electricity bill was overdue. June's bill was higher than expected, and his bank balance wasn't enough to cover it. It wasn't a crisis. Using the mobile app of his microfinance institution, downloaded from the Google Play Store, he instantly secured a ₹2,000 loan. The money was disbursed through the Aadhaar Enabled Payment System (AEPS) [1], allowing him to pay the bill immediately and avoid late payment penalties.

Thousands of miles away, Jagjit was staring at an email that had just landed in his inbox. His company's latest video-editing software has been a global success, but the sudden surge in users has overwhelmed the infrastructure. Customers across the world were experiencing timeouts. The message from his manager in Copenhagen was asking him to fix the issue as quickly as possible.

Jagjit didn't panic. This was exactly the kind of problem he had anticipated. He logged into his company's AWS account that he had just signed up for, ported the application to the Amazon Elastic Compute Cloud (EC2), and redirected traffic by updating the application's DNS records. Within two days, the platform was running at full scale, seamlessly handling the global spike in demand.

All amazing success stories of how American hyperscalers were enabling remarkable productivity and choices for Indian consumers and businesses. This seems too good to be true until it isn’t.

As long as AWS, Microsoft Azure, and Google Cloud continue operating normally, few people ask who ultimately owns the infrastructure powering our digital economy. Yet beneath this convenience lies one of the most significant strategic dependencies modern India has ever created.

This chapter asks a simple question.

What does this reliance cost India?

The answer extends far beyond cloud hosting bills.

It affects where wealth is created, where engineering talent migrates, who writes the rules governing our digital economy, who controls our AI future, and ultimately how much sovereignty India retains over its own digital civilization.

By 2024, Indian organizations were spending well over six billion dollars annually on cloud services, with most of that revenue flowing to American hyperscalers [2]. 



The result is a continuous outflow of wealth that receives remarkably little public attention because it is fragmented across thousands of organizations. It improves shareholder returns in Seattle or executive bonuses in Redmond. More research laboratories in California are built using these profits, and these large companies acquire more Indian AI startups at its infancy using this money, thus expanding its reach and monopoly. It also funds construction of additional data centers abroad thus strengthening their ability to dominate the markets.

Indian companies and consumers become more indentured by these corporate entities.

With one-third of that amount, India could operate a hyperscale for its entire consumer and business population.

Perpetual Talent Loss

Years ago, Jayraj left Kerala for the Gulf in search of better opportunities. The state of Kerala had fully funded his education at a premier public engineering institution in Thiruvananthapuram. Some of his income returned as remittances, but he never did. The talent Kerala had nurtured and developed never came back to create value for it. Many such talents were lost forever.

Similarly, many of India’s highest-paid and most strategically valuable engineers now work for foreign hyperscalers. Their skills are effectively exported permanently. In return, India gains deeper dependence on these companies and on the political systems that support them.

Increasing Influence Deficit

A few months ago, I attended a global technology conference where one of America’s premier hyperscalers was the principal sponsor. Several political and administrative leaders from various countries also participated in the conference. There was a panel exclusively discussing digital infrastructure in South Asia where several key discussions, particularly, but not limited to AI and ethics, happened, and decisions were made. Even discussions about future government policy often assumed these platforms would remain the foundation upon which India's digital economy would continue to grow.

It wasn’t something I hadn’t seen before. The monopoly over decision-making is in the hands of a few companies. However, what was utterly depressing was that there was no Indian company of comparable scale presenting its own vision. No domestic hyperscaler explaining a different process or no competing ecosystem advocating a uniquely Indian approach to AI.

When their technology powers thousands of households and businesses, it is natural that policymakers seek their input more or weigh their input higher because their systems have become critical to the functioning of the economy. They land on every industry association invitee list because people want to hear from them more than anyone else. Colleges and Universities tailor their curriculum to teach their platforms because those are the skills employers’ demand.

It illustrates an important reality: influence follows capability.

I had the opportunity to speak with Sachin [3], an associate professor at the Centre for Data Science at an Indian IIT, who had been invited to propose a topic and participate as a speaker at an international conference in Silicon Valley. He proposed a session on linguistic diversity in algorithmic models, arguing that AI systems used in India had to work reliably across dozens of languages, scripts, dialects, and cultural contexts. A few days later, he received a call from Maurice Chapman, the convenor at the conference. The topic, she explained, was important, but the program committee wanted the session framed more broadly around “responsible AI safeguards” because that theme aligned better with the conference’s sponsors, aka one of the largest hyperscalers in the country. Sachin agreed to adjust the title and emphasis, but the change was telling: an issue central to India’s AI future had to be translated into a language that fit someone else’s priorities.

The episode shows how companies that build foundational infrastructure can also shape the conversations around it. [4] They influence which standards are discussed, which regulations are prioritized, which government decisions receive attention, and which long-term technology strategies become dominant.

The problem is not only that decisions are made elsewhere; it is that, increasingly, the conversation about our future also starts somewhere else.

The AI Levy

Imagine this scenario.

At his Mumbai startup, Aravind, the CEO, opened the leadership meeting with urgency: “Our angel investors want us at the Silicon Valley Data Summit next month. If we launch the beta there, we could secure a Series A that funds us for three years.”

Priya (CFO): “We can’t launch next month. Our runway is four months, and the product costs will drain us before we even reach San Francisco.”

Aravind: “What changed? This wasn’t an issue a few weeks ago.”

Rahul: “OpenAI and Anthropic changed their enterprise API pricing overnight. Input tokens are up 40%, reasoning output tokens have doubled, and our entire product runs on their models.”

Priya: “Our testing pipeline ran all night for the summit. With the price hike, we burned ₹3.5 lakhs ($4,200) in just four hours.”

As AI enters banking, education, healthcare, government, productivity tools, and enterprise software, these payments become routine operating costs. [5] The more India digitizes, the more value flows to the owners of the underlying AI infrastructure. Indian firms may own the application layer, but the foundation keeps collecting rent. Once businesses, datasets, workflows, and models are locked in, the “tax” becomes difficult to avoid.

This is the AI toll: You can certainly build successful businesses. But you are hostage to certain critical raw materials. Unless there is a contingency, which is India’s own AI model.

No issues in using services and products developed and delivered from foreign entities. But complete reliance on them is disconcerting. It puts us at an economic disadvantage because we will be captive of their pricing and financial models. But a more strategic risk is their potential susceptibility to political pressure from their governments. If AI becomes a staple of every digital tool, this is a national security risk.


Political Implications

I had a Russian colleague in the US who had to return for family reasons. He had established firm contacts at many US companies, so he was very confident that he would continue to be able to work for these companies and earn good income. He worked remotely as a contract front-end developer for these companies, while also finding additional clients through freelancing platforms like Upwork and receiving payments in US dollars. He was entirely disconnected from Russian state politics, relying purely on the global internet economy to support his family.

He wrote to me a few years ago that eventually, US geopolitical pressure systematically dismantled his livelihood. Upwork and Fiverr suspended all accounts based in Russia, instantly cutting off his access to clients and ongoing contracts. [6]. The US banned major Russian banks from the SWIFT banking system, and American financial giants like Visa, Mastercard, and PayPal completely halted operations in Russia. Maxim’s US dollar earnings were frozen mid-transit, and his cards stopped working for international purchases. [7]


It is not a failure; it is a clarion call for an insurance policy

In all fairness, we cannot blame foreign hyperscalers for creating this dependency. They built extraordinary businesses by investing hundreds of billions of dollars, attracting exceptional talent, and executing with remarkable consistency.

India has demonstrated the ability to be innovative in many areas. Also, India had to do with what was available to cater to the expectations from millions of Indian businesses, developers, students, and consumers who needed access to world-class infrastructure. It would have been difficult to build domestically at the same speed and scale.

Neither side was wrong. But India now needs to think bigger. We should build our own strong backup, so we are not fully dependent on foreign companies. Every major country tries to control the technologies that matter most to its future. Cloud infrastructure and AI have now become that important.


The Mindset That Holds Us Back

If you ask any T-Hub founder what cloud provider they planned to use for their new product, I am certain they would pick AWS or Azure. If you ask them whether they would ever consider an Indian hyperscaler, if it existed, the answer would be an emphatic NO. I am not surprised, because they want to focus on their core business while ignoring the macroeconomic issues from dependency and overreliance. Firstly, that is not their problem, at least not yet. Secondly, there may be a lack of conviction that India can deliver such a hyperscaler with needed precision and scale.

The cost of dependence is not just on the monthly bill or the dollars leaving the country. It is this assumption that the foundation of our digital future would naturally come from somewhere else.

Many Indian entrepreneurs have grown up in that same environment. They dream of building apps, platforms, fintech products, AI tools, and global companies. But they cannot imagine building the infrastructure underneath them. The idea of an Indian hyperscaler feels too large, too expensive, or too far away.

For the Indian mind, it seems highly unrealistic to be able to successfully complete such a capital-intensive project. Yet every technological giant once looked impossible. The East India Company began as a trading shop. Don’t we all know that Amazon began as an online bookstore while Google began as a research project?

We have our own examples. The greatest of them is ISRO. India’s space program once looked nearly impossible and unnecessary for a newly independent country stuck with some serious survival issues. We still pulled it through even if we had to carry rocket parts on bicycles and bullock carts. [i] Today, ISRO has reduced India’s dependence in three practical ways: launch capability, satellite services, and navigation. [ii]


[1] https://www.npci.org.in/product/aeps

[2] https://my.idc.com/getdoc.jsp?containerId=prAP53678525

[3] Name changed to preserve anonymity. Throughout this book, certain names and identifying details have been altered to protect the privacy of the participants

[4] https://policyreview.info/articles/analysis/platform-power-ai-evolution-cloud-infrastructures

[5] https://www.linkedin.com/pulse/5-layers-ai-where-india-truly-stands-praveen-kumar-b-9664c

[6] https://www.wsj.com/world/europe/russia-war-ukraine-business-sanctions-11647647580

[7] https://mfr.mv/glob/the-retrogressive-effects-of-sanctions

[i] https://pib.gov.in/newsite/printrelease.aspx?relid=165729

[ii] https://www.isro.gov.in/Satellites.html