AI and its spin-off gains

The Artificial Intelligence (AI) industry in the U.S. is on a high, reflected in the trio of major AI upcoming initial public offerings by Anthropic, OpenAI and SpaceX, hogging the headlines and retail and institutional investor funding. Even the globally disruptive Iran war is a distant event from Wall Street’s enthusiastic embrace of AI.

The ebullience may be ebbing. The past few weeks have seen some developments which indicate that AI enthusiasm may be cooling down due to the high cost of its use – enough to give even cash-rich companies a pause. Companies such as Microsoft[i] and Amazon[ii] have asked employees to cut down their use of AI, the former by cutting down subscriptions for Claude and the latter by reduced emphasis on maximising the use of tokens, a measure of AI usage (See box). Uber, an eager user of AI for its activities, has run through its entire AI usage budget for 2026 in just four months[iii].OpenAI CEO Sam Altman acknowledged in a recent media interview that companies are beginning to question the return on investment on AI tools[iv].

The impending public issues of the two largest AI companies – Anthropic and OpenAI – both currently valued at $1 trillion each should also give pause. On May 26, Anthropic raised $65 billion at a valuation of $965 billion[v]. On March 31, OpenAI announced it had raised $122 billion at a valuation of $852 billion. OpenAI also filed for its IPO on June 8, 2026, less than 2 weeks after Anthropic.[vi] Rocket-maker SpaceX, which has a substantial AI component, filed its prospectus with the SEC on May 20 to raise $75-$80 billion, at an overall valuation of $1.25 trillion.[vii] None of these companies is currently profitable, though Anthropic is said to be on track to report its first ever profitable quarter (April-June 2026). Apart from these three, Alphabet (formerly Google) has recently announced a proposed share sale to raise $80 billion to finance AI infrastructure.[viii]

The cutback on AI spend by corporate clients coupled with the rush to tap public markets may be a sign that the AI boom may be slowing – the maturing of an industry or the popping of a bubble, depending on how an observer chooses to view it. While AI may bring about long-term changes in business and the economy, it is also likely to be commodified, making the trillion-dollar valuations unsustainable.

However, it will be a mistake to write off the euphoria entirely, as investments made during booms often lead to long-term innovations. For instance, the optic fibre cables laid during the dotcom boom of the 1990s provided cheap bandwidth, server infrastructure, and trained technical manpower, which helped companies like Amazon, Google, and Facebook become the giants they are today.

The investments in AI are likely to lead to similar spin-off gains. For instance, the power-hungry AI datacentres are leading to innovations in energy, as well as a revival of investments in the U.S. Amazon[ix] and Google[x] have both signed agreements to buy small modular reactors – a promising but expensive technology which looks viable when trillions of dollars are at stake. Microsoft is restarting the Three Mile Island nuclear reactor – which famously shut down after a leak in 1979 – to power its data centres. Meta (Facebook) has signed multiple agreements with companies promoted by Sam Altman and Bill Gates to put up 6,600 MW of nuclear power by 2032.[xi] The U.S. nuclear sector, which had been practically moribund since 1980, is now witnessing innovation not seen since the 1960s.

Scores of investors are finding rich pickings in smokestack stocks, especially in the energy equipment sector and conductive materials like silver and copper. The order backlog for companies like ABB is at record highs. And old Boeing aircraft are being purchased as their jet engines are repurposed to power data centres.[xii]

Why does this matter for India? India has left itself out of the AI cycle almost entirely and is lagging far behind. During 2025 and 2026, there has been a substantial net outflow of foreign capital from the Indian markets – Rs 160,102 crore ($18 billion) in 2025 and Rs 258,717 crore ($27.9 billion) so far in 2026.[xiii] While there are multiple reasons for investors cooling off on India, American tech sucking up hundreds of billions of dollars has definitely played a significant role. This financial outflow accounts for the lacklustre performance of the Indian equity markets over the past two years and has resulted in a loss of business confidence as companies find it difficult to raise capital. The lack of funding seems especially stark in the startup sector, which has seen sharply reduced fund flow compared to the pre-2022 era. A slowdown of the AI boom may see foreign investors return to India, which should revive investments and business activity.

More important is the serious implication for India’s IT services sector, which employs over 6 million and generates annual revenue of over $250 billion.[xiv] This money, in turn, drives much of India’s consumption and the economy. The 40-year-old model of these companies is dated, and little new investment has been made so the sector can cross the chasm. There is a long-held fear that AI will replace entry level jobs in the IT/BPO sector. Sam Altman, CEO of OpenAI[xv] and Dario Amodei, the CEO of Anthropic[xvi] have both warned of large-scale white-collar job losses due to AI in the past. Of late, both have softened their tone, projecting AI as a productivity multiplier rather than a job destroyer.

A recent paper by Vishal Sikka,[xvii] the former CEO of Infosys and an early backer of OpenAI, suggests that there could be limits on Agentic AI being able to perform complex tasks at an affordable cost, as the token costs increase exponentially. Neither of these dimensions will save India’s services companies from a rocky future.

Generative vs Agentic AI

AI can be broadly divided into two baskets. One is Generative AI which responds to queries, questions and prompts to create responses. These are models such as Gemini (Alphabet), Grok (X), and ChatGPT (OpenAI). The other is Agentic AI which is autonomous and works on its own to achieve a high-level goal like fixing a software bug or building a feature. Simply speaking, Generative AI can assist a human worker in becoming more efficient, while Agentic AI seeks to (eventually) replace the human worker entirely. Large corporations typically use Agentic AI, and till recently, had been pushing employees to use more of it. Enterprise models of ChatGPT, or Claude (Anthropic) are examples.Use of both types requires the use of tokens – AI processes text in the form of tokens. The cost of tokens ranges from $1 to $5 per million for input tokens and $5 to $25 per million for output tokens. An input token is what a model reads, while an output token is what a model generates. Generative AI, which most people use, consumes a few hundred tokens per query. Agentic AI consumes much more – hundreds to thousands of times the number of tokens. The cost of using Agentic AI is giving companies such as Microsoft and Amazon pause. In some cases, employees ended up using billions of tokens,[xviii] pushing up costs overall.

Indian tech companies have traditionally spent less on research compared to their counterparts in the West, or even Chinese tech majors, and are left out of the AI race. The sector is now in the unenviable position of waiting for AI valuations to mature so investments can revive. The tech infrastructure created during the dotcom boom helped Indian tech companies by first creating demand for their services (originally body shopping) and then making offshoring/outsourcing possible. It is not clear if any such benefits exist for India’s tech industry this time round.

It is possible that a more innovative set of founders, similar to the tech startups of the past decade, benefit from this round of AI innovation. For the Indian government, this is a good time to make the markets more friendly for long-term investors, including Indian retail investors who have invested heavily locally, by removing irritants such as long-term capital gains tax. A move has been made to eliminate taxes from bond investments; a similar move toward equity investment can finance AI start-ups and may ensure that India gets a foot in the AI door.

Amit Bhandari is Senior Fellow for Energy, Investment and Connectivity.

This article was exclusively written for Gateway House: Indian Council on Global Relations. You can read more exclusive content here.

Support our work here.

For permission to republish, please contact outreach@gatewayhouse.in

References:

[i] https://sqmagazine.co.uk/microsoft-drops-claude-code-github-copilot-cli/

[ii] https://www.businessinsider.com/token-reckoning-amazon-uber-reassess-ai-investments-2026-6

[iii] https://finance.yahoo.com/sectors/technology/articles/ubers-anthropic-ai-push-hits-223109852.html

[iv] https://www.cnbc.com/2026/06/01/cnbc-exclusive-transcript-openai-ceo-sam-altman-speaks-with-cnbcs-david-faber-on-power-lunch-today.html

[v] https://www.anthropic.com/news/series-h

[vi] https://www.cnbc.com/2026/06/08/openai-confidentially-files-for-ipo-prepping-wall-street-for-ai-debut.html

[vii] https://www.sec.gov/Archives/edgar/data/1181412/000162828026036936/spaceexplorationtechnologi.htm