For two decades, Indian IT services companies built one of the most successful global business models in history. Take relatively low-cost engineering talent, sell it to Western enterprises at a multiple of cost, manage execution risk better than competitors, and repeat at scale.
That arbitrage created giants: TCS, Infosys, Wipro, Accenture, Cognizant. Combined, these firms once commanded over $500 billion in market capitalization at their 2021 peak. Today, they trade at far lower multiples, widely described as “ex-growth” but still fundamentally sound.
That framing is too generous.
The reality is harsher: the core unit of value these companies sell — the billable human hour — is being automated away in real time. This is not a cyclical slowdown. It is not about rate cuts, deal delays, or client caution. It is a structural extinction event.
The all-time highs for IT services stocks are unlikely to return — not because demand for software is shrinking, but because the way software is produced has fundamentally changed.
The Business Model: Headcount × Billing Rate
Strip away the jargon and branding, and the IT services model is simple:
Hire engineers in India at $15–40 per hour
Bill clients $50–150 per hour, depending on offshore/onshore mix
Scale revenue by scaling headcount
Defend margins through utilization, pyramids, and process discipline
A typical large enterprise deal looks like this:
~500 developers
Blended billing rate of ~$50/hour
Annual contract value of ~$50 million
Multi-year duration, focused on legacy systems, maintenance, enhancements
This model worked because software development was labor-intensive, coordination-heavy, and difficult to automate. Productivity gains came slowly and were absorbed into larger scopes of work.
That assumption is now broken.
AI Is Compressing the Core Arbitrage — Today, Not “Someday”
The impact of AI on software development is no longer theoretical.
GitHub Copilot users complete coding tasks ~55% faster
AI-generated code now accounts for ~40–50% of code written among active users
Acceptance rates of AI-suggested code exceed 30% and rising
Tools like Claude, Cursor, and Copilot are handling not just syntax, but logic, refactoring, testing, and documentation
This is happening now, across millions of developers, not in pilot programs.
What does a 50% productivity gain mean for an IT services contract?
It means:
500 developers become 250–300
Or the same output is delivered in half the time
Either way, billable hours collapse
And here is the existential problem:
IT services companies cannot adopt AI aggressively without destroying their own revenue base.
Every efficiency gain directly reduces billable hours. Unlike a product company, they cannot decouple revenue from labor input.
Why IT Services Can’t “AI Their Way Out”
Management teams repeatedly talk about:
“AI-first delivery”
“Generative AI practices”
“Productivity-led growth”
But structurally, this is incoherent.
If TCS deploys AI internally and reduces project effort by 30%, it has two choices:
Pass savings to the client → revenue declines
Hide productivity → clients eventually notice and renegotiate
Either way, pricing power erodes.
This is not like manufacturing automation, where capital investment replaces labor and increases margins. IT services sell labor itself. Automating labor automates revenue.
This is why the analogy often used — “AI is just another tool like cloud” — is deeply misleading. Cloud increased complexity and demand for services. AI reduces complexity and human dependency.
The Market Already Knows — But Not Fully
TCS and Infosys trade today at 20–25× earnings, down from 35–40× during the 2021 peak. Many investors see this as fair valuation compression due to slower growth.
But valuation compression implies a lower growth trajectory, not a shrinking addressable model.
What’s unfolding looks more like:
Flat to declining headcount
Price pressure on renewals
Margin defense through layoffs and utilization
Narrative-driven “AI revenues” that are mostly reclassified services
When management presentations start to resemble funeral homes advertising premium coffin designs, it’s worth pausing.
The market is pricing these stocks as mature utilities. The risk is that they are closer to structural decline businesses.
AI Agents Change the Equation Entirely
The next phase isn’t just copilots assisting humans. It’s AI agents replacing entire workflows.
We are entering an era where:
An AI agent takes a feature request
Plans architecture
Writes code
Generates test cases
Debugs failures
Deploys changes
This collapses the need for:
Large junior developer pools
Manual QA teams
L1/L2 support engineers
Large maintenance squads
The traditional IT pyramid — the foundation of margins — becomes top-heavy and economically unsustainable.
This is not a 10-year horizon. Enterprises are already running internal pilots.
“But Demand for Software Will Explode”
Yes, software demand will continue to grow.
But demand growth does not equal revenue growth for labor arbitrage firms.
When software creation shifts from:
“People writing code” to “People supervising AI writing code”
The economic surplus flows to:
AI platform providers
Model owners
Cloud infrastructure
Companies that own IP, not labor
Not to firms whose competitive advantage was hiring more people faster.
The Employment Reality: A Social Shock
The Indian IT-BPM sector employs millions of people. These are not coal miners with transferable manual skills. These are the retrained knowledge workers.
For decades, software was the “safe” career:
Learn Java
Get placed
Climb the pyramid
Retire comfortably
That assumption is breaking.
AI already:
Writes better boilerplate than a 3-year associate
Debugs faster than many senior engineers
Documents code humans never bothered to document
Industry conversations increasingly point to:
25–30% reduction in traditional IT roles over 3–4 years
Not due to recession
But due to permanent productivity shifts
Bench sizes are shrinking. Fresh hiring is slowing. The pyramid is flattening.
The False Comfort of “New AI Roles”
Yes, new roles will emerge:
AI integration engineers
Prompt engineers
Data governance specialists
Workflow designers
But these roles are:
Fewer in number
Higher skill
Not mass-employment replacements
One AI architect replaces dozens of junior developers.
This is not a net employment-neutral transition.
The TCS vs Anthropic Comparison Says Everything
Consider this contrast:
TCS: ~600,000 employees
Anthropic: ~2,000 employees
Anthropic’s output scales exponentially with compute and models. TCS’s output scales linearly with people.
That alone tells you which business model compounds in an AI-native world.
Investment Implications
For investors, the takeaway is uncomfortable but necessary:
Do not anchor to historical multiples (Past peaks were driven by a now-eroding arbitrage)
Treat IT services as melting ice cubes, not stable compounders
Be skeptical of “AI revenue” disclosures (Most are repackaged services, not margin-accretive products)
Expect permanent margin pressure (Clients will demand AI-driven efficiency sharing)
This does not mean these companies vanish tomorrow. They will generate cash for years. But cash flow durability ≠ growth durability.
Utilities don’t reclaim growth multiples. Neither will IT services.
Final Thought: Save for the Winter
The IT winter is closer than most realize.
This is not a call for panic — but for realism.
For investors: adjust expectations. For engineers: upgrade skills, reduce leverage, save aggressively. For policymakers: prepare for a white-collar employment shock.
AI was not designed to “assist” IT services.
IT services is the sector AI was designed to replace.
When the value of a developer hour trends toward zero, the arbitrage that built an empire disappears with it.
And once gone, it doesn’t come back.