Value Capture Has Nothing to Do with Value Creation
February 3, 2026. Anthropic ships a plugin for legal workflows. $285 billion vanishes from global markets.
Same day. Palantir reports earnings. Stock jumps 7%.
Same narrative. Opposite result.
What happened
Anthropic released plugins for Claude Cowork that automates legal work. Contract review, NDA triage, compliance checks. An AI junior associate that never sleeps, never bills by the hour, never asks for a raise.
Same day, Gartner reported Q4 earnings. Numbers were fine. 2026 guidance came in ugly. CEO said the selling environment has gotten tougher. Gartner opened down 31%. Stock has fallen 71% in 52 weeks.
Then contagion. Software stocks had their worst day since April’s tariff panic. ServiceNow, Salesforce, Adobe, HubSpot, Atlassian, DocuSign, Workday, fresh 52-week lows across the board.
By February 4, it jumped continents. Infosys crashed. TCS crashed. Wipro, HCL, all of them. ₹1.8 lakh crore wiped from the top five IT firms in a single morning. Japan, China, London, all red. Thomson Reuters fell 16%.
All from a plugin.
But Palantir went up.
Everyone’s talking about what Anthropic built but nobody’s talking about what market actually repriced.
This wasn’t an AI panic. This was a valuation reckoning that used AI as its excuse.
Infosys and TCS were trading at P/E ratios above 20. Wipro was sitting at 18. It didn’t shake as much. IT services in India have historically traded at 15-18x. Anything above that was premium priced for growth that AI just made harder to promise.
Gold falling, not gold disappearing.
FIIs were already looking to exit after the 2026 budget. Anthropic gave them the headline. Capital was packed and waiting for a taxi. The plugin was the taxi.
The tool affects a tiny slice of Indian IT workforce. Legal, HR, finance back office. They’ll get more optimized, not eliminated. The selloff wasn’t proportional to the threat. It was proportional to the correction the market wanted to execute anyway.
Palantir. P/E of 242. Jumped 7% the same day everything else crashed.
Palantir tells the story of being the company that deploys AI into government and enterprise workflows. Valuation is insane by any traditional metric. But the narrative is credible.
February 3 didn’t price AI capability. It priced narrative credibility.
But core the tension is
There’s a massive gap between creating value and capturing it. Most companies don’t think about this gap. The market thinks about nothing else.
TCS creates billions in value for its clients. So does Infosys. A plumber creates enormous value keeping your house from flooding. Doesn’t mean plumbing is a high-margin business.
Palantir doesn’t create more value than TCS. Might create less. But it captures value differently. Positions itself as irreplaceable infrastructure. Data platforms woven so deeply into government operations that removing them would be like removing the nervous system. Whether that’s true is almost beside the point.
Indian IT’s story has always been “we have smart people who do good work at a fair price.” That gets you a modest multiple.
Palantir’s story is “we are the AI-powered operating system for the world’s most important institutions.” That gets you an absurd one.
Both stories are probably exaggerated. One creates multiples.
Most Indian IT don’t have a credible AI story
Under Salil Parekh, the market never got a credible future narrative. He’s talking about 4,600 AI projects at Davos. Impressive operationally. But the market doesn’t pay for operational updates. It pays for a vision of what you become.
TCS, Wipro, HCL, same problem in different wrappers. They have the assets. They have the clients. Indian IT has been a value trade for a decade. Never a narrative trade.
When you can’t tell the market where you’re going, the market decides for you.
Ravi Kumar at Cognizant is doing something none of the others are. He’s framing Cognizant around a specific gap: “the AI velocity gap.” Massive infrastructure spending, no business value realization yet. Cognizant positions as the builder who bridges that gap.
He’s not saying “we do IT services.” He’s saying “enterprises have spent hundreds of billions on AI infrastructure and haven’t seen returns yet. We’re the company that converts that spend into measurable results.”
Revenue above Street estimates. 28 large deals in 2025. Large deal TCV up nearly 50%. Five mega deals including a billion-dollar engagement. Shifting from linear staffing to agentic delivery models. Partnered with Anthropic and Microsoft.
Same industry. Same macro. Same AI headwinds. Different story.
What the plugin actually did
The plugin is real but it’s not a revolution. Dozens of startups already do this. Harvey AI, Legora, others.
What spooked markets is who built it. Anthropic is the model maker. When the model maker starts shipping the applications, the entire competitive landscape shifts.
You run a chain of burger joints. Your beef supplier has always been a supplier. One morning you wake up and they’ve opened their own restaurant across the street, same beef, half the price. Your food isn’t worse. Your economics just got harder.
That’s Thomson Reuters, RELX, every legal tech company on February 3. Technology partner became competitor.
Gartner’s guidance hit at exactly the wrong moment. A company whose entire business is selling expert advice. CEO saying clients are delaying. In a world where Claude synthesizes the same information in thirty seconds, that reads like a confession.
Gartner stock down 71% in a year.
If one AI agent does the work that used to need ten software seats, companies will ask why they’re paying for ten.
The application layer isn’t safe anymore. Conventional wisdom was models commoditize, apps on top capture value.
Anthropic broke that. When the model maker ships the interface, it’s vertically integrated. Every SaaS company that thought they’d own the layer above just lost that bet.
Cursor’s entire business depends on the model underneath it. The model maker just became a competitor. That math doesn’t work for long.
Panic moves fast. Adoption moves slow. Enterprise clients don’t switch to a Claude plugin over a weekend. Contracts, compliance, institutional inertia. Years, not days.
The market priced five years of disruption into one morning.
A tool that reviews NDAs doesn’t replace the team building a bank’s core trading platform. BFSI clients can’t rip out their IT services and swap in AI agents. Security, regulatory constraints, institutional knowledge that lives in people. The market treated this plugin like a neutron bomb. It’s more like a very efficient new stapler for one department.
Where value goes
Models are commoditizing. Electricity layer. Electricity changed the world but the power plant was never the best business.
Application layer is under pressure. Companies with deep integrations and proprietary data survive. Thin interfaces over an API don’t.
Harness and context is where value capture goes. Whoever controls the agent’s memory, permissions, integrations, and the user’s context owns the relationship. Model is muscle. Harness is the brain.
Look at what Anthropic did when Clawdbot showed up. Open-source project that wrapped Claude into an always-on personal agent. Persistent memory, messaging integration, real computer access. 60,000 GitHub stars in weeks. The company that just caused a $285 billion selloff reacted to Clawdbot like it was an existential threat. Tightened enforcement on third-party harnesses. Pushed trademark claims. Forced a rebrand. Twice.
Why would the model maker panic about an open-source wrapper? Because the wrapper was capturing the value. The context, the memory, the user relationship.
The model does the work. The harness decides what work is worth doing. That’s where the margin lives.
Back to the bounce
The selloff fades. Markets overshoot on fear.
But the question remains. Who has a story.
US tech gets absurd multiples on narrative. Indian IT gets modest ones. Indian IT is arguably more capable at enterprise delivery than Palantir.
India doesn’t have the insane valuations. We’re far below. A company here getting even a fraction of Palantir’s multiple would look completely different. The opportunity is to position as the credible alternative that makes Palantir’s promises real at a fraction of the cost. Nobody in Indian IT is telling that story.
Markets don’t pay for value. They pay for the story about value.
With inputs from Kashi