deepak varier

[deepak varier @Deepak_Varier]

The off switch was always there. Fable 5 is just the first time we all watched someone use it.

[deepak varier @Deepak_Varier]

Anthropic launched Fable 5 days ago calling it state-of-the-art. Last week the US government ordered it switched off — for every user, everywhere, within 24 hours.

Most people are covering this as an AI safety story. I think that's the wrong story.

[deepak varier @Deepak_Varier]

If your product has a foundation model as load-bearing infra — agents, copilots, workflow tools — this is the week to actually test your fallback model in production, not just have one configured and untested.

[deepak varier @Deepak_Varier]

India built a $300B IT services industry on one model.

Smart people. Delivered at scale. Charged by the hour.

AI doesn't break the "smart people" part.

It quietly breaks the "charged by the hour" part.

That doesn't show up in one earnings call.

[deepak varier @Deepak_Varier]

The companies are placing the right bets.

Infosys + Anthropic. Wipro's AI-native unit. TCS platforms.

But TCS's $2.3B in AI revenue is 7–8% of a $30B business.

The core is still priced the old way. The transition is real. It is early.

[deepak varier @Deepak_Varier]

If you're building AI products for enterprise clients, this is the frame that matters:

When your product makes a team 30% more efficient — does it make you more valuable, or does it just reset what they expect to pay?

Usually both. The question is which one moves faster.

[deepak varier @Deepak_Varier]

I watch this from the client side — inside a large Indian enterprise that uses these services.

The renewal question has changed.

Before: how many engineers do we need?

Now: how many do we actually still need?

Contract by contract, the pricing power is compressing.

[deepak varier @Deepak_Varier]

The obvious fix: shift to outcome-based billing.

Stop charging for effort. Start charging for results.

It sounds clean. It is not.

For a 600,000-person org running thousands of client engagements — both sides need to change simultaneously. That is a large, slow move.

[deepak varier @Deepak_Varier]

HCL's CEO said the quiet part out loud.

"AI deflation" — 3–5% revenue dip as AI compresses delivery costs.

I suspect that number exists in every major Indian IT company's internal forecast.

He was just the one who said it on an earnings call.

[deepak varier @Deepak_Varier]

TCS: best margins in four years. Net profit +12%. AI revenue $2.3B.

Stock fell 9% in a single session.

Good results. Market selloff. One thing explains it. 🧵

[deepak varier @Deepak_Varier]

FY26 numbers are specific.

Revenue per employee rose 3–4% across TCS, Infosys, HCLTech.

But TCS total revenue slipped 0.5%. 23,460 employees cut. Fresher hiring down.

Fewer people. Higher output per person. Flat revenue.

That gap is the whole story.

[deepak varier @Deepak_Varier]

Indian IT's pricing model runs on one logic:

More complex work → more people → higher revenue.

AI breaks the middle step. The work stays complex. Fewer people needed. Value delivered is similar. The bill is smaller.

That is the whole problem in three sentences.

[deepak varier @Deepak_Varier]

Full breakdown here, using Wispr Flow as the case study:

https://t.co/eP2Rloc3H3

Curious what you're seeing if you're building in India, MENA, or ASEAN.

[deepak varier @Deepak_Varier]

Wispr Flow has 14% of its users in India.

2% of its revenue.

I assumed this was a pricing problem. It's not. It's a fundamentally different kind of problem.

[deepak varier @Deepak_Varier]

Voice AI isn't too early for India. Demand is real.

The problem is Western SaaS economics were bolted onto it. That doesn't survive at ₹10/month scale.

The founders who rethink the cost model, not just the price, will own this market.