How the Day Actually Played Out

Morning open (positive): Sensex opened up 42.51 points (0.06%) at 74,901.50, and Nifty added 39.75 points to open at 23,454.05, as crude oil cooling below $100/barrel and softer US 10-year bond yields lifted sentiment. Asian markets broadly supported this — South Korea’s Kospi jumped 1.9%, Hong Kong’s Hang Seng rose 0.36%, and Wall Street had closed Monday near an all-time high.

But by the close, IT and financials dragged the whole market into the red:

• Nifty50 closed at 23,329, down 85.30 points (0.36%) — snapping a four-day winning streak
• Sensex closed at 74,529.08, down 329.91 points (0.44%)
• Nifty MidCap and SmallCap also finished slightly lower (−0.08% and −0.23%)
• TCS, Trent, and Tata Consumer Products ended as the day’s top Nifty50 losers

The IT Sector’s Specific Damage

Through the session, individual stock moves included:

• HCL Tech: down as much as 2.44% intraday, closing among the sharpest losers — one report clocked it down 1.82% by midday, another at −2%
• Tech Mahindra: down 1.4–1.5%
• Infosys: down roughly 0.7–1.4% depending on the reading
• TCS: ended as one of the Nifty50’s top three losers outright

Why the Reversal — More Than Just H-1B

While the H-1B fee fallout is the sector-specific driver for IT, several other forces converged on the broader market Tuesday:

• Geopolitical uncertainty and elevated global bond yields kept investors cautious even after the morning’s positive cues
• Foreign Institutional Investors (FIIs) turned net sellers, offloading ₹576.20 crore worth of equities the previous session (Sept 21) — though Domestic Institutional Investors (DIIs) stayed net buyers, purchasing ₹2,797.27 crore, partially cushioning the fall
• Crude oil, despite cooling roughly 7% over the prior five sessions, saw a slight rebound intraday after touching a low of $99.90/barrel — reviving some inflation-related caution

The Structural IT Story Behind the Daily Noise

This isn’t just a one-day blip — IT has been a serial underperformer through 2026, and the reasons compound each other:

• Elevated US interest rates discourage corporate tech spending, since tighter monetary policy makes companies cut client technology budgets and delay non-essential software projects — directly hitting revenue growth for Indian IT exporters, who earn the majority of their revenue overseas
• Trump-era tariffs, AI-led disruption to traditional outsourcing/services models, and the US-Iran war’s economic ripple effects have all weighed on the sector over the past year
• Indian IT firms draw over 50% of revenue from the US alone, making them unusually exposed to American policy and macro swings compared to sectors with more domestic-facing revenue

The “Two-Speed Market” Framing Holds Up

Your instinct on this is right — it’s a genuinely clean editorial narrative: on the same day, Coal India (+2.6%), Jio Financial (+2.5%), Dr Reddy’s (+1.5%), Adani Enterprises (+1.3%), and HDFC Life (+1%) were among gainers, while IT names sat at the bottom of the leaderboard. Of the 50 Nifty stocks, 28 traded positive and 21 negative — so breadth was roughly balanced, but the concentration of pain in IT specifically (a sector that’s been bleeding all year) makes for a strong visual “IT vs everyone else” segment.

The Timeline — This Has Been Building Since Last Year

• September 2025: Trump first imposed the $100,000 H-1B fee, with only a one-day implementation window (anyone entering the US after 12:01am Sept 21, 2025 was covered) — Nasscom called this deadline “unrealistic and disruptive.” Indian IT stocks lost ₹85,500 crore in market cap in a single day on that original announcement, with the Nifty IT index falling nearly 3%.
• December 2025: A US federal judge upheld the fee as lawful.
• June 2026: A different federal judge — District Judge Leo Sorokin — struck it down, ruling bluntly that “the President had no power or delegated authority to impose a tax on H-1B petitions.” That ruling was stayed pending appeal, so the fee kept operating anyway.
• August 25, 2026: The administration formally proposed codifying the fee as a $103,265 regulation.
• September 18–19, 2026: Trump signed a new proclamation extending the restrictions for another 12 months, through September 21, 2027.

The Numbers That Matter for India Specifically

• Indians accounted for 71% of all approved H-1B petitions in FY2024 (USCIS data) — by far the largest nationality group, with China a distant second at ~12%.
• Trump’s own proclamation claims the fee has already caused a 92% decline in H-1B registrations from large IT outsourcing companies — effectively confirming the policy is working as intended against Indian IT firms’ traditional staffing model.
• India’s IT sector is worth $315 billion, drawing roughly 57% of its revenue from the US market.

Important Nuance: It’s Not as Catastrophic as It First Sounds

Two clarifications matter for accuracy:

• The $100,000 fee applies only to new H-1B petitions, not existing visa holders or routine renewals — a White House clarification that came out within hours of the original 2025 announcement and eased some panic.
• Indian IT companies have been structurally reducing H-1B dependence for years — shifting toward hiring local US staff instead of rotating talent from India — which is precisely why Nasscom argues the fee’s real-world bite may be smaller than headlines suggest, even as it still objects on principle.

Nasscom’s Position (India’s IT Industry Body)

Nasscom’s response after the September 19 extension was notably measured but pointed:

• It warned the changes “can potentially have ripple effects on America’s innovation ecosystem and the wider job economy,” and would disrupt business continuity for onshore projects, requiring companies to “work closely with clients to adapt.”
• Back in August, Nasscom highlighted that the Indian IT industry has invested over $1.1 billion in US STEM talent pipelines and worker upskilling — arguing this context should factor into any fee decision.
• Some industry voices have said they see no immediate adverse impact over the next 6–12 months since the hike only affects the next application cycle — but caution a deferred impact is still likely if the policy holds.

The Separate “9-11” Fee Rule — Why It’s the Sneakier Story

This is the part with less press coverage but arguably more direct financial bite: effective September 9, 2026, the expanded Biometric Entry-Exit Fee rule now charges employers $4,000 per H-1B extension and $4,500 per L-1 extension — even for the same employee, same company, just extending an existing visa. Analysts flagged this as the sharper sting for the stock selloff that week, since it converts what was once an occasional cost into a recurring annual line item specifically for firms built around onshore Indian engineers.

Additional Political Layer

Trump also signed an executive order directing federal agencies to apply greater scrutiny to H-1B applications from companies conducting layoffs — a rule that could complicate matters for IT majors simultaneously downsizing in some units while trying to bring in specialized talent elsewhere, adding a compliance headache on top of the direct cost increase.

Bottom Line for Your Segment

This is really a three-layer squeeze on Indian IT: (1) the $100,000 new-application fee, now locked in through 2027, (2) the newer $4,000–$4,500 recurring renewal fee hitting existing staff, and (3) added scrutiny tied to layoffs — landing on a sector already down roughly 24–37% for the year depending on the stock. Nasscom’s own language — “ripple effects,” “business continuity…disrupted” — signals real concern even while downplaying panic.

The AI Energy Management Alliance (AEMA)

Launched September 16, 2026 by Google, Nvidia, and startup Emerald AI, with 18–20 launch partners including Anthropic, National Grid, AES, Constellation, NRG, RWE, and others. It’s actually a relaunch — Emerald AI’s CEO Varun Sivaram noted an earlier “Advanced Energy Management Alliance” existed back in 2014 but had largely gone dormant.

The core idea: Treat AI data centers as flexible electricity users rather than facilities demanding constant maximum power. In practice, a flexible data center could shift non-urgent computing to off-peak hours, draw on on-site batteries, use local generation, or briefly slow down less time-sensitive workloads during grid stress — while critical jobs keep running.

The pitch to utilities: In exchange for this flexibility, AEMA wants faster, larger grid connections — arguing this could unlock roughly 100 gigawatts of capacity already sitting unused on the existing US grid, since Sivaram noted the grid runs at only about 50% utilization on average, and each 10% gain in utilization could lower electricity rates by roughly 3.4% (per a cited Brattle Group estimate). The group also claims the approach could save about $733 million in system costs per gigawatt of new buildout avoided.

Real deployments already happening:

• Google has built roughly 1 gigawatt of demand-response capacity into agreements with several US utilities
• Nvidia’s DSX Flex software (built for its upcoming Rubin GPU generation) lets operators schedule AI training around grid stress or even push stored power back to utilities during shortages
• Emerald AI and Nvidia have completed six global demonstrations of flexible data centers
• Later in 2026, Nvidia, Digital Realty, and Emerald AI plan to switch on what they call the “world’s first power-flexible AI factory” in Virginia, drawing nearly 100MW

Texas Freezes New Data Center Hookups

This is the other side of the same coin — and it’s a dramatic story. In early September 2026, grid operator ERCOT, at Governor Greg Abbott’s direction, became the first major US grid operator to stop processing new large-load interconnection requests, including data centers.

The numbers behind it:

• Texas’s interconnection queue ballooned from ~48 gigawatts in 2023 to 474 gigawatts today — nearly a tenfold jump in three years
• That’s more than five times Texas’s own record peak electricity demand
• ~90% of the new power requests are data centers, per Abbott
• Nationally, interconnection requests across major US grids now top 700 gigawatts — roughly 10x the 60–70 GW the entire US data center fleet actually draws at any given moment

The real problem — “ghost demand”: Developers are filing speculative requests at multiple utilities simultaneously to secure a queue position cheaply, often without a signed customer, secured financing, or any real intent to build — because a queue position itself has resale value, or because landowners file to make a parcel more attractive to buyers. PUCT Chairman Thomas Gleeson summed up the core anxiety: “When you don’t know what is real, you really don’t know how to build the infrastructure for it.”

Evidence it’s working elsewhere: Utilities using stricter screening have seen huge drops in “real” demand — Exelon cut its high-probability estimate by 40% (to ~11 GW) after imposing collateral requirements, and Ohio’s AEP saw its data-center pipeline fall by more than half after charging a $100,000 grid-study fee upfront.

Other states following suit: Pennsylvania’s Gov. Josh Shapiro signed a similar executive order in August requiring stricter permitting for projects over 25MW — of more than 100 announced Pennsylvania projects, only about 20 had actually applied for permits under the new rules.

Timeline: ERCOT aims to complete its audit by December 2026, feeding results into its next long-term load forecast. Until then, no new interconnection agreements above the review threshold move forward.

Why These Two Stories Belong Together

They’re really two responses to the same underlying crisis — AI’s power demand outpacing grid capacity to plan for it. AEMA is the industry’s “cooperative” answer (make data centers grid-friendly to earn faster hookups), while Texas’s freeze is the regulatory “brakes” answer (stop trusting inflated demand numbers until they’re verified). Together they connect directly to your earlier data center political backlash segment — this is the utility/engineering layer underneath that political story.