Normal Guy

Intelligence Digest โ€” Aug 2โ€“3, 2026

The Week Everything Got Cheaper At Once

Iran stands down, AI tokens hit new lows, and India's Q1FY27 earnings reveal who actually makes money โ€” and who just makes revenue.

Brent โˆ’3.2% Nifty +1.6% S&P 500 +0.9% KOSPI โˆ’5.1% US 30Y 5.20% AI Token Index โˆ’88% since 2023 Gold +289t Q2 CB buys

01 · The Big Picture

Three things got cheaper in the same week, and they do not point the same way. Oil fell because Iran and the United States stepped back from a fight. AI fell because Chinese and American labs are racing each other to the bottom on price. And Indian company earnings went up โ€” but the cost of earning those profits went up faster for many of them. Each of these is a big deal on its own. Together, they form a triangle that forces every investor to pick a side: do you bet on the physical world getting cheaper, the digital world getting cheaper, or the companies that sit between the two?

The numbers that frame the week. The frontier AI token cost index (a measure of how much it costs to use the most powerful AI models) has fallen to 12% of where it was in March 2023. Central banks bought 288.9 tonnes of gold in the second quarter of 2026 โ€” 62% more than the same period last year โ€” which tells you that even as they talk about peace and cheaper oil, the people who manage national reserves are quietly hedging against something. The S&P 500's second-quarter earnings per share (the profit each share of the index earned) grew 47.4% compared to the same quarter last year, but strip out the biggest tech companies and other income items, and the number drops to 28.8%. India's power consumption in July hit 170.7 billion units, 10.9% more than last year โ€” which is one of the most reliable real-time indicators that the economy is actually running hot, not just reporting numbers.


02 · Iran Stands Down

The strike was called off. On August 1โ€“2, the United States cancelled military strikes against Iran. Saudi Arabia mediated. The deal includes opening the Strait of Hormuz โ€” the narrow waterway through which roughly a fifth of the world's oil passes every day. Oil prices fell below $79 a barrel, and US stock futures rallied on the news. The immediate fear โ€” that a military confrontation could close Hormuz and send oil above $100 โ€” is off the table for now.

But the language stayed aggressive. The US president called Iran "duplicitous" and declared that Iran "will never have a nuclear weapon" and that "the Strait of Hormuz is controlled by the United States." That is a claim of ownership over international waters, and it matters because it tells you the underlying tension has not gone away โ€” what changed is the timing, not the intent. If Iran resumes enrichment or tests a weapon, the strikes come back. This is a pause, not a peace.

What it means for India. Cheaper oil is the single fastest transmission channel into the Indian economy. Every $10 drop in Brent saves India roughly $15 billion a year on its import bill. That flows directly into the current account (the country's balance of incoming and outgoing money with the rest of the world), which flows into the rupee, which flows into how much room the Reserve Bank of India has to keep interest rates where they are. But last edition's warning about the Hormuz sulphur chokepoint โ€” that sulphur and sulphuric acid from Gulf refineries feed Indian fertiliser and speciality-chemical plants โ€” has not changed. If the truce breaks, those supply lines break first.


03 · The AI Price War

AI got dramatically cheaper this week. The frontier token cost index โ€” which tracks the price of using the most capable AI models โ€” has collapsed to 12 on a base of 100 in March 2023. That means what cost $100 to run through an AI model two years ago now costs $12. The proximate cause is a price war between American and Chinese labs. DeepSeek released its V4 Flash model on July 31, priced at $0.14 per million input tokens and $0.28 per million output tokens โ€” roughly one-fiftieth the cost of the most expensive Western models. Alibaba's Qwen released its 3.8-Max model with 2.4 trillion parameters (the internal weights that determine what the model knows), claiming performance on par with the best Western labs at $2 per million input tokens.

The margin collapse is real. If you are a company that built a business selling AI tools to other businesses โ€” charging $300 for every thousand queries, say โ€” and the underlying cost of running those queries just fell by 50 to 100 times, your customers will eventually notice. One analysis framed it bluntly: the big labs like Anthropic and OpenAI are now "killing AI-native B2B SaaS" by offering the same capabilities at a fraction of the price that startups need to cover their costs. Chinese models have taken 46% market share by mid-2026. The quality gap is narrowing while the price gap is widening.

But somebody is paying for all this. The five largest technology companies โ€” Meta Platforms, Oracle, Alphabet, Microsoft, and Amazon โ€” have accumulated $1.65 trillion in off-balance-sheet commitments (money they have promised to spend but which does not appear as debt on their main financial statements, typically through special-purpose vehicles and long-term lease structures). This is confirmed and GAAP-compliant, but it means the real cost of building AI infrastructure is much larger than what their balance sheets suggest. Meta's chief financial officer said the company is "compute-supply constrained for the foreseeable future" โ€” meaning they want to spend even more, and the limiting factor is physical hardware, not money.

Read this twice

The 61% number is misleading. A widely-cited figure claims cloud service provider capital expenditure (the money spent on data centres, servers, and networking equipment) is growing at a 61% compound annual growth rate to reach $1.1 trillion by 2027. The 61% is actually a single-year jump from 2024 to 2025 โ€” not a rate that compounds over multiple years. Current estimates for 2027 spending are $1.2 to $1.4 trillion, higher than $1.1 trillion. The direction is right, the framing is wrong.

For India, this is a mixed gift. Cheaper AI helps every Indian IT services company offer more to clients for less. But it also means the premium that Indian engineers can charge for building AI systems is falling just as fast. The "anti-AI trade" thesis โ€” the idea that India benefits precisely because it does not depend on AI revenue โ€” is gaining traction. Sterlite Technologies won a โ‚น960 crore order, its stock up over 400% this year, riding the data-centre cabling wave. HFCL added โ‚น522 crore. The infrastructure underneath AI is where Indian companies are actually making money โ€” not in the models themselves.


04 · The Leverage Hangover

Last week's momentum crash is still reverberating. Morgan Stanley's momentum index โ€” which tracks the performance of stocks that have been going up the most โ€” fell 17.4% in four trading sessions, the worst stretch ever recorded. The trigger was the liquidation of a hedge fund that had built positions four times the size of its actual capital ($45 billion in total positions), and when it collapsed, Citadel absorbed the book. The US 30-year Treasury yield hit 5.20%, a 19-year high. And yet the equal-weight S&P 500 (where every company counts the same regardless of size) hit a record high during the wreckage โ€” meaning the damage was concentrated in the most popular, most crowded positions, not in the market as a whole.

MicroStrategy (now called Strategy) is selling bitcoin. The company sold 3,588 bitcoins for $216 million, at an average price of $63,957 โ€” well below its average cost of $75,419 per coin. It is selling at a loss to fund a $400.7 million quarterly preferred dividend (a fixed payment the company promised to shareholders of a special class of stock), which is eight times larger than a year ago. The company still holds 842,138 bitcoins. This is not a panic sale, but it is the first time the company has been a net seller, and it tells you that even the most committed bitcoin holder has a price at which the carrying cost becomes unmanageable.

Prediction markets hit a record. Kalshi and Polymarket combined reached $50.59 billion in trading volume in July โ€” an all-time high. But the number is flattered by the World Cup. Once the tournament ended, open interest (the total value of bets that are still active) collapsed from $2 billion to $1.2 billion. Twelve state regulators are now suing these platforms. The volume spike may be the peak.


05 · India's Earnings Season

Q1FY27 is the busiest reporting season in years, and the picture is mixed. Across the 58 companies covered in detailed earnings tables this window alone, the median profit-after-tax growth (what is left after all expenses, interest and taxes are paid) was about 27.3% year-on-year. That sounds strong. But dig into the individual results and you see a pattern: revenue (the total money coming in from sales) is growing comfortably, while profit margins (the share of revenue that becomes actual profit) are under pressure from input costs, competition, and one-off items.

The standout results. TVS Motor reported record July sales of 6.30 lakh units (one lakh = 100,000), up 38% from last year. Electric two-wheeler sales more than doubled, up 158%. Bajaj Finance โ€” the largest non-bank lender โ€” earned โ‚น5,986 crore in net profit, up 27%, with assets under management (the total value of loans and investments it manages) reaching โ‚น5.47 lakh crore. Its net interest income (the difference between what it earns on loans and what it pays on deposits) grew 23% to โ‚น12,571 crore.

Ather Energy crossed a milestone. The electric scooter maker reported its first-ever positive EBITDA (earnings before interest, taxes, depreciation and amortisation โ€” essentially, operating profit before accounting adjustments) of โ‚น9.45 crore. Revenue grew 89% to โ‚น1,217 crore. The net loss narrowed 71% to โ‚น51 crore. This is a company that has been losing money since it was founded, so reaching positive operating profit โ€” even at a thin 0.8% margin โ€” matters. But electric two-wheeler penetration (the share of all two-wheelers sold that are electric) hit 11.2% in July, up from the full-year average of 6.5%. That is a peak-month number, not a steady-state one.

The catches inside the good numbers. DLF, the real estate developer, saw revenue fall 52.9% year-on-year โ€” a seasonal effect, since Q1 is typically slow for property handovers, but the magnitude is striking. JM Financial reported net profit down 36% to โ‚น292 crore even as revenue rose 8% to โ‚น1,200.5 crore โ€” costs grew faster than income. MakeMyTrip, the travel company, earned $9.1 million in net profit, down 64.7% from last year, despite a 5.6% rise in revenue to $285.6 million. When revenue goes up and profit goes down, the question is always: is the company spending to grow, or is it losing pricing power?

Mahindra & Mahindra's electric SUV division flipped from a โ‚น101 crore loss to a โ‚น288 crore profit. This is a clean turnaround in a single quarter โ€” the kind of number that makes you look twice. The company has not published the detailed breakup yet, so whether this comes from volume, pricing, lower battery costs, or accounting reclassification is not clear. But if the number holds and the division stays profitable, it changes the valuation math for the whole company.

The speciality chemicals revival. Bhagiradha Chemicals reported profit-before-tax growth of 601% year-on-year โ€” but last year's base was near zero, so the percentage is dramatic while the absolute number is small. Chemcon Speciality Chemicals grew EBITDA 98%. The sector spent two years in a downturn (global destocking โ€” when customers used up their existing inventory instead of buying new stock). The current quarter suggests that cycle may be over, but one quarter does not confirm a trend.


06 · The Pharma Turn

The CDMO opportunity is getting concrete. CDMO stands for Contract Development and Manufacturing Organisation โ€” a company that makes drugs or drug ingredients on behalf of other pharmaceutical companies, the way a factory might make shoes for a brand that does not own a factory. WuXi AppTec, the world's largest CDMO, reported 420,000 compounds in its pipeline โ€” the number of different drug molecules it is working on for clients. A detailed analysis of global CDMO capabilities โ€” tracked through Drug Master Files (official registrations of drug ingredients with regulators) โ€” shows that Indian CDMOs are still far behind Chinese ones in sheer breadth, but are catching up in specific niches.

Gufic Biosciences is at an inflection point (the moment when a company's growth curve shifts from slow to fast). The company is being positioned as an Indian CDMO play that has been investing in capacity for years and is now starting to see those investments turn into revenue. The detailed analysis covers its injectable capacity, lyophilisation (freeze-drying, a way to preserve biological drugs) capabilities, and regulatory approvals across markets.

Biocon may be turning. The biosimilar maker (a company that produces cheaper copies of biological drugs after the original patent expires) is being reframed as a potential turnaround. Yesafili, its biosimilar for the eye drug Eylea, has launched in the US market. The thesis is that after years of underperformance, the company's scale finally matches the opportunity. Glenmark Pharma reported revenue up 23% and EBITDA up 38% in Q1FY27.

The niche plays. Panacea Biotec has secured Paclitaxel approvals (an anti-cancer drug) in the US, UK and Africa. Viyash Life Sciences is positioned to supply Albendazole (a deworming drug) to the US during a shortage. Dishman Carbogen Amcis is making ADC payloads (antibody-drug conjugates โ€” a new generation of cancer treatments that attach a toxic drug to an antibody that targets only cancer cells). These are small companies making highly specific, hard-to-substitute products where demand exceeds supply.


07 · India Keeps Building

The government opened โ‚น84,084 crore for offshore energy exploration. The Samudra Manthan scheme (named after a mythological churning of the ocean) is the largest single allocation for finding oil and gas under the seabed around India. The number has been confirmed at exactly โ‚น84,084 crore, and it covers 600 or more MMTOE (million metric tonnes of oil equivalent) of reserve accretion (adding to the country's known underground energy reserves) โ€” not production. Finding oil is different from pumping it; this money is for the finding part.

โ‚น45,000 crore for border railways. The railway expansion along India's borders โ€” both the western border with Pakistan and the northern border with China โ€” is being accelerated. This serves a dual purpose: military logistics (moving troops and equipment faster) and civilian connectivity. Separately, the Vande Bharat train crossed the Chenab bridge โ€” the world's highest railway bridge โ€” in a test that combines infrastructure engineering with national prestige.

LIC is selling shares worth up to โ‚น31,000 crore. The government is offloading a 2.5% stake in the Life Insurance Corporation of India through an OFS (Offer for Sale โ€” the government sells its own shares to the public without the company issuing new ones). There is a greenshoe option (an option to sell an additional 4% if demand is strong enough). This is the government raising cash by selling its stake in the country's largest insurer. The circulating figure of โ‚น32,000 crore is about โ‚น1,000 crore too high.

India's power consumption tells the real story. July 2026: 170.7 billion units of electricity consumed, 10.9% more than July 2025. This is not a number that can be fudged or flatered by accounting โ€” it is the actual electricity that flowed through the grid. When power consumption grows at double digits, it means factories are running, air conditioners are on, and the economy is physically active. The number is consistent with private-sector capital expenditure (spending on new factories and equipment) growing at roughly 20%, with new factories, data centres, and warehouses being the main destinations for that spending.

The block deals tell you where the money is moving. Paytm saw a block deal (a large, negotiated sale of shares between institutional investors) worth โ‚น963.6 crore โ€” not โ‚น2,000 crore as circulated; the actual number is barely half. Meesho had a block in the range of โ‚น1,200 to โ‚น1,540 crore, not โ‚น1,900 crore. Arvind Ltd raised โ‚น500 crore through a QIP (Qualified Institutional Placement โ€” the company issues new shares to large institutional investors). Early investors in Indian startups are cashing out, and the exit prices are consistently lower than what is being quoted in casual conversation.


08 · Defence and the New Orders

India's defence exports hit โ‚น384 billion in FY26, up 62.7% from the previous year. This is a record. The five-year compound annual growth rate (the rate at which the number has grown each year, compounded) is 35.6%. The government's target for FY29 is โ‚น500 billion, which implies only 9.2% annual growth from here โ€” far below the rate already being delivered. Either the target is conservative, or the government expects the growth rate to slow sharply.

The AMCA engine deal may be cleared by August 15. The AMCA (Advanced Medium Combat Aircraft) is India's next-generation fighter jet, and the engine โ€” the most complex and expensive component โ€” has been the bottleneck. If the Cabinet Committee on Security approves the engine deal on Independence Day, it removes the last major procurement hurdle. India is also testing a 3,760-kilometre missile between August 6 and 7 from Abdul Kalam Island.

MKU won a โ‚น660 crore contract for 4,000 NW-3000 night-vision devices, to be delivered by end of 2026. South Korea is reportedly evaluating 120 Tejas fighter jets, though this claim could not be independently verified โ€” it may be a confusion with South Korea's own KF-21 Boramae programme. Pakistan has deployed 250 Chinese-made SH-15 155mm howitzers along the border, though this too is unconfirmed.


โ—† Visual Intelligence


โ—† Connecting the Dots

Three inputs just got cheaper at the same time: energy (Brent below $79 after Iran stood down from its planned strike), artificial intelligence (the AI token price index hit 12 against a base of 100 in March 2023, with DeepSeek V4 at $0.14 per million tokens), and capital costs for India's physical buildout (lower oil means a smaller import bill, roughly $15 billion saved per $10 drop in crude). Meanwhile, leverage is breaking in the West โ€” the 30-year US Treasury yield crossed 5.20%, momentum strategies lost 17.4% in four trading sessions, and Strategy (formerly MicroStrategy) sold 3,588 Bitcoin at a loss to service convertible debt.

India sits at the intersection of all three tailwinds. Its Q1FY27 earnings season is delivering median profit-after-tax growth of roughly 27%, led by companies building real things: TVS Motor shipped 6.3 lakh units (up 38%), Ather Energy posted its first positive EBITDA at โ‚น9.45 crore (that means operating profit, before interest and taxes, turned positive for the first time since the company was founded), Mahindra's electric SUV division swung from loss to โ‚น288 crore profit, and Bajaj Finance grew PAT 27% to โ‚น5,986 crore. Power generation hit 170.7 billion units in July (up 10.9%), the government sanctioned โ‚น84,084 crore for submarine construction under Samudra Manthan, and border railway spending reached โ‚น45,000 crore.

The pattern is clear: as virtual things (AI tokens, SaaS pricing power, crypto-backed leverage) deflate, physical things (power plants, submarines, EV batteries, optical fibre, pharmaceutical intermediates) are where margin and conviction are migrating.

The sharpest edge in this data: India's infrastructure companies โ€” Sterlite Technologies, KEI Industries, KEC International, Polycab โ€” benefit from both tailwinds simultaneously. Cheaper oil lowers their input costs (copper smelting, transport, polymer insulation). Cheaper AI lowers their design, procurement, and project-management overhead. And the government is writing the cheques: โ‚น84,084 crore for submarines, โ‚น45,000 crore for border railways, and a growing share of a power grid that consumed 170.7 billion units last month alone. These are companies selling copper and fibre into a buyer who cannot delay. The question is not whether they have demand โ€” it is whether global copper supply (Chile peaked at 34% of world output; mine development lead times are up 41%) can keep up.

โ—† Surprising and New

WuXi AppTec's compound library hit 420,000. This is the Chinese contract research and manufacturing company (known as a CDMO โ€” a company that develops and manufactures drug molecules on behalf of pharmaceutical firms). 420,000 compounds means it has one of the largest catalogues of ready-to-test chemical entities on earth. For Indian CDMOs like Gufic Biosciences, Laurus Labs, and Divi's Laboratories, WuXi's scale is both a threat (it can undercut on price) and an opportunity (Western pharma firms are diversifying supply away from China under "China+1" policies, and India is the obvious second source).
ISM Manufacturing PMI jumped to 55.6. The ISM (Institute for Supply Management) surveys American factory purchasing managers every month. A reading above 50 means the sector is expanding. 55.6 is the strongest print since early 2022. New orders, production, and employment all accelerated. This matters for India because US factory restocking drives demand for Indian auto parts, speciality chemicals, and IT services spending.
Paytm's โ‚น963.6 crore merchant payment block. That is the total amount that was temporarily frozen across Paytm merchant accounts following a regulatory directive. The company's stock dropped on the news, but the block applies to a fraction of its overall payment volume. Still, it signals continued regulatory scrutiny of India's payments infrastructure companies.
Momentum factor's โˆ’17.4% drawdown. Momentum investing means buying stocks that have been going up and shorting stocks that have been going down. When this strategy loses 17.4% in four days, it signals that the market's most popular trades are reversing violently. This kind of "factor crash" happened in 2007, 2009, and 2020. It does not predict a recession, but it does mean that crowded positions โ€” including in AI stocks โ€” are being forcibly unwound.
Meesho targeting โ‚น1,200โ€“1,540 crore IPO. Meesho is India's largest social commerce platform (it lets small shopkeepers resell goods via WhatsApp and other messaging apps). If priced at the top of range, it would be one of the largest Indian consumer-internet IPOs of 2026. The timing โ€” right as US momentum stocks crash โ€” suggests Indian tech founders see a window to list domestically rather than chase US capital.

โ—† Stock & Sector Tracker

CompanySectorSignalSentiment
Ather EnergyEV / AutoFirst +ve EBITDA โ‚น9.45crBullish
TVS MotorAuto6.3L units, +38% growthBullish
Bajaj FinanceNBFCPAT โ‚น5,986cr (+27%)Bullish
Mahindra (EV)EV / AutoLoss โ†’ โ‚น288cr profitBullish
Sterlite TechnologiesTelecom Infraโ‚น960cr optical fibre orderBullish
Glenmark PharmaPharmaEBITDA +38%Bullish
BioconPharma / BiosimilarYesafili US launchBullish
Bhagiradha ChemicalsSpecialty ChemPBT +601% (low base)Bullish
KEI IndustriesCables / InfraPower demand 170.7 BUBullish
Polycab IndiaCables / InfraGrid capex beneficiaryBullish
KEC InternationalT&D / InfraBorder rail + gridBullish
Gufic BiosciencesCDMO / PharmaAt inflection pointBullish
Laurus LabsCDMO / PharmaChina+1 beneficiaryBullish
LICInsuranceOFS up to โ‚น31,000crMixed
MeeshoE-commerceIPO โ‚น1,200-1,540crMixed
PaytmFintechโ‚น963.6cr merchant blockCautious
DLFReal EstateLuxury slowdown signalsCautious
JM FinancialNBFCLeverage risk in drawdownCautious
MakeMyTripTravel TechHigh PE, momentum unwindCautious
Strategy (MicroStrategy)BTC TreasurySelling 3,588 BTC at lossCautious

โ—† Where the Energy Is

These are the themes where the most informed accounts โ€” across all six groups โ€” showed the highest conviction and deepest engagement in the last 24 hours.

EV Penetration Hits Escape Velocity. Ather Energy's first positive EBITDA, TVS Motor's 38% volume surge, and Mahindra's eSUV profitability flip all landed in the same 24-hour window. This is not one company's story โ€” it is the sector crossing the line from "subsidy-dependent experiment" to "self-sustaining business." Battery costs, charging infrastructure, and consumer willingness have all crossed their respective thresholds in India. The next leg is component localisation: who makes the motors, controllers, and battery management systems domestically.
CDMO as India's Next Export Engine. With WuXi at 420,000 compounds and Western pharma actively diversifying away from Chinese manufacturing dependence, Indian CDMOs are the obvious second source. Gufic Biosciences at inflection, Biocon's Yesafili US launch, and Glenmark's 38% EBITDA growth all point the same direction. The constraint is not demand โ€” it is building GMP-certified manufacturing capacity (GMP = Good Manufacturing Practice, the global quality standard for drug factories) fast enough to absorb the redirected orders.
Data Centre Cabling & Grid Overload. India's power grid delivered 170.7 billion units in July, up 10.9%. Every data centre, every EV charger, every factory expansion needs copper cable and optical fibre. Sterlite Technologies just booked โ‚น960 crore of orders. Meanwhile, global copper supply is tightening โ€” Chile's share peaked, mine lead times are up 41%. The companies laying cable into India's grid have pricing power that is getting stronger, not weaker.
Defence Exports and Border Infrastructure. Samudra Manthan (โ‚น84,084 crore for six submarines), border railways (โ‚น45,000 crore), and the broader "Atmanirbhar" defence ecosystem are creating a multi-year order pipeline for companies in shipbuilding, electronics, and heavy engineering. This is government-backed demand with 5-10 year visibility โ€” the kind of certainty that private capex cycles cannot match.
Q1FY27 Earnings Momentum. The median PAT growth of roughly 27% across reporting companies is not a blowout, but it is broad-based. More importantly, the quality of earnings is shifting: revenue growth is being driven by volume (units shipped, power consumed, loans disbursed) rather than price hikes. Volume-driven growth is stickier and compounds better than price-driven growth. The companies that are winning are those selling more of what they make, not charging more for the same thing.

Normal Guy