Welcome back to a new edition.
Before the stories, one question since this the Apple launch week:
If no iPhones sold at all next year, how much of Apple would still be standing?
That is what Tim Cook spent his years as chief executive building, and what John Ternus inherited on 1 September. Seven minutes of our video will change how you read our first story, watch it now: Tim Cook’s Apple Legacy: From Product Company to $4 Trillion Ecosystem at Global Markets by Vested.
Apple launched its first ₹3 Lakh folding phone and gained $164B the next day, then found something almost nobody is talking about.
Also, OpenAI said its newest AI can do virtually anything a human can. An independent lab tested the same AI and found it less likely.
And back here in India, the finance minister asked the RBI to push the digital rupee harder, in a year when its circulation shrank.
We’ll get to the interesting bits.
But before that, we built an interesting challenge to feed your competitive spirit.
Each week, we bring out a new edition of the “Global Investing Challenge.”
Think you can guess the stock behind this week’s wild stories?
Click (Quiz) or on the image below to take the challenge and prove that you are a smarter global investor than most.
Coming back to the week’s updates, let’s start with how the global markets fared this week.
The World in a Week
All eight rows fell, and oil is why most of them did.
Fighting near the Strait of Hormuz pushed crude above $100 a barrel. Expensive oil makes everything else cost more, so it arrives as inflation. US price data on Thursday and Friday was not alarming but not reassuring either, and most traders now expect the Federal Reserve to raise rates next week. Europe raised this week, Japan is expected to follow.
Higher rates hit gold, down 1.8%, and Bitcoin, down 1.5%, directly. Neither pays you anything to hold, so when cash and bonds start paying more, people sell.
The S&P 500 (▼0.8%) and Nasdaq (▼0.7%) fell four days straight, then bounced on Friday as oil eased. The FTSE 100 (▼1.7%) got no rescue from its oil majors, because the same inflation pushed a Bank of England rate cut further out of reach. The Nikkei 225 (▼1.6%) and SSEC (▼1.1%) both closed hours before Friday’s rebound, so neither caught it.
India had the hardest week. The Nifty 50 fell 2.1%, a fifth consecutive losing week. India buys most of its oil abroad, so a crude spike lands as a bigger import bill. Foreign investors sold again, and the rupee had its worst week since May.
Now, coming to this week’s stories.
News Stories
The untold story of Apple’s ₹3 Lakh iPhone Duo
If you have been living under the rocks, Apple (AAPL) launched its first folding phone iPhone Duo on 9th September, priced at ₹2,99,900 to ₹4,49,900 in India. That’s some ₹1.35 Lakh more pricey than iPhone 18 Pro.
And soon it was meme material –
It is also the first launch Apple has run without Tim Cook in charge, since John Ternus took over as chief executive on 1 September. Ternus had this to say on the launch –
“iPhone Duo is the most transformational change to iPhone since the original.”
Investors agreed fast. On 10th September, the first full session after the launch, Apple rose 3.56 percent to close at $326.57. That added about $164 billion to what the company is worth, taking it to roughly $4.77 trillion.
Source: AAPL on Vested
Now hold that $164 billion next to the size of the prize. Counterpoint Research, which tracks phone shipments, reckons folding phones are about 2% of all smartphones sold, and expects Apple to ship roughly 6 million this year, second behind Samsung.
Six million phones at $1,999 each is about $12 billion of revenue. The market added more than thirteen times that in one session, for a phone nobody has received.
Gene Munster of Deepwater Asset Management, who has covered Apple for most of his career, called it the first product since AirPods that people genuinely want. If he is right, $164 billion could likely be justified.
Make of it what you will.
Now here is the part almost nobody wrote about, and it is the bit that matters for your money.
The companies that have to build this phone did not celebrate. In South Korea on 10 September, BH, which makes the bendable circuit boards behind the screen, fell 3.41%. And LG Innotek, which supplies camera modules, rose 0.72 percent.
Korean analysts reported Apple had asked suppliers for prices 30% below existing levels.
That is the Apple everyone knows, ordering in huge volume, naming its price, and suppliers accepting because losing Apple is worse.
Except this may be the most Korea-dependent iPhone Apple has built, and on the expensive parts that arrangement is breaking down. It had to hand Samsung Display an exclusive deal to make the folding screen for up to three years.
On memory chips, the parts that store your photos and apps, the only real suppliers on earth are Samsung, SK hynix and Micron. There is no fourth name to threaten them with.
An industry official, speaking to Asia News Network, said –
“Apple usually uses its enormous purchasing power to push component prices down. With this product, that formula may work less well.”
You can see it in the cost.
For instance, on the one-terabyte iPhone 18 Pro Max, memory alone is estimated to have added nearly $300 to the bill of materials, almost a sixth of the price.
Anyways, here is the part worth sitting with.
For twenty years, Apple’s supply chain story was Apple holding the whip. The AI boom has quietly handed memory makers the pricing power instead, because every data centre on the planet now wants the same chips your phone needs.
Keep those three names in mind, because they turn up again in the next story.
OpenAI’s AGI statement is likely a stretch
On 3 September, OpenAI released GPT-6 Astra, and its president Greg Brockman said the line that set off the week.
“Welcome to the AGI era.”
Asked directly whether Astra was AGI arriving, he was looser: “I think it might be about this model.” OpenAI put the phrase in the room without formally claiming the milestone, then let everyone else argue.
For starters, AGI means Artificial General Intelligence, and OpenAI’s own version is a system that can do all economically valuable work as well as or better than humans. That is the bar it set itself.
In fact, Astra is genuinely strong. OpenAI’s scorecard puts it at 97.6 percent on FrontierMath, a set of research-level maths problems, and OpenAI helped pay to build that test.
Then came the number that matters, from ARC Prize, an independent body that runs its own intelligence tests.
Astra scored 99.9% on the ARC-AGI-3 test when run through OpenAI’s own setup. The same model scored 62.7% on the same test run through a neutral one. OpenAI’s scorecard prints the 99.9% and does not mention the 62.7%.
The difference is the harness, the scaffolding around a model while it sits the test. The neutral one is identical for every company and lets the model keep only the notes it writes down. OpenAI’s lets it hold on to its internal working and reuse it.
Stranger still, the 99.9 run used about $19,000 of computing time and the 62.7 run used about $26,000. Which means more money, worse score.
Source: ARC Prize
ARC Prize was careful about what that proves:
“we are not claiming that it is AGI”
Here is why a definition argument belongs in a markets newsletter.
For almost seven years one contract put a number on AGI. Microsoft’s (MSFT) rights to OpenAI’s technology were written to end once AGI arrived, and by late 2024 the trigger was financial: AGI would be declared when OpenAI could generate roughly $100 billion of profit for its earliest backers. On 27 April 2026, that clause was deleted.
So the only contractual definition of AGI that ever existed was removed four months before OpenAI announced the AGI era had begun.
Now if you follow the money to these AI giants, you get more interesting insights.
For instance, look at who funded Anthropic’s May round. Alongside the usual investment funds, it lists Samsung, SK hynix and Micron as strategic infrastructure partners. Those are the three memory makers Apple can no longer lean on, selling chips into the AI build-out and buying a stake in the customer at the same time.
Anyways, here is the part worth sitting with.
The capability is real and moving fast. What is missing is an agreed way to measure it, and without one the story becomes whatever the best-funded party says.
One capability is not in dispute. Astra scored 100% on a test of finding and exploiting software flaws, and OpenAI restricted those features at launch.
On a side note, Anthropic’s threat report this month found attackers using AI to finish break-ins in hours that once took weeks. Hold that, because it surfaced in a speech here in India.
India’s shrinking digital rupee needs a big push
At the Global Fintech Fest in Mumbai on 11 September, finance minister Nirmala Sitharaman spent part of her address on exactly that problem:
“A leading lab, I’m not naming here, now judges autonomous cyber capability to be moving faster than its existing controlling mechanisms. The controls that they put in place are not catching up.”
She did not say which lab, and the description matches what Anthropic published this month almost exactly, iykyk 🙂
But the part that matters most for your money was a pilot that ran on Monday.
On 7 September, REC Limited, a state-owned lender to the power sector, completed India’s first tokenised corporate bond issue inside the market regulator’s testing sandbox.
It raised ₹500 crore, paying 7.30% a year for one year and nine months, and investors bid ₹796 crore for it. Payment, allotment and listing all finished the same day.
For starters, tokenised means this. Normally the bond moves on one system, the cash moves on another, and someone matches them up afterwards. Here, both sat on the same shared ledger and moved together in one step, so neither could travel without the other.
The money side of it was the digital rupee, which is ordinary rupees issued by the RBI itself in digital form rather than as notes. The FM framed why that matters quite well –
“Every tokenization architecture being tested across the world eventually returns to the foundational question: What is the money leg made of?”
She then asked the RBI to push its digital rupee pilots further and sharpen what the currency can do.
Which brings us to the awkward number.
Source: RBI Annual Report (2025-26)
So, the digital rupee in circulation shrank 24% in the year the RBI was expanding its pilots.
Usage is growing, though. In July, RBI governor Sanjay Malhotra put the pilot at 12 million users and more than 175 million transactions worth about ₹40,000 crore since launch.
Now set that against UPI. In August 2026 alone, UPI handled 24.51 billion transactions (or 791 million a day) worth ₹29.82 trillion.
Source: NPCI
Meaning –
UPI does more transactions in one day than the digital rupee has done in its entire existence.
Anyways, here is the part worth understanding, and it is not the obvious conclusion.
The digital rupee is not failing at being UPI, it was never trying to be. UPI already moves money between people and shops brilliantly. What a digital rupee can do, and UPI cannot, is settle instantly and finally against an asset on the same ledger, which is what the REC bond needed.
So a small circulation number is not proof this is going nowhere. It says the real use is institutional, and that use only started working this week.
One thing follows for you directly: this was an institutional pilot, and retail investors cannot buy tokenised bonds through normal channels yet.
The bottom line
All three announcements had a similar pattern.
Apple added about $164 billion in market value for roughly 6 million phones that ship from 23 October. Gene Munster says it could be justified. But the Korean suppliers who have to build the phone did not celebrate, and the memory makers who sell chips to both Apple and the AI labs are the ones quietly gaining pricing power.
OpenAI said the AGI era has begun. An independent lab scored the same model at 62.7 on a neutral test against 99.9 on OpenAI’s own setup. The only contractual definition of AGI that ever existed was deleted four months before that announcement. The capability is real, but not the measurement.
And India’s finance minister asked the RBI to push the digital rupee harder, days after the first tokenised bond settled in it beautifully. The digital rupee in circulation shrank 24% last year. UPI does more transactions in one day than the digital rupee has done since launch. But the digital rupee is settling an asset and its cash in one step.
In each case, the claim arrived before the scale. That is not unusual.
What is unusual is that the same three companies, Samsung, SK hynix and Micron, keep turning up on the winning side of all three stories.
Either the claims catch up to the prices, or they do not… only the future will tell.
That is a story for a future Vested Shorts. We will be here when it drops.
