On Wednesday the market split the AI buyers into two groups: those paying from their own cash, and those borrowing. By Thursday evening that line was gone. The one big buyer that funded itself sold 25 billion USD of bonds. Then Seoul sold the suppliers that get paid in advance.
What happened
- Microsoft (MSFT) priced a $25 billion bond in 10 parts, from 2 to 40 years. It is its first US investment-grade bond sale since 2017, one day after reporting $18.6 billion of positive free cash flow.
- Alphabet (GOOG) fell 12.6% on Thursday, its worst day as a public company, after negative free cash flow funded by new bonds and share sales.
- Amazon (AMZN): cloud growth accelerated to +39% with a $578 billion backlog, and management expects over 40% growth in 2027 and 2028. But spending reached $77 billion in one quarter, trailing free cash flow turned negative $42 billion, and 2027 spending will be “meaningfully more”. Shares fell 7-12% after hours.
- Intel guided below expectations on weak PC demand.
- Friday in Seoul: the KOSPI fell 3.5%, SK hynix 7%, Samsung 3%. Foreign investors sold about 1.6 trillion won.
- Bonds stayed calm. The US 10-year yield fell to 4.98% on a day with 3.4% GDP growth and $25 billion of new long-dated corporate paper.
Why it matters
All four large cloud buyers now use outside capital. That changes what to watch. Their 2027 spending plans are all going up, so supplier demand has a floor. But that floor is now partly financed by the bond market. The new gauge is not the order book. It is whether investors keep lending on good terms.
There is a difference inside the borrowers. Microsoft borrows from strength: top credit rating, positive cash flow. Alphabet and Amazon borrow to fill a cash gap. I rank them differently.
And Seoul sold the wrong layer. SK hynix (SKHY) is paid with customer deposits and says it is sold out through 2029. It fell more than Samsung on its customers’ borrowing news. That is positioning, not a change in memory contracts.
What changed in the portfolio
- Microsoft: both clips closed by trailing stops near 530 and 521, about +40% each. Same day as the bond. Rules, not prediction.
- Memory: the staggered stop ladder sold its top steps. Two Micron (MU) clips and one memory fund clip closed in small profit with near-zero slippage. A Korea fund clip filled 1.5% under its stop at the Frankfurt open, the known gap risk.
- Treasuries: a second small iShares 7-10 Year Treasury (IEF) clip was added at the US open. On its first day, bonds rose while tech fell.
- Alibaba and BYD clips were reset at the same size, lower base.
- Cash is now about 16.5% of the portfolio. AI hardware is about 17%, the lowest in two months.
What I watch next
- Today’s US open: Korea-linked funds will reprice Seoul’s fall. Some memory stops may fill in the opening auction, where slippage is usually worst.
- Samsung: the one pre-written add stays conditional on a calm London close. No chasing.
- Taiwan Semiconductor (TSM) and ASML (ASML) are below their weekly lines. No adds this week.
- Amazon failed my funding test. It is not this week’s AI add.
- Next week: Apple, AMD, the US midterms, Lumentum (LITE) results and possible Marvell (MRVL) deal terms.
Risks, stated openly
- If a big tech bond deal is cut or prices badly, spending guides could follow, and suppliers would feel it.
- Record results keep selling off. Positioning can beat fundamentals for weeks.
- Stops protect gains but can sell into a gap.
- Futures still price a likely December rate hike. A hot inflation print hits growth stocks and bonds together.
The lesson
Who pays for AI is still the right question. This week added a second one: who lends to the payer, and on what terms? The answer is calm today. The portfolio enters the next test with lower AI exposure, more dry powder and a duration hedge that just worked.
Patience is a position too.
Snapshot: portfolio data as of 08:17 UTC, Friday 30 October 2026 (US prices = Thursday closes plus after-hours quotes).
Sources: Amazon Q3 2026 results · Microsoft prices $25B bond sale · Seoul Economic Daily: KOSPI −3.5% · Closing Look, 29 Oct 2026
This is not financial advice. Past performance does not guarantee future results, and copy trading carries a real risk of losing money. Always do your own research.