From Leaders to Laggards: Is Memory Waiting on Micron? Memory gave back Tuesday's gains on Wednesday. SanDisk fell 3.73%, SK Hynix 3.12% and Micron 2.22%. The selling carried on into Thursday's pre-market, then all three bounced on Friday as oil eased after the Trump–Xi meeting. They fell together and bounced together, so this is one trade, not three. The real test is Micron's report after the close on September 30, which is about 4 AM on October 1 in Singapore. Wall Street expects $50.91 billion in revenue and $31.49 in earnings per share. Management guided about 86% gross margin, but also said price increases are moderating. The quarter also has 14 weeks, one more than usual. Bull case: 16 take-or-pay contracts lock in about $100 billion through 2030. HBM4 has already shipped over $1 bil
$ServiceNow(NOW)$ up 14.7% in eight sessions, straight into Wednesday's print ServiceNow closed at $158.96 on 23 Oct, up 14.7% from $138.61 two weeks ago. Q3 results are Wednesday 28 Oct after the close. Analysts drove most of the move. Barclays raised its target to $160 from $134, Wells Fargo to $190 from $175, Piper Sandler to $175 from $140, and Bernstein to a Street-high $266 from $248. RBC set $176 and Mizuho has Outperform at $170. There was also a new NVIDIA partnership on an enterprise AI agent, and AI contract value is above $1bn. The stock is now about 3% above the MarketBeat consensus target of $153.90, and J.P. Morgan's target is $150. The business hasn't changed, and my scorecard is still 73/100. The run-up is the risk. After Q1,
$SpaceX(SPCX)$ SpaceX's First Profit as a Public Company, and Musk's "Trillion-Dollar Revenue" Claim: I Fact-Checked the Viral Video and Ran It Through My Sell-Put Gates Written Wednesday 21 Oct, SGT, after Tuesday's report. A video about SpaceX's results is everywhere right now. Some of it is accurate, some isn't, and the most important numbers aren't in it at all. Here's what the company actually reported. The quarter in numbers - Revenue: $14.6 billion, up 161% year over year, against about $12.9 billion expected - Operating profit: $2.79 billion, after a $143 million operating loss in Q2 - Net income: $4.4 billion, or 35 cents a share against a 12-cent estimate. A year earlier it lost $1.1 billion. - Adjusted EBITDA: $8.8 billion One importan
Goldman Upgraded CoreWeave. Nebius Got the Bounce. Why the Neocloud Split Is About Balance Sheets, Not Demand Written Tuesday 20 Oct, SGT, using Monday's US closes. What happened on Monday Goldman Sachs analyst Gabriela Borges upgraded CoreWeave to Buy from Neutral. Usually that's enough to move a stock that's down 25% in a week. It wasn't. - CoreWeave: up about 2.2% intraday to $62.74, then reversed to close down 0.44% at $61.12 - Nebius: closed up 2.88% at $170.24, after rising as much as 4.64% intraday on a reiterated Truist Buy rating with a $355 target That's a gap of about 3.3 percentage points in one session between two companies that, a week earlier, traded as the same stock. Last week CoreWeave fell 25.21% and Nebius fell 25.15%, almost exactly the same. So the split started on Mo
Stocks to Watch Today (20 Oct): One Green Day After the Worst Week Since June. Here's What I'm Watching and Why I'm Not Chasing It Written Tuesday morning SGT, using Monday's US closes. What happened on Monday After last week's AI infrastructure selloff, tech finally bounced. - Nasdaq: +1.11% to 26,430.89 - S&P 500: +0.41% to 7,690.73 - Dow: -0.07% to 51,966.03 - Philadelphia Semiconductor Index: +1.61% to 11,316.34, with 26 of 30 members higher Memory stocks, last week's biggest losers, led the rebound: - Western Digital +6.09% - SanDisk +5.53% - SK Hynix +4.67% - Micron +2.97% to $904.90 Bank of America said memory valuations have bottomed, with 2027 to 2028 consensus P/E ratios at just 2x to 3x. The other notable moves: - Amazon +4.95%. Morgan Stanley named it a top pick and raised
$CoreWeave Fell 12%, But the Bond Market Said It First: Will the Credit Warning Spread? A Put Seller's Read Written Monday 19 Oct, SGT, using Friday's US closes. Most people saw the headline: CoreWeave down 11.91% on Friday to $61.39. But the stock was the last thing to move. The bond market had been warning about this for weeks. What actually happened - Stock: down 25.21% for the week, from $82.08 to $61.39, just above its 52-week low of $59.51. It is down 55.91% over the past year. - Bonds: the October 2031 notes briefly yielded more than 16% on Thursday, and some CoreWeave bonds traded near 79 cents on the dollar. - Credit default swaps: the five-year CDS hit an all-time intraday high of 1,194.5 basis points and closed Friday at 1,117. Bloomberg data puts the implied probability of defa
$CloudBigThree AWS vs Azure vs Google Cloud: three hyperscalers, one earnings week, and why I'm not selling puts on any of them yet Microsoft and Alphabet both report on Wednesday 28 October after the US close. Amazon follows on Thursday 29 October. The Fed decision also lands on the 28th. For those of us in Singapore, the numbers hit our screens between roughly 4am and 6am on the 29th and 30th. AMZN is a core name in my sell-put book, and MSFT and GOOGL sit in my second tier. Instead of reviewing each one alone, I put the three side by side this time. Same business at the centre (renting cloud and AI compute), three very different stocks. All prices below are Friday 16 October closes. Cloud numbers are for the quarter ended 30 June, which all three reported in late July. 1. THE CLOUD SCOR
$Meta Platforms, Inc.(META)$ Meta: up 23% since late August, results on Wednesday, and why puts don't work at this share price Meta closed at $709.33 on 16 Oct, up 23% from $576.14 on 27 Aug and 8.8% below its 24 Sep peak of $777.59. Q3 results come out Wednesday 21 Oct after the close, confirmed by the company. The move started with Muse, Meta's personal AI agent, which has a free tier and paid plans at $20 and $100 a month. It hit a record 264,000 US downloads on 19 Sep. On 21 Sep the stock jumped 11.4% on about 2.6× normal volume, after Wells Fargo raised its target to $796 from $640. On 16 Oct it fell 3.7% with the rest of mega-cap tech on inflation and AI-financing worries. The business is strong, but spending is absorbing the growth.
$NVIDIA(NVDA)$ NVIDIA: down 9.8% in seven sessions, and why I'm still not selling puts NVIDIA closed at $214.13 on 16 Oct, down 9.8% from $237.47 on 8 Oct and 12.1% below its record of $243.57. Nothing changed in the company. Rates and a funding scare moved the whole chip sector. On 14 Oct, September CPI showed inflation at 3.7%, above expectations. The 10-year yield rose more than 7 basis points, and the market-implied chance of a Fed hike on 28 Oct jumped from 22% to 44%. The Philadelphia semiconductor index fell 2.8% and NVIDIA 3.2%. The same day CoreWeave fell 15–16% after a chart of its roughly 15% bond yields went viral, which fed worries about how the AI build-out is financed. NVIDIA fell another 1.8% on 15 Oct and 1.4% on 16 Oct. On
Memory Crushed, AI Chips Rallied. Is Semiconductor Money Rotating Internally? The Answer Came Three Days Later Written Saturday 17 Oct, SGT, using Friday's US closes. First, check the date on the numbers The figures in this topic are real, but they come from Tuesday, Oct 13. That day Nvidia closed up 0.97% at $228.52, AMD rose 1.90% to $605.53 and Intel gained 1.44% to $102.81. Micron fell 3.62% to $953.19, its fourth straight loss, after China's CXMT said it plans to launch a 4F2 DDR5 server memory module by year-end. On that one session, it did look like rotation: money leaving memory and moving into compute. Now look at what happened over the next three sessions. The scoreboard: Tuesday close to Friday close - Nvidia: $228.52 to $214.13, -6.30% - Micron: $953.19 to $878.84, -7.80% - Bro
CXMT Targets Server Memory: Is the Memory Repricing Done? Here's What the Numbers Say, and Why I'm Not Selling Puts Yet Written Friday 16 Oct, SGT, after Thursday's US session. Last weekend in Shanghai, CXMT president Cao Kanyu said China's largest DRAM maker plans to launch a DDR5 RDIMM server module built on its new 4F² cell design by the end of 2026. A 4F² cell takes up roughly a third less area than the 6F² design the industry still uses, so you get more memory per chip. CXMT says it holds more than 300 patent families on the architecture. That's the headline. Here's what wasn't said: no speed, no capacity, no yield and no volume. It's a product launch target, not a commitment to mass production or to qualification at the big server platforms. The market didn't wait for the details. Fr