Memory prices still look supported, but the market is no longer a clean one-way trade: the bull case is that tight supply and AI-driven demand keep pricing elevated into 2027, while the bear case is that the sector has already priced in a lot of good news and can correct sharply on any sign of slower pricing or capex normalization. SanDisk’s recent pullback looks more like profit-taking after an extreme run than a fresh demand collapse, but it also shows how fragile sentiment is in this group.SanDisk fell about 3.5% recently after a huge rally, with commentary pointing to broad profit-taking across memory and storage rather than company-specific weakness. The sector has been volatile because investors are debating whether AI demand can keep overwhelming supply long enough to justify today’

For me, 50% is already a very meaningful contribution, and I think AI can maintain a large share of S&P 500 $S&P 500(.SPX)$ earnings growth. The AI cycle is no longer just about chips. It is spreading into cloud, data centres, networking, software and productivity gains. That said, I would not expect AI spending to grow at this pace forever. Oracle shows both sides of the story: huge future demand, but also massive CapEx and cash flow pressure. The market will increasingly reward companies that can turn AI demand into real earnings and cash flow. I remain bullish on AI long term, but I prefer selective accumulation rather than chasing. For me, the next phase is not about who spends the most, but who can turn that spending into sustainab

I think the 50% share can hold in the near term, but it is becoming a much higher bar to clear. Goldman itself expects AI infrastructure beneficiaries to drive roughly half of S&P 500 earnings growth, while Q2 data showed AI infrastructure already contributing about one-third of EPS growth. The key question is no longer whether companies will spend on AI—they clearly are. It is whether that spending converts into recurring revenue and margins. If hyperscalers keep expanding capex, the suppliers can continue winning. But if financing costs rise or ROI disappoints, the earnings concentration becomes the market's biggest vulnerability. So my view: 50% can persist, but it probably cannot keep rising indefinitely. AI remains the engine; valuation and cash-flow discipline decide how far the

Exactly. All the overnight drunk sellers. Not like everyone doesn’t need ChatGPT anymore and stopped needing tokens.

On Friday the August CPI report landed, traders took the odds of a 25 basis point hike this week from 75 per cent to close to nine in ten, and $S&P 500(.SPX)$ closed 0.86 per cent higher all the same, ending a four-session slide; $Dow Jones(.DJI)$ closed 0.98 per cent higher, a gain of more than 500 points. The bet on higher rates got bigger. The buyers came back. August CPI rose 3.4 per cent year on year, level with July and in line with expectations; month on month it rose 0.4 per cent against 0.1 per cent in July. The gasoline index rose 3.9 per cent on the month and accounted for a third of the entire rise in goods prices. Core CPI eased to 2.4 per cent year on year

My pick is ①: Inference / AI Agents. Selling off Asian chip stocks just because AI leaders said "slow down" is a knee-jerk overreaction. Slowing down frontier training doesn't mean stopping construction. Foundation models are already big enough. The money is shifting from chasing model benchmarks to monetizing inference and agents. High-frequency inference consumes HBM and DRAM even more persistently. The memory logic is intact. ASICs and networking will benefit as demand diversifies. Today's broad sell-off is a textbook headline-driven panic. Don't hand over your core chips. Let Micron's Sept 30 earnings show the real demand numbers. I'm watching for dips, not joining the bears.

Memory prices still look supported, but the market is no longer a clean one-way trade: the bull case is that tight supply and AI-driven demand keep pricing elevated into 2027, while the bear case is that the sector has already priced in a lot of good news and can correct sharply on any sign of slower pricing or capex normalization. SanDisk’s recent pullback looks more like profit-taking after an extreme run than a fresh demand collapse, but it also shows how fragile sentiment is in this group.SanDisk fell about 3.5% recently after a huge rally, with commentary pointing to broad profit-taking across memory and storage rather than company-specific weakness. The sector has been volatile because investors are debating whether AI demand can keep overwhelming supply long enough to justify today’

I’d pick ① Inference / AI Agents. The biggest mistake is treating slower frontier-model development as the same thing as weaker AI demand. These are two different cycles. If model training slows, the next growth engine could be inference: AI agents, enterprise automation, search, coding and customer service. Once millions of businesses start using AI continuously, compute demand becomes recurring rather than concentrated in giant training runs. That shift could also benefit ASICs, networking, HBM and data-center infrastructure, especially where cost, latency and power efficiency matter. So I wouldn’t call today’s selloff the end of the AI cycle. I’d call it a rotation from “build smarter models” to “use AI everywhere.” The real warning sign would be falling cloud capex, weakening GPU util

As hyperscale technology conglomerates — such as Microsoft, $Alphabet(GOOGL)$ Alphabet, $Amazon.com(AMZN)$ Amazon, and Meta—signal potential moderation or pacing adjustments in their capital expenditure (capex) growth rates, public markets have begun pricing in an inevitable AI infrastructure "reality check." In this article, we will be looking at how to navigate the AI Capex deceleration, as deceleration in hyperscale AI capex does not mark the end of the semiconductor super-cycle; 1. The Capex Reality Check: Why Hyperscalers Are Recalibrating AI Infrastructure Spend The primary catalyst behind market anxiety regarding semiconductor stocks is the widening gap between hyperscaler infrastructure capex and

$NVDA$   Enflame, which is backed by tech giant Tencent, is considered one of China's so-called "four little dragons" of AI chipmaking and is the last of that group to go public. The other three all surged on listing, and have remained higher since. In December, MetaX soared nearly 700% on its first day of trading, while Moore Threads gained over 400% on its trading debut. Biren jumped 76% on its IPO in January. Investors are betting on the ability of domestic AI chipmakers to replace Nvidia in China. International chipmakers led by Nvidia accounted for nearly 60% of China's AI accelerator market in 2025, according to IDC data cited in Enflame's prospectus, translated by CNBC. Nvidia

🌟🌟🌟The market is currently walking a tightrope & Friday's looming inflation data is the ultimate wild card.  If the inflation numbers come in hot, the broader market will likely face a sharp interest rate panic.  This may hit high beta technology & semiconductor stocks the hardest. The storage sector is already showing signs of a fierce tug of war.  For example, while Wall Street remains euphoric, the CEO of flash memory giant $KIOXIA HLDGS CORP(KXIAY)$ openly warned that prices have risen high enough which may cap short term upside. My strategy to play the storage surge: I would let the inflation data digest over the weekend so that I

For me, a US$638 billion backlog is impressive—but a backlog is only a promise until it becomes revenue, cash flow and ultimately free cash flow. Tonight, I would focus on three things: RPO conversion, AI revenue growth, and cash generation. If Oracle can show that major AI contracts are moving into actual revenue faster than expected, while keeping margins under control, the backlog starts to look like a genuine earnings engine rather than a headline number. The bigger question is capex. Oracle is spending heavily to build AI infrastructure before customers fully pay for it. That creates a dangerous gap if financing costs stay high. So I would not buy simply because the backlog is huge. I want evidence that AI demand is converting into cash faster than Oracle is converting cash into data

For me, the key is whether $Oracle(ORCL)$ can turn its huge AI backlog into real revenue and cash flow. A $638 billion backlog sounds impressive, but it means little if execution cannot keep up with the capital spending required. I would watch cloud growth, AI demand, contract wins and especially free cash flow. If Oracle shows that AI investments are starting to generate stronger cash returns, I would be more comfortable investing behind the backlog. I also want to see whether management can maintain strong growth without continuously increasing its spending burden. I remain bullish on AI infrastructure long term, but I do not want to chase the story based on backlog alone. I want the numbers to prove it first. If the results are strong, I would

On Wednesday $S&P 500(.SPX)$ closed 0.48 per cent lower, its third session in a row heading down. $Meta Platforms, Inc.(META)$ rose 6.55 per cent the same day to close at US$653.69. The whole market was backing away while one of the largest companies in it put on six and a half points, and those look like two unrelated things. They are one thing. What is holding the indexes down is rates: Brent crude moved back above US$100 for the first time in two months, the inflation print lands on Friday, the ten-year Treasury yield was pushed to 4.85 per cent and the thirty-year broke above 5.30 per cent. The Treasury stepped in that day and raised the cap on its buybacks of ten-

Why did Nvidia slide in the last two days?