Hike Odds Near Nine in Ten: Is the Market Right to Look Past It?

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 from 2.5 per cent, a third straight decline and the lowest reading since April 2021. What got dearer was energy. The sticky part came down. Oil fell back from its high for the week on Friday, letting go of the force that had held prices up the day before.

Where the rebound money went was plain enough on the day. $Oracle(ORCL)$ closed 1.74 per cent lower at US$150.28, down again in its second session after results. Yet the capital expenditure guidance it gave on its earnings call for financial year 2027 is US$90 billion to US$95 billion, and its chief financial officer, Hilary Maxson, named the rack, cooling and networking suppliers that money goes to: Hewlett Packard Enterprise and Dell. $HP Inc(HPQ)$ closed about 12 per cent higher at US$62.09, its highest close since it began trading in November 2015; $Dell Technologies Inc.(DELL)$ rose about 11.5 per cent to finish above US$567, also a record.

The same money is an outlay on Oracle's books and an order on theirs. Friday ran that split cleanly: the side fronting the cash was sold and the side collecting it was bought. US$90 billion to US$95 billion is a year of guided spending rather than purchase contracts already signed, and nothing says how much of it ends up with these two. Where the money comes from is a separate matter: Oracle says its net cash outlay comes to no more than US$70 billion once customer prepayments and financing arrangements are taken out, which is to say that part of this buying cannot move until the prepayments and the outside funding are lined up. The repricing that a name-check brings lands the same day. The orders land a quarter at a time.

That line does not stretch to every seller. The memory names did not keep up on Friday: $SanDisk Corp.(SNDK)$ closed 3.50 per cent lower at US$1,633.35 while the technology hardware sector rose 1.46 per cent, so it fell against its own sector, touching US$1,616.80 during the session on about 9.3 million shares, more than 40 per cent below the previous session's volume, and no named piece of bad news of its own can be found anywhere in public reporting; $Micron Technology(MU)$ closed 0.22 per cent lower at US$975.26; $SK hynix(SKHY)$ closed 0.94 per cent higher at US$190.07. What got named was the server builders, not every tier of supplier above them.

Chip shares closed higher across the board: $Marvell Technology(MRVL)$ rose 4.03 per cent to US$236.10; $Intel(INTC)$ rose 2.61 per cent to US$102.94, taking back the US$100 mark it had lost the session before; $Advanced Micro Devices(AMD)$ rose 2.49 per cent to US$516.13; $NVIDIA(NVDA)$ closed 0.03 per cent lower at US$218.29.

The large caps that do not live off the pace of model releases rose as well: $Apple(AAPL)$ closed 1.75 per cent higher at US$332.27; $Alphabet(GOOG)$ closed 1.53 per cent higher at US$335.45; $Tesla Motors(TSLA)$ closed 0.52 per cent higher at US$365.44. The case put for Alphabet is that Gemini Enterprise is already used by close to 90 per cent of the Fortune 100, alongside a partnership with Accenture. Broadcom managed only 0.32 per cent, closing at US$361.99 and trailing the rest of the group; cybersecurity names as a group rose only 0.44 per cent, and $CrowdStrike Holdings, Inc.(CRWD)$ closed 1.02 per cent lower. Goldman Sachs holds up a different measure: AI is delivering half the earnings growth of the S&P 500.

$Oracle(ORCL)$ also disclosed a share sale plan on Friday: its founder, Larry Ellison, was to sell up to 50 million shares under a 10b5-1 arrangement, about US$7.5 billion at that day's close, running to 24 October. The plan was cancelled on Saturday, with the company confirming that no shares were sold and that no other selling arrangement is in place. Cancelling it changes the insider signal. It does not change the arithmetic: turning US$664 billion of remaining performance obligations into money means building the capacity before any of it arrives, and the capital budget already runs at more than twice the cash the business itself generates. Oracle is down about 22 per cent this year.

The weekend brought something else. Dario Amodei, the chief executive of Anthropic, publicly called for slowing the push on frontier model capability; Sam Altman of OpenAI and Elon Musk of Tesla each came out in support, and all three back bringing in independent safety evaluation; Altman also said OpenAI will not list this year. Nvidia is reported to be in talks to put as much as US$10 billion into Anthropic when it lists. No party has announced a halt to training or a cut to capital spending. All of it came after the close, and none of it is in Friday's prices.

The next gate is Wednesday, when the rate decision is published at 2am Beijing time on 17 September. The coverage that carries the Fed whisperer label has the central bank all but certain to raise this week, and once a cycle like that opens it rarely stops at one. What Friday actually priced was whose books a given piece of capital spending lands on: the side carrying it as an outlay fell again, and the two carrying it as orders both set records. If rates do move up another notch, the side paying up front is the first to get repriced.

The above is personal analysis, not investment advice.

💬 【Talking Point】

Goldman Sachs puts AI at half the earnings growth of $S&P 500(.SPX)$. Do you think a share that size can hold?

💰 【Bounty】

Drop your view in the comments and there are Tiger Coins in it for you! 🎁

🔔 Better shared than saved — tag a friend and split the coins!

# AI Giants Call for Slowdown — Can Chip Stocks Hold?

Disclaimer: Investing carries risk. This is not financial advice. The above content should not be regarded as an offer, recommendation, or solicitation on acquiring or disposing of any financial products, any associated discussions, comments, or posts by author or other users should not be considered as such either. It is solely for general information purpose only, which does not consider your own investment objectives, financial situations or needs. TTM assumes no responsibility or warranty for the accuracy and completeness of the information, investors should do their own research and may seek professional advice before investing.

Report

Comment3

  • Top
  • Latest
  • Shyon
    ·19:31
    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 sustainable profits.

    @Tiger_comments @TigerStars @TigerClub @Marktomarket

    Reply
    Report
  • 苏36
    ·17:58
    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 engine can take the market.

    @Marktomarket [捂嘴]

    Reply
    Report
  • Jerry Lam
    ·17:43
    我觉得 “AI贡献标普500一半盈利增长”短期还能维持,但很难长期一直维持这么高的占比。

    原因是现在AI资本开支还处在高强度扩张阶段,NVDA、AVGO、服务器、网络、云基础设施这些公司,确实能把AI需求快速转成收入和利润,所以AI对指数盈利的拉动会非常集中、非常显眼。

    但后面市场一定会开始追问第二层问题:这些利润到底来自“别人继续砸CapEx”,还是来自AI本身创造了足够多的新收入和现金流。 如果只是云厂商不断加预算,而应用端变现跟不上,那高增长很难无限延续。

    所以我更看好的是两类公司:一类是 先收钱的卖铲人,另一类是真正能把AI变成订阅、广告、云收入和Agent付费的应用平台。真正危险的是那些 先重资产投入、现金流却长期跟不上 的公司。

    一句话:AI还能继续贡献很大一部分盈利增长,但下一阶段市场会从“谁花钱最多”转向“谁最先把AI变成真钱”。

    Reply
    Report