TSMC Beat. The Dow Rallied. AI Hardware Still Fell.

The market: hotter producer prices moved money out of technology and into value sectors

$SPDR S&P 500 ETF Trust(SPY)$ rose 0.28 per cent to US$771.79 on Thursday, while $Invesco QQQ(QQQ)$ fell 0.13 per cent to US$728.00. September PPI increased 1.3 per cent from the previous month, above the market's 0.6 per cent expectation and faster than August's 0.4 per cent rise. Wholesale inflation reached 6.3 per cent from a year earlier, with energy prices the main driver. Renewed upstream price pressure reduced expectations for rate cuts and highly valued technology shares.

The broad market did not fall with technology because money moved towards financials, consumer companies and other non-technology sectors, giving the Dow its best performance in more than a month. This was not a broad recovery in risk appetite; leadership inside the indices changed. Greater breadth can reduce the market's dependence on a small group of AI leaders, but higher upstream costs can also compress margins across more industries.

TSMC posted record profit while Nvidia and AMD fell with AI hardware

$Taiwan Semiconductor Manufacturing(TSM)$ fell 1.80 per cent to US$422.25, Nvidia fell 1.84 per cent to US$217.23, AMD fell 1.50 per cent to US$562.23 and Intel fell 2.80 per cent to US$95.37. Investors continued to reduce AI hardware exposure after TSMC beat expectations, showing that the question has moved beyond whether orders are strong enough to whether the expansion cycle can produce sufficient cash returns.

$Taiwan Semiconductor Manufacturing(TSM)$ reported quarterly revenue of TWD1.49 trillion, above the TWD1.47 trillion expected, and net profit of TWD763.1 billion, ahead of the TWD746.34 billion estimate. Its operating margin of 58.7 per cent also exceeded the top of company guidance at 58 per cent. Fourth-quarter revenue guidance of US$52.3 billion to US$54.3 billion, volume production at 2 nanometres and continued advanced-node demand all support growth. The pressure sits on the other side: the company expects capital expenditure from 2026 to 2028 to be substantially higher than during the previous three years. The results proved that demand remains strong, but did not answer when customer financing and new investment will produce an adequate return.

CoreWeave shares rebounded while the cost of insuring its debt set a record

$CoreWeave, Inc.(CRWV)$ rose 5.94 per cent to US$69.69 and Nebius rose 1.29 per cent to US$188.41. Both neocloud shares rebounded after the previous session's sharp decline, but the cost of insuring CoreWeave's debt remained at a record. The equity market was trading an oversold rebound while the credit market continued to raise the price of default risk, producing opposite signals.

Demand for GPU compute from model companies has not disappeared, and stable contract delivery can support a recovery in the shares. A neocloud still has to finance equipment and data-centre expansion continuously, however, and high funding costs consume cash alongside depreciation. A one-day rebound asks whether the selling went too far. Credit pricing asks whether long-term capital still wants the risk, and that does not improve automatically after a single positive session.

Rocket Lab delayed Neutron's first launch and fell for a seventh session

$Rocket Lab USA, Inc.(RKLB)$ fell 6.64 per cent to US$55.25 after the company delayed Neutron's first launch until 2027, extending the decline to a seventh consecutive session. Neutron is an important product in Rocket Lab's attempt to enter the larger launch market. The delay pushes potential revenue further out and lengthens the development-spending period.

ARK added shares during the pullback, while Rocket Lab continued to highlight its record on Mars missions, showing that the market has not rejected its existing launch and space-systems capability. What changed was the price of time. The longer Neutron takes to enter service, the longer investors must wait to test whether it can compete with SpaceX on launch efficiency. A revised launch milestone and cost control will matter more for valuation than short-term buying.

iPhone Duo pre-orders begin on Friday, putting Apple's upgrade case to the test

$Apple(AAPL)$ fell 0.28 per cent to US$342.00. Pre-orders for the company's first foldable phone, iPhone Duo, begin on Friday, giving the market its first test of whether the device can create a new reason to upgrade. The foldable design can lift the average price of premium products and place Apple Intelligence, hardware design and ecosystem services in the same upgrade cycle.

A product launch is not proof of demand. Pricing, durability and application support still require consumer validation, and a single change in form may not reverse slower upgrades in a mature market. Pre-orders will show whether early adopters are willing to pay. Retention and usage data will determine whether this becomes a durable product cycle.

This is personal analysis and not investment advice.

💬 【Talking Point】

Which operating result would best show that this cycle of AI capital spending is earning an adequate return?

💰 【Bounty】

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

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

# TSMC Profits Hit Record Highs — Why Is AI Hardware Still Being Sold?

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

Comment(7)

  • Top
  • Latest
  • 長髮哥
    ·10-18 11:18
    在評估當前這一輪龐大的超大型企業(Hyperscaler)人工智能資本支出(CapEx)是否獲得足夠回報時,雲端運算基礎設施的營收成長率(Cloud Revenue Growth)與雲端業務未履約合同金額(Cloud Backlog) 是市場和華爾街公認最能表明「支出已轉化為實質回報」的營運指標。
    雖然許多企業會看內部效率或生產力的提升(如代碼生成效率、客服自動化),但對整體科技大循環(AI Super-cycle)而言,唯有基礎設施層面的強勁變現才能證明數千億美元的晶片與數據中心投入並非泡沫。
    Reply
    Report
  • Lanceljx
    ·10-17 11:36
    The strongest evidence would be sustained growth in AI-related revenue and free cash flow that outpaces the growth in AI capex.

    Revenue growth alone is not enough. Companies can generate more AI sales while spending even faster on GPUs, data centres and power. I would look for rising utilisation of AI infrastructure, improving operating margins and, ultimately, higher free cash flow and ROIC.

    If hyperscalers can show that each additional dollar of AI investment is producing growing incremental cash returns, the spending cycle becomes much easier to justify.

    For me, improving ROIC alongside accelerating free cash flow would be the clearest sign that AI capex is creating economic value rather than simply fuelling an expensive infrastructure race.

    Reply
    Report
  • 靖润
    ·10-16
    最能说明AI资本支出获得足够回报的,不是台积电的利润,而是云厂商的经营现金流。


    具体看一个指标:资本支出增速与云收入增速的差值。如果微软、亚马逊、谷歌的云收入增速开始超过资本支出增速,并且自由现金流同步回升,那才叫回报。台积电利润再炸裂,也只是上游卖铲子的,它赚的是云厂商花出去的钱,不是AI应用赚回来的钱。


    现在PPI走高,资金从科技转向价值,AI硬件跌,就是在表达对这个回报的怀疑。台积电业绩好但股价跌,说明市场不信上游,要看下游。接下来盯着微软的Azure AI收入、亚马逊的AWS AI收入、以及它们资本支出占经营现金流的比例。这个比例如果开始回落,科技股才有第二波。如果还是资本支出猛涨、AI收入不温不火,那AI硬件还得跌。就这么简单。
    Reply
    Report
  • Kentzw
    ·10-16
    For me, it’s sustainable free cash flow growth. 💰 Revenue growth is encouraging, but AI companies need to show that the money invested in chips, data centres and power is generating real returns. Spending billions is easy to announce; proving it pays off is the real test. 📊
    Reply
    Report
  • 吉3186
    ·10-16
    我认为,最关键的指标是自由现金流(Free Cash Flow,FCF)持续增长。
    AI需求强劲、营收创新高,只能证明市场有需求;真正重要的是,企业投入巨额资金建设数据中心、购买GPU之后,能否赚回成本并创造更多现金。
    我会重点观察三个方面:
    自由现金流持续改善:资本开支增加后,公司仍能产生充足现金,而不是长期依赖借债。
    投资回报率(ROIC)提高:新增投资带来的利润,是否足以覆盖资金成本。
    客户持续付费:AI订单和长期合约能否转化为实际收入及现金回款,而不只是账面增长。
    以台积电为例,创纪录利润和强劲指引是好消息,但未来几年资本开支大幅增加,仍需要观察投资回报。对于 CoreWeave,则要特别关注债务成本、现金流和偿债能力。
    我的结论:AI资本支出是否成功,不能只看营收和订单有多大,而要看投入的每一美元,最终能创造多少可持续的现金回报。
    Reply
    Report
  • Jerry Lam
    ·10-16
    **我认为最能证明AI资本开支获得足够回报的指标,是新增投资带来的自由现金流,而不是营收增速或GPU采购规模。**

    台积电的强劲财报证明上游芯片需求真实存在,但供应商盈利,并不意味着下游算力运营商同样能够赚钱。

    我重点关注三个指标:

    ① **投入资本回报率(ROIC):** 新建数据中心的回报能否持续高于加权平均资本成本(WACC)?如果融资成本高于投资回报,规模越大反而可能越不划算。

    ② **GPU利用率与单位经济效益:** 算力出租率提高的同时,扣除电费、维护、折旧和融资成本后,单个项目能否产生合理利润?

    ③ **自由现金流改善:** 在考虑持续资本开支后,企业是否能够减少对外部融资的依赖?

    CRWV股价反弹,但信用风险指标仍然承压,正说明市场开始区分增长故事和偿债能力。

    **我的判断是:AI需求仍有增长空间,但行业将从比拼资本开支规模,逐渐转向比拼资本使用效率。**

    **一句话总结:真正健康的AI投资,不是花100亿美元创造更多订单,而是每投入1美元,最终能够持续创造超过资本成本的现金回报。**

    Reply
    Report
  • ToNi
    ·10-16
    look tsla
    Reply
    Report