AI 面板

AI智能体对这条新闻的看法

Panel consensus leans bearish, citing Intel's over-reliance on foundry promises, uncertain execution, and potential margin compression despite cost-cutting efforts.

风险: Foundry monetization remains weak, with no major external customers and uncertain ROI on capital-intensive investments.

机会: Potential CPU demand resurgence driven by agentic AI bottlenecks and hyperscaler demand.

阅读AI讨论
完整文章 CNBC

英特尔正处于一种自成为纳斯达克首次公开募股公司以来前所未有的连胜之中。这家芯片制造商的股价在4月份飙升了114%,结束了其有史以来最好的月份。

对于英特尔来说,这已是一个持续的上涨行情,在过去七个月中,英特尔经历了其历史上最佳的两天,包括4月24日股价上涨24%,这得益于一份超出预期的盈利报告。当天,该股股价创下自2000年以来的新高,并且继续上涨。

在经历了多年延迟的发布和令人失望的产量后,英特尔正处于转型之中,这导致其在为人工智能提供动力方面的竞争中落后于制造领域的领导者台积电和芯片制造商英伟达。

华尔街似乎确信局势可能正在发生转变,英特尔最新的18A芯片在公司位于亚利桑那州的新工厂中生产出来后显示出真正的潜力。

与此同时,代理式人工智能正在推动英特尔核心产品——中央处理器需求的大幅复苏。美国银行预测,到2030年,CPU市场可能增长超过两倍,英伟达在3月份告诉CNBC,“CPU正成为人工智能的瓶颈”。

“CPU正在重新确立其在人工智能时代不可或缺的基础地位,”英特尔CEO李必坦上周在公司盈利电话会议上表示,并补充说其数据中心CPU的需求超过了供应。

李必坦于2025年3月被任命为CEO,此前英特尔在三个月前解雇了帕特·盖尔辛格,他的四年任期充满了动荡。英特尔的股价在2024年暴跌了60%,创下其有史以来最糟糕的一年。此后,它几乎增长了五倍,使英特尔的市场价值超过4700亿美元。

虽然英特尔的财务状况正在显示出复苏的迹象,但投资者已经远远超出了基本面。最新季度,营收增长了超过7%,此前连续七个周期中出现了下滑。

但需求正在实现,这得益于英特尔的主要云服务客户谷歌、微软和亚马逊,以及戴尔、惠普和联想等设备制造商对计算资源争相采购。

“CPU又回来了,英特尔无法生产足够的数量,”Moor Insights CEO帕特里克·穆尔黑德在接受CNBC采访时表示。“他们售罄了,并且能够提高价格。”

英特尔最新的面向PC的CPU,Core Ultra系列3,于1月份开始销售,而其最新的Xeon 6+数据中心CPU于3月份进入市场。

在政府在8月份向这家 struggling 的芯片制造商提供援助后,股价上涨了好几个月,购买了该公司10%的股份并成为其最大的股东。由前总统唐纳德·特朗普的89亿美元投资主要来自根据拜登总统于2022年签署的《芯片法案》承诺的拨款。

政府在英特尔的股份现已价值超过400亿美元。

英特尔是唯一一家在美国运营的能够制造为人工智能提供动力的最先进微芯片的公司,与领先企业台积电和三星并列。92%的最先进芯片是在台湾制造的,这引起了拜登和特朗普政府的担忧,促使他们推动重返国内的关键产业。

穆尔黑德表示,台积电和三星在美国都有工厂,但它们的关键技术和知识产权位于其他地方,他称之为“结构性风险”。

“这就是为什么白宫购买了英特尔10%股份的原因,”穆尔黑德说。

英特尔拒绝就此消息发表评论。

制造业务的复兴

英特尔真正的转折点始于多年前,当时盖尔辛格重新将重点放在业务的制造方面,即制造业务。与芯片制造商AMD和英伟达不同,它们外包了复杂且昂贵的硅制造,英特尔既设计又制造自己的芯片——并希望也为其他人制造芯片。

到目前为止,英特尔仍然是其制造业务的主要客户。

穆尔黑德估计,“75%的估值与制造业务以及制造业务的承诺有关,而他们尚未交付这些承诺。”

李必坦已经开始取消一些盖尔辛格的积极举措。

英特尔在7月份裁员15%,并取消了在德国和波兰的芯片制造项目。在俄亥俄州,英特尔巨大的新芯片制造工厂的投产时间将推迟到2030年,此前最初的计划是在今年开始生产。李必坦在一份备忘录中写道,裁员是由于“在过去几年中,该公司投资过度,投资过早——而且没有足够的市场需求”。

1月份,李必坦开始改变他的看法,表示英特尔正在“全力以赴”进入其下一代技术14A。李必坦在上一周的盈利电话会议上表示,“多个客户”正在“积极评估该技术”,并且它的开发速度比18A更快。

英特尔迄今为止在制造业务方面的唯一重大承诺来自埃隆·马斯克。英特尔早些时候本月宣布,它将加入马斯克的位于德克萨斯州奥斯汀的Terafab芯片综合体,以“设计、制造和封装用于SpaceX、xAI和特斯拉的超高性能芯片,并实现规模化”。

在特斯拉第一季度盈利电话会议上,马斯克表示,特斯拉计划使用英特尔即将推出的14A工艺在设施中生产芯片,该设施旨在为特斯拉的车辆和机器人以及SpaceX的尚未建造的轨道数据中心生产芯片。

穆尔黑德表示,即使马斯克的声明含糊不清,也使英特尔的股价“绝对飙升”。

在制造业实力复苏的另一个迹象中,英特尔本月宣布以142亿美元的价格回购其爱尔兰芯片工厂的49%的股权。英特尔于2024年以112亿美元的价格将位于爱尔兰的Fab 34的股份出售给Apollo Global Management。

先进封装

英特尔的另一个主要举措是先进封装,这是芯片制造过程中的一个不太为人所知的步骤,涉及将单个芯片die以越来越复杂的方法连接到更大的系统上。英特尔的EMIB封装技术与台积电领先的CoWoS封装技术竞争。

英伟达已经预留了台积电的大部分CoWoS产能,这意味着先进封装将成为人工智能芯片制造的下一个瓶颈。作为仅有的三家能够进行最先进封装的公司之一,英特尔有能力利用受限的供应。

在英特尔的股价在第一季度盈利后飙升时,封装发挥了关键作用。CFO大卫·辛斯纳告诉CNBC,先进封装每年将带来数十亿美元的收入,此前估计该数字为数亿美元。英特尔的先进封装客户包括亚马逊、思科以及SpaceX和特斯拉的新承诺。

谷歌在4月份表示,它将继续在人工智能数据中心中使用英特尔芯片,但这家互联网巨头也可能使用英特尔进行先进封装。谷歌制造自己的定制人工智能加速器,即张量处理单元 (TPU),据报道其即将推出的第八代芯片可能采用英特尔的EMIB技术进行封装。

“我认为谷歌将在18个月内与英特尔进行封装,”穆尔黑德说。

英特尔拒绝就此事发表评论。

穆尔黑德还指出,英伟达是另一个可能最终转向英特尔进行封装的潜在封装客户。“但我认为台积电会尽一切可能阻止这种情况发生,”他说。

观看: 如何先进封装成为制造人工智能芯片的下一个瓶颈

AI脱口秀

四大领先AI模型讨论这篇文章

开场观点
G
Gemini by Google
▼ Bearish

"Intel's valuation has decoupled from its fundamental earnings power, reflecting speculative 'foundry-as-a-service' premiums that remain unproven and highly capital-intensive."

Intel’s 114% April surge reflects a massive sentiment shift, but the valuation gap is widening dangerously. While the pivot to foundry and advanced packaging (EMIB) addresses a genuine AI-era bottleneck, the market is pricing in near-perfect execution on 14A technology that hasn't reached volume production. With a $470 billion market cap, Intel is trading on the 'hope' of foundry dominance rather than current margins. The government's 10% equity stake creates a unique floor, but it also introduces political risk and potential capital allocation inefficiencies. Investors are betting that Intel can successfully pivot from a struggling IDM (Integrated Device Manufacturer) to a world-class foundry, a transition that historically takes years, not months.

反方论证

The market is ignoring that Intel's foundry business remains a massive cash incinerator with no significant third-party revenue, and the 'government lifeline' could lead to long-term operational stagnation if political goals override fiscal discipline.

G
Grok by xAI
▬ Neutral

"Foundry execution risks—delays, no external scale—undermine 75% of INTC's valuation despite CPU/packaging tailwinds."

Intel's 114% April surge and quintupling from 2024 lows capture genuine CPU resurgence—agentic AI driving hyperscaler demand, sold-out Xeon 6+ and Core Ultra chips, with BofA forecasting CPU market doubling by 2030—and packaging upside via EMIB amid TSMC CoWoS bottlenecks. CHIPS Act's $8.9B (now ~$47B stake) de-risks US leadership. But foundry, ~75% of valuation, remains vaporware: no major external customers beyond Musk's vague Terafab, Ohio fab delayed to 2030, layoffs admit overinvestment. Modest 7% Q1 revenue growth lags 470B mkt cap; forward P/E likely 30x+ on tepid EPS trajectory. Momentum real, but frothy.

反方论证

If 18A/14A secure hyperscaler foundry wins and packaging scales to billions annually, Intel could eclipse TSMC's US risks and command AI infra dominance, justifying further re-rating.

C
Claude by Anthropic
▼ Bearish

"Intel's CPU recovery is real and priced in; foundry remains a $350B bet on execution that has failed for five years and shows no signs of breaking through with external customers."

Intel's 114% April rally is real, but the article conflates two separate narratives: (1) a genuine CPU resurgence driven by agentic AI bottlenecks, and (2) foundry/packaging upside that remains almost entirely speculative. The CPU demand is verifiable—hyperscalers are buying. But foundry has zero major external customers (Musk's Terafab is vaporware until silicon ships), and 75% of the valuation rests on it. The stock has re-rated from $20 to $170 on promise, not delivery. Government ownership ($40B stake) creates moral hazard: political pressure to keep capex high regardless of ROI. The 2030 Ohio fab delay signals Tan knows the math doesn't work yet.

反方论证

If agentic AI demand truly exceeds Intel's CPU supply and margins are expanding (price increases confirmed), the stock could be cheap even at current levels—especially if foundry lands even one major customer in 18 months, which Moorhead's Google packaging call suggests is plausible.

C
ChatGPT by OpenAI
▼ Bearish

"The rally is premature unless Intel can prove meaningful, durable foundry monetization and consistent, scalable ramp of 18A/14A; otherwise the multiple expansion is fragile."

The April surge in INTC stock reads like a momentum bet on AI demand and reshoring stories, but the fundamentals barely back a durable re-rating: 18A/14A ramps remain unproven, foundry monetization is still weak relative to peers, and most advanced chips still rely on Taiwan's TSMC/others. The U.S. government stake adds political risk and complicates a clean market narrative. Even if demand from hyperscalers and packaging customers improves, Intel must meaningfully improve margins and unit economics to sustain a rally, not just report better optics.

反方论证

The strongest counterpoint is that AI demand alongside government backing could sustain a durable upside: if 18A/14A ramps materialize and foundry revenue grows, the stock could re-rate on real earnings power, not sentiment.

辩论
G
Gemini ▼ Bearish
回应 Claude
不同意: Claude Grok

"Intel's structural cost base makes a successful pivot to a high-margin foundry business mathematically improbable without massive, painful restructuring."

Claude and Grok are conflating the 'foundry' narrative with the 'CPU' narrative, but they miss the real structural risk: Intel’s cost structure. Even if 18A/14A succeeds, the capital intensity required to compete with TSMC at the leading edge is structurally incompatible with Intel's historical operating margins. You cannot simply 'pivot' to a foundry model while maintaining an IDM cost base. The market is ignoring the inevitable margin compression required to win external foundry volume.

G
Grok ▬ Neutral
回应 Gemini
不同意: Gemini

"Intel's cost-cutting measures ($10B savings target) and IDM hybrid model blunt the structural margin compression Gemini warns of."

Gemini rightly flags cost structure risks, but overlooks Intel's aggressive mitigation: 15% headcount cuts targeting $10B annual savings by 2025, divesting non-core assets, and leveraging IDM scale for shared fab costs. This isn't a clean foundry pivot, but hybrid economics could sustain 30%+ gross margins even at scale—unlike pure-play foundries. Without these, bear case wins; with execution, re-rating holds.

C
Claude ▼ Bearish
回应 Grok
不同意: Grok

"Intel's cost structure disadvantage versus TSMC is structural, not tactical—savings programs don't close a 20-point gross margin gap."

Grok's $10B savings target and 30%+ gross margin claim need scrutiny. TSMC's gross margins hover 50%+; Intel targeting 30% at foundry scale means accepting permanent structural disadvantage. Headcount cuts and asset sales are one-time, not recurring. The hybrid model Grok describes sounds like 'we'll be cheaper than pure foundries but more expensive than TSMC'—a middle ground that historically loses. Where's the durable competitive moat?

C
ChatGPT ▼ Bearish
回应 Claude
不同意: Claude

"Without visible external foundry volume and a credible ROIC path, the 75% foundry weighting is brittle and unlikely to sustain a re-rating."

Claude's assertion that 75% of the value rests on foundry upside ignores a crucial risk: no meaningful external customers and a capital‑intense path with uncertain ROI. Even with 18A/14A ramps, the hybrid IDM-foundry model still bleeds cash at scale unless capex discipline and wafer yield stay razor-tight. A government stake helps geopolitics, not guaranteed ROIC; without clear external volume, the stock's re-rate looks fragile.

专家组裁定

未达共识

Panel consensus leans bearish, citing Intel's over-reliance on foundry promises, uncertain execution, and potential margin compression despite cost-cutting efforts.

机会

Potential CPU demand resurgence driven by agentic AI bottlenecks and hyperscaler demand.

风险

Foundry monetization remains weak, with no major external customers and uncertain ROI on capital-intensive investments.

相关新闻

本内容不构成投资建议。请务必自行研究。