NVIDIA Entered CPU War: ARM the Winner, Intel or AMD the Losers?

At Computex, Huang announced Vera CPU entering mass production and RTX Spark crashing into the PC market — NVIDIA is now officially in the CPU business. The market voted with its feet: $ARM Holdings(ARM)$ surged +15.7% to $409, the biggest free-rider winner; $NVIDIA(NVDA)$ +6.3% to $224; while the x86 duo took it on the chin — $Advanced Micro Devices(AMD)$ -1.2% to $510, $Intel(INTC)$ -4.7% to $109. A Barclays report just ranked the winners and losers of this $100B+ CPU war.

For years the semiconductor spotlight was on GPUs. But over the past six months, AI workloads have shifted from training to inference and agents (Agentic AI) — GPUs "compute," CPUs "manage": calling tools, routing sub-agents, tracking task completion. That's CPU work.

The CPU-to-GPU ratio has shifted from ~1:2 to near 1:1 — a completely different standing. Market-size forecasts have doubled in six months: NVIDIA sees a $200B TAM by 2030; AMD pegs the server CPU market at $120B with its own share above 50%.

NVIDIA: the most dangerous new variable

Vera is NVIDIA's first standalone data-center CPU, taking direct aim at Intel Xeon and AMD Epyc. Huang's numbers: 3x faster than x86 on SQL, 6x on data processing, ~1.8x overall on common agent tooling, with Anthropic and OpenAI as early customers. ~$20B of CPU revenue visibility this year.

But there's an opaque variable: nearly all of NVIDIA's CPUs ship bundled inside NVL rack systems — no standalone ASP, and the revenue is "embedded" in rack estimates, not broken out. Which means the market may not have fully priced its CPU footprint yet.

Intel: the most cornered

The target was raised from $65 to $100, but Barclays was blunt: that's ASP-driven valuation repair, not share improvement. The verdict — "ongoing core share loss, structurally pressured profitability, no evidence of a manufacturing-leadership comeback."

The starkest number is the 2030 CPU EPS contribution: AMD ~$19, ARM ~$7, Intel just ~$1.5. x86's efficiency disadvantage on agentic workloads + a lagging process node + getting squeezed by both AMD and NVIDIA = a structural bind no single product cycle fixes.

ARM: does nothing, collects royalties

ARM's logic is the simplest: whoever builds the CPU, if it's on Arm, they pay the royalty. NVIDIA's Vera is Arm; AMD is migrating parts of its line to Arm. Barclays lifted the target from $250 to $360, with ~$7 of 2030 CPU EPS contribution. It never steps onto the battlefield to fight for share — the most stable way to benefit, and why it spiked 15% in a single day.

Whose side are you on in the CPU war?

Is the market simply rewarding the certainty of "royalty collection"?

Can Huang's Vera really take share from x86, or is it just an add-on bundled into NVL racks?

Intel's target was lifted to $100, but it's only valuation repair — do you think it has a comeback left in this cycle?

Leave your comments to win tiger coins~

# Computex: NVIDIA Windows PC Arrives — Can Microsoft Turn the Tide?

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

Comment18

  • Top
  • Latest
  • Intel Target Lift: Valuation Repair Without a Cyclical ComebackThe upward revision to a $100 price target reflects pure balance-sheet valuation repair and does not signal a major cyclical comeback. The reassessment is driven by structural cost-cutting, asset optimization, and government-subsidized foundry infrastructure rollouts. Intel's core data center and client processor businesses still face market share erosion against rival x86 and Arm architectures. The operational turnaround is a long-term capital expenditure story that will not yield a dominant cyclical product comeback in the near term.
    Reply
    Report
  • TimothyX
    ·10:52
    TOP
    The CPU-to-GPU ratio has shifted from ~1:2 to near 1:1 — a completely different standing. Market-size forecasts have doubled in six months: NVIDIA sees a $200B TAM by 2030; AMD pegs the server CPU market at $120B with its own share above 50%.
    Reply
    Report
  • Chrishust
    ·04:41
    1. In this cpu war $ARM Holdings(ARM)$ has the better position, however $NVIDIA(NVDA)$ is also a strong contender
    2. Yes $ARM Holdings(ARM)$ has a high valuation due to the value of royalties of the arm chipset instruction set
    3. Yes, $NVIDIA(NVDA)$ new cpu has higher performance that legacy x86 cpu chipsets
    4. $Intel(INTC)$ is highly valued for its current position in the market which is second to amd and nvda
    Reply
    Report
  • Shyon
    ·00:02
    I’m leaning toward $ARM Holdings(ARM)$ in this CPU war. Its business model is the most attractive because it benefits no matter who wins. Whether it’s $NVIDIA(NVDA)$ , $Advanced Micro Devices(AMD)$ , or hyperscalers building Arm-based CPUs, ARM collects royalties without having to fight for market share directly. That certainty helps explain the stock’s strong reaction.

    NVIDIA is still the biggest wildcard. Vera may not replace x86 overnight, but within NVIDIA’s AI ecosystem it doesn’t need to. If customers are already buying NVL racks, adopting Vera becomes a natural extension. The market may also be underestimating how much CPU revenue is embedded inside those systems.

    For Intel, I’m not convinced a comeback is near. A higher target driven by valuation repair is different from improving competitiveness. Intel still faces pressure from AMD and NVIDIA, and until I see clear market-share gains and product leadership, I remain cautious.

    @Tiger_comments @TigerClub @TigerStars

    Reply
    Report
  • NVIDIA Vera: A Tied Product Bundle over Standalone DisruptionJensen Huang's Vera CPU will operate primarily as an integrated add-on bundled into NVL racks, rather than capturing organic standalone x86 market share. Vera features 88 custom Olympus cores designed explicitly to orchestrate data flow and execute sandboxed code loops for agentic AI workloads. However, its proprietary NVLink-C2C interconnect confines its true value to dense, high-end NVIDIA server layouts. It will not replace x86 processors in general-purpose enterprise servers that lack proprietary accelerators.
    Reply
    Report
  • The Royalty Collection Model: Rewarding Recurring Cash CertaintyThe market is aggressively rewarding the structural certainty of IP and royalty collection. Monetizing foundational architectural blueprints creates highly resilient, recurring cash flows with zero direct manufacturing exposure. Fabricating leading-edge silicon incurs soaring multi-billion-dollar capital expenditure risks and yield uncertainties. In a volatile macro environment, investors value the recurring royalty model as a safe, highly defensive software-like toll booth on all global computing.
    Reply
    Report
  • CPU War Realignment: Standing with the Merchant EcosystemThe alignment belongs to the merchant ecosystem of customizable compute rather than any single chip giant. The market is transitioning away from proprietary hardware lock-ins. The winning side is the one that prioritizes open, flexible deployment metrics like performance-per-watt and seamless software compilation. Customers increasingly reject being forced to choose between legacy x86 architectures and rigid proprietary stacks.
    Reply
    Report
  • Mkoh
    ·09:16
    Nvidia is entering with strong Arm-based chips like Grace (data center) and upcoming Vera/PC designs to challenge x86 in AI servers and laptops. This boosts ARM's momentum in efficiency-focused workloads.

    However, AMD and Intel aren't losers yet. x86 remains dominant in legacy software, gaming, and high-performance desktops/servers. The market is fragmenting into specialized chips rather than one winner. ARM gains share (especially AI), but coexistence is likely. Nvidia adds competition for all.

    Too early to declare victors—performance, software ecosystem, and pricing will decide. (

    Reply
    Report
  • 北极篂
    ·07:50
    所以如果问我站哪边,我会偏向“卖铲人逻辑”——Arm拿版税、NVIDIA吃生态,而Intel短期更像在争取不被边缘化。
    Reply
    Report
  • 北极篂
    ·07:50
    反而Intel让我最担心。现在的问题已经不只是产品慢半拍,而是市场信心出现结构性流失。过去x86是默认标准,但当客户发现AI时代不一定非用x86不可,护城河就开始松动了。
    Reply
    Report
  • 北极篂
    ·07:50
    至于Vera能不能真的抢走x86市场?我认为答案是“会,但不会一夜之间”。因为它最大的优势不是单挑Intel Xeon或AMD EPYC,而是绑定整个NVL系统。如果客户已经买了NVIDIA整套AI基础设施,顺手采用Vera其实是顺理成章的事。换句话说,Vera更像“生态入侵”,而不是传统CPU战争。
    Reply
    Report
  • 北极篂
    ·07:50
    我觉得市场这次最聪明的地方,是第一时间先买Arm。因为它的商业模式太舒服了——谁赢都要交“过路费”。不管是NVIDIA的Vera,还是未来更多Arm服务器芯片,只要跑在Arm架构上,Arm就能稳定收版税。这种确定性,在波动大的半导体行业非常值钱。
    Reply
    Report
  • 北极篂
    ·07:49
    过去AI时代大家只盯着GPU,因为训练模型拼的是算力。但现在进入Agentic AI阶段后,事情开始变了。GPU负责计算没错,但谁来调度任务、管理Agent、处理数据库和工具调用?这些其实都是CPU在干的活。所以CPU的重要性,明显被重新估值了。
    Reply
    Report
  • It will benefit the consumers. With new players coming into the market, we will see a price war between the manufacturers and the sellers.
    Competition will bring down the price of the CPU at least for a period of time... Let's save up to buy a new PC
    Reply
    Report
  • moliya
    ·06:24
    I think Nvidia is stands out all of the chip maker and going to stand tall out all of them....
    Reply
    Report
  • highhand
    ·00:12
    nvda is breaking out with such low forward PE. just buy to ride the upside
    Reply
    Report
  • ECLC
    ·10:54
    Waiting for benefits of price war.
    Reply
    Report