Amid a sharp rise in its stock price, is Google's growth only just beginning? With an accelerating AI ecosystem and TPU advancements, which beneficiary stocks should investors watch?
The landscape of the AI chip market is quietly undergoing a transformation. $NVIDIA (NVDA.US)$As attention grows and rapid expansion continues, Google is quietly building a robust AI infrastructure empire. While market focus remains on OpenAI, $Microsoft (MSFT.US)$、 $NVIDIA (NVDA.US)$While market attention is concentrated on the AI ecosystem led by OpenAI and others, Google is leveraging its unique approach to demonstrate overwhelming strength.
The parent company of Google $Alphabet-C (GOOG.US)$The stock price has continued to hit record highs recently, with its market capitalization exceeding $3.5 trillion.（The company's market capitalization surpassed the $3 trillion mark in September.). In the third quarter, revenue exceeded $100 billion for the first time, with Google Cloud’s growth rate recovering; backlog reached $155 billion, and annual capital expenditure was raised from $91 billion to $93 billion. Moreover,Google announced last week its most powerful chip to date, 'Ironwood,' which is scheduled for full deployment over the coming weeks..The performance data of this chip is impressive, directly disrupting the current competitive landscape of the AI computing power market.
The initial market response has been enthusiastic, with companies such as Anthropic, Lightricks, and Essential AI already placing orders. Anthropic, an AI startup,will utilize up to 1 million new TPUs.They plan to operate their own Claude model. BofA is,BofA estimates that this partnership will generate up to $10 billion in annual revenue for Google.according to the calculations.
According to reports, the performance of Ironwood chips in training and inference has improved fourfold compared to the 6th generation TPU,a fourfold increase,with Google evaluating them as being 'designed to handle the most demanding workloads.' This includes a strategic aim to eliminate data bottlenecks in the training and inference of large-scale AI models.a strategic objective,There is.
In order to achieve this goal, Ironwood adopts a cutting-edge interconnect architecture. According to Google,a single AI supercomputer unit (POD) can connect up to 9,216 Ironwood TPUs and access 1.77 PB of shared high-bandwidth memory.This is essential for the operation and scaling of existing large-scale, data-intensive AI models.
This indicates that Google is engaged in a high-stakes competition with rivals such as Microsoft $Amazon (AMZN.US)$ 、 $Meta Platforms (META.US)$ and other competitors,with the aim of building future AI infrastructure.Moreover, it highlights that Google’s TPUs are directly challengingthe GPU-dominated AI chip market.indicating $NVIDIA (NVDA.US)$It shows that it is directly challenging the GPU.
According to analysis, many large language models and AI workloads still rely heavily on NVIDIA's GPUs.Google's TPU is a custom chip specialized for AI workloads, with the potential to demonstrate significant advantages in terms of cost, performance, and energy efficiency.There is.
For Google Cloud customers, Ironwood offers two configurations based on workload requirements: a 256-chip setup and a 9,216-chip setup. A single POD can accommodate up to 9,216 TPUs, delivering a total computing power of 42.5 ExaFLOPS, significantly surpassing NVIDIA’s current GB300 NVL72 system, which provides 0.36 ExaFLOPS.
This is one of the reasons why Anthropic directly purchased one million TPUs.
According to analysts,The value of Google's TPU business may currently be undervalued.There is.The TPU industry is transitioning from 'internal development' to 'full commercialization,' and is expected to rapidly establish its market position in the coming years.
In September of last year, Gil Luria of D.A. Davidson pointed out that Google's TPU has narrowed the gap with NVIDIA and become the 'best alternative.' He believes that if the TPU business were integrated with Google's DeepMind AI research institute, its value could reach $900 billion.
Ben Reitzes, an analyst at Melius Research, also expressed high expectations for Google's TPU, referring to it as 'the most mature application-specific integrated circuit (ASIC),' which is a chip specialized for specific AI tasks. Reitzes pointed out that by leveraging TPUs, Google can innovate rapidly, thereby contributing to both Broadcom’s AI revenue and the growth of Google Cloud.
Furthermore, This 'vertical integration' model has transformed cloud service providers (CSPs) from mere 'purchasers of computing power' into 'definers and suppliers of computing power.'This transformation fosters a closely collaborative industrial chain where profits extend from traditional GPU manufacturers to upstream IP/EDA, foundries, advanced packaging, and downstream server ODMs as well as key components such as optical modules, PCBs, and liquid cooling systems.These developments provide leading companies in these fields with structural growth opportunities, allowing leaders in each sector to leverage their strengths while facing clear market opportunities.。
Google is both the definer and the largest beneficiary of the ecosystem.Its investment value depends on the success of its entire AI strategy, with the TPU serving as a core component.Through vertical integration from software (such as TensorFlow and JAX) to hardware (TPUs), Google generates significant synergy effects.Currently, Google is transitioning its TPU strategy from internal use to external availability, aiming to penetrate third-party data centers through models such as HaaS (Hardware as a Service) and explore new markets.
● The key partners for chip design and CoWoS contract manufacturing
In realizing TPU chips, design services and CoWoS contract manufacturing represent the most valuable processes. $Broadcom (AVGO.US)$ and $Taiwan Semiconductor (TSM.US)$are the "suppliers of shovels," with the former offering scarce ASIC design capabilities and the latter providing superior manufacturing expertise.
As Google's long-standing design service partner, $Broadcom (AVGO.US)$the company is deeply tied to Google through its extensive IP portfolio and is expected to continue benefiting from an accelerated XPU (a general term for various processors) rollout anticipated beyond 2026.
$Taiwan Semiconductor (TSM.US)$TSMC's exclusive position in its advanced logic process and CoWoS packaging technology ensures the production capacity of TPUs. Amidst the continuously growing demand for AI chips, TSMC's production capacity has become a stable source of revenue.
$Amkor Technology (AMKR.US)$One of the world’s leading semiconductor outsourced assembly and testing (OSAT) companies, expanding advanced packaging capabilities in Arizona, USA, alongside TSMC, to meet the backend demand for high-performance computing and data center chips.
● Module and Rack Assembly
$Celestica (CLS.US)$Focused on switches/routers and AI networking equipment for data centers, offering an 800G/1.6T switching system platform equipped with Broadcom chips. Designed to handle increasing traffic in hyperscale data centers.
$Jabil (JBL.US)$As a global leader in manufacturing and supply chain, announced the J-422G server and liquid cooling/rack integration solutions for AI workloads, enhancing cooling management capabilities through acquisitions.
$Flex Ltd (FLEX.US)$Deploying an integrated infrastructure platform for AI data centers that incorporates power, cooling, and computing into pre-designed modular reference designs, aiming to increase deployment speed by up to 30%. Flex also offers direct liquid cooling, rack power, and end-to-end integration and operational services.
● PCB, Copper Clad Laminates, Cables
$TTM Technologies (TTMI.US)$One of North America’s largest PCB manufacturers, achieving record growth with high-layer-count/large backplanes for AI and data centers.
$Panasonic Holdings (6752.JP)$Providing polymer capacitors and thermal dissipation materials (such as graphite sheets) for servers and base stations, excelling in reliability under high-temperature and high-ripple current environments.
$Amphenol (APH.US)$A leader in high-speed copper cables/connectors, having commenced mass production of 800G and promoting 1.6T interconnect formats (such as OSFP and QSFP-DD), meeting the demand for high-speed backplanes/cables in hyperscale data centers.
● Network
$Lumentum (LITE.US)$Providing optical switches (OCS) and high-speed optical devices, reducing network latency and energy consumption in large-scale AI clusters, playing a key role in the construction of 'AI Network Factories.'
Google is the only full-stack AI company.
The world is $NVIDIA (NVDA.US)$Amid growing attention, Google's AI strategy has been quietly taking shape, building its own "ecosystem."The market may beunderestimating the value of Google's 'full-industry chain vertical integration.'In this AI competition, Google has established strong barriers to entry.
According to analysis,Google is the only major technology company in the AI field that possesses full-stack capabilities encompassing 'model (Gemini) - computing power (Google Cloud) - chips (TPU)',with its own computational infrastructure, making it independent of external dependencies. This vertical integration provides Google with unparalleled flexibility and cost management,establishing a robust barrier to entry.。
Google controls every layer of the AI value chain. From AI-specialized TPU accelerators (v6e Trillium, v7p Ironwood) and Arm-based Axion CPUs, to the aggregated system architecture known as the 'AI Hypercomputer,' and further to the 'AI Operating System' provided by the Vertex AI platform, Google powers its core intelligence engine, the Gemini model, and delivers services to billions of users worldwide through applications like NotebookLM and Workspace.
Google's AI strategy、is a meticulously designed global project aimed at building an unprecedented 'full-stack' vertically integrated empire, spanning from chip design at the lowest level to application ecosystems at the highest level, and extending even to future space computing initiatives..
In addition to revenue from cloud services and TPU rentals,Google holds approximately 14% of Anthropic's shares and a significant stake of 5% to 10% in SpaceX.Earlier this year,the rise in SpaceX’s valuation boosted Alphabet’s unrealized gains on its books.These financial assets provide evaluation leverage tied to economic cycles, even outside the cloud and TPU businesses.
The most advanced and perhaps most critical move in Google’s AI strategy is、a solution to the ultimate constraint of AI: 'energy.'In November 2025, amid a sharp increase in computing power and electricity demand, Google announced a 'moonshot' research project called 'Project Suncatcher.' The vision of this project isto build scalable AI data centers powered by solar energy in space (low Earth orbit, LEO).It is to do.
According to Google's official blog, the team has announced a partnership with Planet to launch two prototype satellites as early as the beginning of 2027, aiming to test hardware in space. Google pointed out that the growing demand for AI computing power could far exceed the increase in terrestrial energy supply.
AI is energy-intensive, and the sun is the largest energy source in the solar system. In space, solar panels can achieve efficiency up to eight times higher than on Earth, enabling near-continuous operation around the clock throughout the year while minimizing reliance on large batteries.
Meanwhile, on October 29 local time, $Alphabet-A (GOOGL.US)$Google announced its Q3 earnings, with its cloud business accelerating growth and Q3 revenue reaching $151.57 billion.a 34% increase year-over-year.The backlog of orders for the cloud businesssurged by 46% quarter-over-quarterto $155 billion.
According to a research report by Morgan Stanley, Google Cloud's revenue growth potential significantly exceeds the market's general expectations, even under conservative assumptions.There is a possibility that revenue growth in 2026 could exceed 50%.It is said that there is.
Notably, in response to surging demand, Google has upwardly revised its capital expenditure forecast for the second time within the year. Capital expenditures for 2025 are expected to range between $91 billion and $93 billion.
As computing power becomes a scarce resource and the cloud market shifts towards inferencing applications, computing costs have become a key focus. By leveraging its mature TPUs, Google can offer lower-cost computing services and expand its market share. Future attention will be on the growth trends of TPU contracts.
Source: moomoo, publicly available market data
This article utilizes partial automated translation.
-moomoo News Sherry
This article utilizes partial automated translation.
-moomoo News Sherry
Disclaimer: Moomoo Technologies Inc. is providing this content for information and educational use only.
Read more
Comment
Sign in to post a comment
Market Insights
Hot Topics
Trump's 'Taco deal' sparks market rebound : Market rebound sustainable ?
1. If tariffs pause, which sectors benefit first? 2. Trump's "threat-compromise" cycle normalizes—how to hedge across assets?
Moo Options Explorer
Apr 7 17:28
Weekly roundup: Tariff turmoil and recession fears dominate
高松宮 : Google, you too! Following Alibaba's entry into semiconductor production, which has begun to challenge Nvidia's stronghold, Google is also entering the market, intensifying competition even further!