Google (Alphabet) AI Value Chain

The Google (Alphabet) AI value chain is the network of public companies that build, supply, and assemble the hardware behind Google’s artificial-intelligence stack. Rather than owning Alphabet stock alone, this concept spans the custom TPU silicon, foundry, memory, optical, networking, and server suppliers that turn Google’s record AI capital spending into working data centers.

Category	Thematic / AI infrastructure
Representative companies	11
Related ETF	TIGER Google Value Chain ETF (0190Y0)
Last updated	2026-06-04

Key takeaways

The Google (Alphabet) AI value chain is a way to own Google's AI supply chain (the chip, foundry, memory, optical, networking, and server suppliers) instead of just Alphabet stock.
Its center of gravity is custom silicon: Google's TPU, co-designed with Broadcom since 2016, anchors the chain and runs Search, Gemini, and, under a 2025 deal, Anthropic's Claude.
Alphabet guided 2026 capital expenditure to $175-185 billion, up from $91.4 billion spent in 2025, and that spending flows directly to the suppliers in this chain.
Representative names span Alphabet (GOOGL), Broadcom (AVGO), TSMC (TSM), the optical layer (Lumentum, Ciena, Arista, Zhongji Innolight), Marvell, Synopsys, Amphenol, ASML, and ODMs like Celestica and Quanta.
The simplest way to own the whole chain in one ticket is the TIGER Google Value Chain ETF (0190Y0) from Mirae Asset, the Korea-listed fund that tracks the Akros Google Value Chain Index across Alphabet, Broadcom and the optical, networking and semiconductor suppliers.

What is the Google value chain?

Most investors get exposure to Google by buying Alphabet. This concept does something narrower: it follows the money Alphabet spends. To run Search, Gemini, and Google Cloud, Google designs its own AI chips and pours money into data centers, and that spending lands on a chain of suppliers, chip designers, the foundry that makes the silicon, memory and packaging vendors, the optical and networking firms that wire the clusters, and the contract manufacturers that bolt the servers together. Owning the chain is a bet that the build-out is real regardless of whether Alphabet’s own ad margins hold up.

The catalyst is capital spending, and the number is large. Alphabet spent $91.4 billion on capital expenditure in 2025 and guided 2026 capex to a range of $175-185 billion, roughly double the prior year (CNBC). That sits inside a hyperscaler build-out where combined 2026 capex is tracking toward $700 billion or more across the largest cloud players (Fortune).

Why is custom silicon (the TPU) the center of the chain?

What makes Google’s chain different from a generic semiconductor basket is the TPU, Google’s own AI accelerator. Google is one of the few buyers running its own custom silicon at scale instead of relying entirely on merchant GPUs, and it does not build that chip alone. Broadcom has co-designed Google’s TPUs since 2016, a partnership now extended through 2031 that also covers the networking and packaging silicon in Google’s AI racks (Capacity Media). The physical chip is fabricated by TSMC: the seventh-generation Ironwood TPU runs on TSMC’s 3nm (N3P) process with CoWoS advanced packaging (Tom’s Hardware).

Google pitches Ironwood as a generational step built for running AI, not just training it.

Ironwood is our seventh-generation Tensor Processing Unit (TPU), our most performant and scalable custom AI accelerator to date, and the first designed specifically for inference. It scales up to 9,216 liquid cooled chips.

— Google

The TPU story also reaches beyond Google’s own products. In October 2025 Anthropic agreed to expand its use of Google Cloud TPUs to up to one million chips and more than a gigawatt of capacity coming online in 2026, a deal reported to be worth tens of billions of dollars (Data Center Dynamics). Anthropic framed the choice around the chip’s economics:

Anthropic selected TPUs for their price-performance and efficiency, informed by its experience training and serving models on TPUs.

— Anthropic

What does the money look like for suppliers?

The capex is already showing up in supplier results. Broadcom, the chain’s chip engine, reported fiscal Q4 2025 revenue of $18.0 billion, up 28% year over year, with AI semiconductor revenue rising 74% to $6.5 billion and a total AI order backlog of $73 billion (The Motley Fool). Management guided AI semiconductor revenue to roughly $8.2 billion in fiscal Q1 2026, about double the year-ago figure (The Motley Fool).

Zoom out and the demand sits in a market still growing fast but no longer guaranteed to accelerate. Omdia put the AI data-center chip market at $207 billion in 2025 and reaching $286 billion by 2030, while flagging that custom ASICs like the TPU are taking share from GPUs even as overall growth likely peaks (Omdia). That share shift is the bull case for owning the chain: the suppliers most tied to custom silicon and the infrastructure around it can grow even if merchant-GPU spending plateaus.

Who should consider it?

This fits an investor who wants the “picks and shovels” of the Google AI trade rather than Alphabet stock alone: ownership of the chipmakers, foundry, optical, and assembly vendors that get paid as Google builds. The driver is concrete spending. Alphabet spent $91.4 billion on capital expenditure in 2025 and guided 2026 capex to $175-185 billion, roughly double the prior year (CNBC), and that money lands directly on these suppliers. It is already showing in results: Broadcom, the chain’s chip engine, grew AI semiconductor revenue 74% year over year to $6.5 billion in fiscal Q4 2025 and carried a $73 billion AI backlog (The Motley Fool).

Size it as a small thematic growth satellite, not a core holding or an Alphabet substitute. The basket leans on one customer’s capex, so a pause in Google’s build-out would hit every layer at once, and many constituents are high-beta semiconductors and contract manufacturers trading at rich multiples. The market backdrop is no longer a one-way bet either: Omdia puts the AI data-center chip market at $286 billion by 2030 but flags that overall growth is likely peaking even as custom ASICs like the TPU take share (Omdia). It is not for an investor who wants diversified core equity, who cannot stomach single-customer concentration, or who wants to avoid the currency and custody complexity of suppliers cross-listed in Taiwan, Korea, and China.

What are the layers of the chain?

The chain stacks from sand to server. The foundry layer is TSMC, which fabricates the leading-edge silicon, and ASML, whose lithography machines make that fabrication possible. The chip layer is Broadcom and Marvell on custom accelerators and networking, with Synopsys supplying the chip-design software used to tape them out. The memory and packaging layer brings in Samsung and Amkor. The interconnect layer is optical and networking, Lumentum, Ciena, Arista, Zhongji Innolight, and connector maker Amphenol, the same suppliers covered in optical networking. Finally the assembly layer is the ODMs, Celestica, Quanta, Wistron, and Hon Hai, that build the finished AI servers.

Which companies represent the Google value chain?

Company	Sector	What it does
Alphabet (GOOGL)	Internet & AI Platform	Parent of Google; designs the TPU accelerators and builds the data centers at the heart of the chain through Google Services, Google Cloud, and Other Bets.
Broadcom (AVGO)	Semiconductors · Custom AI Silicon	Co-designs Google's TPUs and supplies the networking and packaging silicon for Google's AI racks; its custom-accelerator (XPU) business is the chain's chip engine.
Taiwan Semiconductor (TSM)	Semiconductors · Foundry	Manufactures the TPU and the wider chain's leading-edge chips, fabricating Ironwood on its 3nm (N3P) process with advanced CoWoS packaging.
Lumentum (LITE)	Optical · Components & Lasers	Supplies lasers and optical transceivers that move data between TPUs inside and across Google's AI clusters.
Arista Networks (ANET)	Networking · Data-Center Switching	Builds the high-speed Ethernet switches that wire large AI clusters, a layer Google scales alongside its own networking silicon.
Ciena (CIEN)	Optical · Networking Systems	Provides optical transport and data-center interconnect systems that link Google's data centers over long distances.
Marvell Technology (MRVL)	Semiconductors · Data Infrastructure	Supplies data-infrastructure chips, optical DSPs, and custom-silicon design services used across hyperscale AI networks.
Synopsys (SNPS)	Software · Chip Design (EDA)	Provides the electronic design automation tools and design IP that engineers use to tape out custom AI accelerators like the TPU.
Amphenol (APH)	Components · Connectors & Cabling	Makes the high-speed electrical and fiber-optic connectors and cable assemblies that stitch AI servers and racks together.
ASML Holding (ASML)	Semiconductors · Lithography Equipment	Supplies the EUV and advanced lithography systems without which TSMC cannot fabricate the leading-edge silicon used in the chain.
Amkor Technology (AMKR)	Semiconductors · Assembly & Packaging	Provides outsourced semiconductor assembly, test, and advanced packaging that complements foundry output for AI accelerators and memory.

What are the risks?

A single-customer supply chain concentrates both the upside and the danger.

Single-customer dependence. The thesis leans on one company’s capital spending; if Google slows its AI build-out, suppliers across every layer feel it together.
High beta and rich valuations. Many constituents are cyclical semiconductors and contract manufacturers trading at full multiples that leave little room for disappointment.
A possibly peaking market. Omdia flags that AI data-center chip growth is likely peaking even as custom ASICs gain share (Omdia).
Geopolitical and listing complexity. Key suppliers sit in Taiwan, Korea, and China, adding currency, custody, and supply-chain risk that a pure Alphabet position avoids.

TIGER Google Value Chain ETF (0190Y0) · Mirae Asset Global Investments
Korea-listed ETF tracking the Akros Google Value Chain Index, holding Alphabet, Broadcom and the chip, optical and networking suppliers.
Akros Google Value Chain Index (AGOOGVC) · Akros
The benchmark the TIGER ETF tracks: up to 20 Google value-chain companies listed in the US, Korea, China and Taiwan.

Frequently asked questions about the Google value chain

What is the Google (Alphabet) value chain?

It is the chain of public companies that design, fabricate, supply, and assemble the hardware behind Google's AI, from custom TPU chips through foundry, memory, optical, networking, and server suppliers. The concept lets investors own that supply chain rather than only Alphabet stock. The chain is anchored by Alphabet, which guided 2026 capital expenditure to $175-185 billion (CNBC).

Who makes Google's TPU?

Google designs the TPU, but co-designs it with Broadcom, a partnership that dates to 2016 and now runs through 2031 across networking and packaging silicon too (Capacity Media). The chips are fabricated by TSMC; the seventh-generation Ironwood TPU runs on TSMC's 3nm (N3P) process with CoWoS advanced packaging (Tom's Hardware).

How much is Alphabet spending on AI infrastructure?

Alphabet spent $91.4 billion on capital expenditure in 2025 and guided 2026 capex to a range of $175-185 billion, roughly double the prior year (CNBC). That sits inside a broader hyperscaler build-out where combined 2026 capex is tracking toward $700 billion or more (Fortune).

Which suppliers benefit from Google's AI build-out?

Broadcom is the largest single beneficiary on the chip side; its AI semiconductor revenue rose 74% year over year to $6.5 billion in fiscal Q4 2025, with a $73 billion AI backlog (The Motley Fool). Others include TSMC (foundry), Lumentum, Ciena and Arista (optical and networking), Marvell, Synopsys (chip-design software), Amphenol, ASML, and ODMs such as Celestica and Quanta.

What are the risks of investing in the Google value chain?

The chain depends heavily on one customer's capital spending, so any pause in Google's AI investment hits suppliers across layers at once. Many constituents are high-beta semiconductors and contract manufacturers trading at rich valuations, and the market itself may be peaking: one forecast sees the AI data-center chip market reaching $286 billion by 2030 but warns growth is likely slowing (Omdia).

Is there an ETF for the Google value chain?

Yes. In Korea the TIGER Google Value Chain ETF (0190Y0) from Mirae Asset, listed in May 2026, packages the chain into a single listing and tracks the Akros Google Value Chain Index (Mirae Asset). The fund holds 20 names spanning Alphabet, Broadcom, and the optical, networking, and semiconductor suppliers across Google's AI build-out (The Asia Business Daily). Always check fees, holdings, and risk before investing.

Is the Google value chain a good long-term investment for an individual investor?

For a patient investor who already holds broad index funds, it works as a small thematic growth position rather than a core holding or an Alphabet substitute, because the basket leans on one customer's capex and many constituents are high-beta semiconductors at rich multiples. The long-term driver is real spending that has begun landing on suppliers: Alphabet's 2026 capex guidance of $175-185 billion is roughly double its $91.4 billion in 2025 (CNBC), and Broadcom's AI semiconductor revenue already rose 74% year over year to $6.5 billion in fiscal Q4 2025 with a $73 billion backlog (The Motley Fool). Sizing should match a multi-year horizon, since a pause in Google's build-out would hit every layer at once.

How does the Google value chain fit an institutional or thematic AI mandate?

It expresses hyperscaler-capex demand through liquid, identifiable leaders at the custom-silicon, optical, and ODM layers with a clear anchor in Alphabet, so it serves as a targeted sleeve inside a technology or AI-infrastructure allocation rather than a diversified core. Unlike a broad chip index, the thesis is tied to one buyer whose spend is concrete: Alphabet guided 2026 capex to $175-185 billion (CNBC), part of a hyperscaler build-out tracking toward $700 billion or more in 2026 (Fortune). The trade-offs an allocator weighs are heavy overlap with broad semis, single-customer concentration, and the custody and tracking complexity of names cross-listed in Taiwan, Korea, and China.

Is the Google value chain a sensible way for an advisor to add AI exposure for a client?

It lets a fiduciary spread a client's AI bet across Google's whole supply chain instead of a single mega-cap, capturing the build-out even if Alphabet's own margins compress, and the holdings are liquid enough to model and rebalance. The structural tailwind is durable spending: Alphabet's 2026 capex guidance of $175-185 billion roughly doubles its 2025 outlay (CNBC). It is not a substitute for diversified core equity, though, because the market backdrop is no longer one-way: Omdia puts the AI data-center chip market at $286 billion by 2030 but flags that overall growth is likely peaking (Omdia), so suitability depends on the client's risk tolerance and existing exposures.

Sources & references

Tech AI spending approaches $700 billion in 2026, cash taking big hit · CNBC, 2026-02-06
Expanding our use of Google Cloud TPUs and Services · Anthropic, 2025-10-23
Google and Anthropic confirm massive 1GW+ cloud deal with up to one million Google TPUs · Data Center Dynamics, 2025-10-23
Broadcom (AVGO) Q4 2025 Earnings Call Transcript · The Motley Fool, 2025-12-12
Broadcom locks in long-term Google TPU deal through 2031 · Capacity Media, 2025-10-24
Ironwood: The first Google TPU for the age of inference · Google, 2025-04-09
The custom AI ASIC state of play (May 2026) · Broadcom deals, Google TPUs, Meta MTIA & beyond · Tom's Hardware, 2026-05-01
AI data center chip market to hit $286bn, growth likely peaking as custom ASICs gain ground · Omdia, 2025-08-01
Big Tech is about to spend $700 billion on AI this year · Fortune, 2026-04-30
TIGER Google Value Chain ETF (0190Y0) — product page · Mirae Asset Global Investments, 2026-05-12
Mirae Asset lists 'TIGER Google Value Chain ETF' for one-ticket exposure to Google's AI ecosystem · The Asia Business Daily, 2026-05-12