COMPANIES · AI CHIPS & NETWORKING

Broadcom

Designs semiconductors and provides infrastructure software, including custom AI accelerators and networking products.

DirectAI chips & networking
What it does

Designs semiconductors and provides infrastructure software, including custom AI accelerators and networking products.

Why it matters

It designs custom AI accelerator chips ('XPUs') for hyperscalers such as Google and Meta, giving those companies an alternative to buying finished GPUs from NVIDIA.

How it makes money

Sells semiconductors (including custom AI chips and networking silicon) and infrastructure software, increasingly with AI-related products as a large and fast-growing share of revenue.

Geography

United States

AI stack position

AI chips and networking. The companies designing the processors, accelerators and networking equipment that do the actual computing behind AI, including GPUs like NVIDIA's, and the high-speed connections that link thousands of them together.

Connections (11)

Every connection below is tied to a readable source, date, classification and statement of its limits. A company source establishes what that company stated; it does not necessarily provide independent confirmation.

Current relationships (2)

Broadcom co-developing multiple generations of Meta's MTIA custom AI silicon Meta Platforms
Co-develops withCurrent
Broadcom and Meta extended their partnership for Broadcom to help develop and deploy technology supporting multi-gigawatt scale of Meta's custom MTIA AI silicon across future chip generations.

Source: 2026-04-14 · reviewed 2026-09-20

Evidence: Primary-source statement · Extended, ongoing multi-generation custom-silicon collaboration

Limits: Describes an extended development partnership and a multi-gigawatt capacity target; does not confirm current production volumes, unit economics, or which foundry fabricates the resulting chips.

Broadcom Announces Extended Partnership with Meta to Deploy Technology to Support Multi-Gigawatts of Meta's Custom Silicon, MTIA ↗
TSMC manufactures custom AI ASICs for Broadcom
Manufactures forCurrent
TSMC manufactures Broadcom's latest custom AI ASICs on its 2-nanometer process using 3.5D advanced packaging, according to trade-press reporting on Broadcom's custom-silicon roadmap.

Source: 2026-03-05 · reviewed 2026-09-22

Evidence: Secondary reporting (unconfirmed by primary parties) · Reported operating

Limits: Trade-press reporting rather than a direct statement from TSMC or Broadcom; no volumes, contract value, or product names are disclosed.

Broadcom-TSMC 3.5D AI chips give ASIC leader an early edge over Nvidia ↗

Dated or announced (6)

Samsung Electronics signs MOU to supply memory, foundry and packaging to Broadcom
SuppliesDated / announced
Samsung Electronics and Broadcom signed an MOU under which Samsung will supply HBM for Broadcom's next-generation AI accelerators, manufacture Broadcom wireless-communications chips on its 2nm-and-below foundry process, and provide 2.3D/2.5D advanced packaging, in a partnership valued at over $200 billion through 2030.

Source: 2026-07-25 · reviewed 2026-09-20

Evidence: Primary-source statement · MOU signed July 2026; deliveries and production phased through 2030, not yet a confirmed operating supply relationship

Limits: Reflects a signed MOU with a 2030 horizon; does not confirm that HBM shipments or foundry production for Broadcom have begun, nor how the $200B figure breaks down by year or product line.

Samsung Electronics and Broadcom Expand Strategic Collaboration Across Memory and Foundry Technologies ↗
Samsung Electronics will manufacture 2nm chips for Broadcom
Manufactures forDated / announced
Samsung Electronics and Broadcom signed an MOU valued at more than $200 billion across memory and foundry through 2030, under which Samsung's 2-nanometer-and-below foundry process would manufacture Broadcom products including wireless broadband communications solutions, alongside 2.3D/2.5D advanced packaging.

Source: 2026-07-25 · reviewed 2026-09-22

Evidence: Primary-source statement · Announced / planned

Limits: This is a memorandum of understanding, not a completed or binding supply contract; the $200B figure spans both memory and foundry over five years and is not a confirmed, itemized order.

Samsung Electronics and Broadcom Expand Strategic Collaboration Across Memory and Foundry Technologies ↗
Broadcom supplies silicon implementation and networking for the custom chips of OpenAI
SuppliesDated / announced
OpenAI said Broadcom provides the silicon implementation, manufacturing and Tomahawk networking behind Jalapeno, OpenAI's first custom inference accelerator, with initial deployment targeted by the end of 2026.

Source: 2026-06-24 · reviewed 2026-09-24

Evidence: Primary-source statement · Announced; not confirmed operating

Limits: Announcement of a chip and a deployment target. It does not establish that any Jalapeno units have been produced or installed, the volumes or prices involved, or that Broadcom is OpenAI's only silicon partner.

OpenAI and Broadcom unveil LLM-optimized inference chip ↗
Broadcom co-developing and supplying Google's TPUs through 2031 Google (Alphabet subsidiary)
Co-develops withDated / announced
Broadcom and Google extended their long-standing collaboration in which Broadcom co-designs and helps manufacture Google's custom TPU chips, with the arrangement now running through 2031.

Source: 2026-04-06 · reviewed 2026-09-20

Evidence: Secondary reporting (unconfirmed by primary parties) · Extended multi-year co-development and supply arrangement disclosed via securities filing and press reporting

Limits: Reporting on a securities disclosure describes contract duration and scope; it does not itself state unit volumes, dollar value, or exclusivity of the TPU relationship.

Broadcom agrees to expanded chip deals with Google, Anthropic ↗
Broadcom supplies custom silicon powering Google TPU capacity committed to Anthropic
Provides infrastructure toDated / announced
As part of an expanded Google-Broadcom-Anthropic compute agreement announced April 6, 2026, Anthropic will gain access to multiple gigawatts of Google TPU capacity (co-designed with Broadcom) beginning in 2027, building on capacity already coming online in 2025. This record does not establish the precise gigawatt breakdown between the initial and expansion tranches, and Broadcom's role is as Google's custom-silicon co-design partner, not as a direct compute provider to Anthropic.

Source: 2026-04-06 · reviewed 2026-09-20

Evidence: Primary-source statement · New compute commitment announced by Anthropic; capacity to come online starting 2027

Limits: Confirms an expanded Google-Broadcom-Anthropic compute agreement and Broadcom's role as Google's TPU co-design partner; does not itself establish the precise gigawatt figures for each tranche, and should not be read as Broadcom directly renting or leasing compute capacity to Anthropic -- Google is the compute provider, Broadcom the silicon co-design partner.

Anthropic expands Google and Broadcom compute deal ↗
Broadcom co-developing 10 gigawatts of OpenAI-designed custom AI accelerators OpenAI
Co-develops withDated / announced
OpenAI and Broadcom announced a strategic collaboration to co-develop and deploy 10 gigawatts of OpenAI-designed AI accelerators together with Broadcom Ethernet networking, with rack deployment targeted from the second half of 2026 through 2029.

Source: 2025-10-13 · reviewed 2026-09-20

Evidence: Primary-source statement · Definitive collaboration announced; deployment phased 2026-2029

Limits: Confirms a signed co-development and deployment plan; does not confirm current shipped accelerator volume, contract dollar value, or exclusivity.

OpenAI and Broadcom announce strategic collaboration to deploy 10 gigawatts of OpenAI-designed AI accelerators ↗

Comparisons (3)

Marvell Technology competes with Broadcom
Competes withComparison only
Marvell and Broadcom are reported as rival design partners for hyperscaler custom AI silicon; Broadcom shares fell more than 5% after Marvell's expanded custom-chip agreement with Google was announced.

Source: 2026-08-20 · reviewed 2026-09-22

Evidence: Analytical comparison · Ongoing market competition

Limits: Market price reaction is used as evidence of perceived competitive overlap, not a statement by either company; does not establish relative market share.

Google-Marvell Deal Expands Custom Silicon Past the TPU ↗
Astera Labs competes with Broadcom
Competes withComparison only
Industry analysis of Astera Labs' IPO identifies Broadcom and Marvell as the dominant incumbents in datacenter connectivity (PCIe retimers, CXL controllers), holding more than 80% combined revenue share, against which Astera Labs competes as a specialized entrant.

Source: 2024-03-17 · reviewed 2026-09-22

Evidence: Analytical comparison · Ongoing market competition

Limits: Reflects a 2024 market-structure analysis at the time of Astera Labs' IPO; does not reflect any later shift in relative market share.

Astera Labs IPO - The Next Connectivity Superhero or Steamrolled By Competition? ↗
Broadcom enables custom alternatives to NVIDIA
Competes withComparison only
Broadcom designs custom AI accelerator ('XPU') silicon for large customers, offering an in-house alternative to buying finished NVIDIA GPUs.

Source: Source undated · reviewed 2026-09-17

Evidence: Analytical comparison · Product comparison

Limits: Broadcom's custom-silicon business model is structurally different from NVIDIA's merchant-GPU model; this is a market-structure comparison, not a bilateral agreement.

Broadcom AI accelerator (XPU) custom compute ↗
Evidence for this profile

Broadcom Q3 FY2026 results ↗
Reviewed

Explore it interactively

Open this profile in the database See it on the relationship map

Inclusion in this database is not an endorsement of a company or its securities. Nothing here is a recommendation to buy or sell anything, and none of it is individualized investment advice. See the disclosures. Data: AI Map v3.22, released .