Advanced Packaging Is Escaping CoWoS Scarcity—But Nvidia’s “$1.5B to Amkor” Deal Is Not Verified in Primary Sources Yet

What we can confirm vs. what we can’t

The headline claim ($1.5B, multi-year Nvidia→Amkor) did not clear verification in-session

I could not verify the exact “Nvidia’s $1.5B multi-year Amkor advanced-packaging deal” from primary sources in this session. SEC filings search calls errored, and key navigations returned no usable content, so the load-bearing dollar figure and contract details are treated as unconfirmed.

This article therefore pivots to a rigor-first version of the thesis: advanced AI packaging is the scarce link, OSAT capacity expansions matter, and Nvidia’s manufacturing model makes routing decisions central. But the specific deal size ($1.5B), timing, and whether it truly bypasses TSMC CoWoS cannot be stated as fact without the primary documents.

Verified base layer

Nvidia doesn’t package in-house—so its supply chain routing is structurally “OSAT-dependent”

A key enabling fact for any “routing around CoWoS” argument is that Nvidia does not run packaging/test inside its own fabs. Instead, it relies on external manufacturing/subcontractors for packaging and testing—meaning that contract structure with OSATs can materially change lead times and capacity availability.

Nvidia (overview)

Contract-manufacturing model (no in-house packaging claimed here)

Nvidia overview retrieved via data tool indicates it distributes products via an ecosystem of OEMs and subcontractors; packaging/test are handled by independent subcontractors (exact contract numbers not verified this session).

Verified company roles

Amkor is the kind of supplier that can add advanced-package capacity (OSAT role is real)

Amkor is a major external semiconductor packaging and test provider. OSATs like Amkor offer advanced packaging services (including advanced system-in-package and flip-chip/stacked packages, depending on customer and platform), which makes them a natural “second routing path” when a foundry’s advanced packaging nodes are constrained.

What Amkor is (service scope) vs. what this session could not verify (Nvidia-specific deal size/scope)
Layer	Verified in-session?	What we know	What we don’t know
OSAT role	Yes	Amkor provides external packaging and test services (end-to-end packaging/test and advanced packaging offerings).	Specific technologies used in the alleged Nvidia contract (not disclosed/verified).
Deal size ($1.5B)	No	No primary document confirming the $1.5B figure was retrieved.	Treat as unconfirmed.
Multi-year contract	No	No primary document confirming multi-year duration was retrieved.	Treat as unconfirmed.
CoWoS routing bypass	No	Routing around CoWoS is plausible in concept if OSAT capacity expands, but the specific mechanism for Nvidia→Amkor is not verified here.	We cannot state that the deal routes around TSMC CoWoS without evidence.

Data we can ground in fundamentals

Amkor’s financial profile signals capital intensity typical of packaging expansions—but this session can’t tie it to the specific alleged Nvidia deal

Even without the deal details, investors can still read OSAT build-out cycles in results: packaging/test providers often invest ahead of volume ramps. In this session, the fundamentals tools returned Amkor and Nvidia income statements and key metrics, but they cannot be used to retroactively prove a specific $1.5B contract.

Nvidia revenue growth (FY 2023–FY 2025, tool data)

This is not proof of an Amkor contract. It only anchors the macro context of why advanced packaging capacity would be strategically contested.

Unit: USD (reported)

FY 2023

26,974,000,000

FY 2024

60,922,000,000

FY 2025

130,497,000,000

Amkor revenue (FY 2023–FY 2025, tool data)

Again: not attributable to a specific Nvidia contract without primary evidence, but useful for spotting the direction of packaging/test demand.

Unit: USD (reported)

FY 2023

6,503,065,000

FY 2024

6,317,692,000

FY 2025

6,707,981,000

Second-order winners/losers framework (what to look for)

If routing truly shifted away from CoWoS, the second-order beneficiaries would be OSAT + packaging substrate + advanced test—here’s the checklist

OSAT capacity substitution: confirm whether the agreement specifies “advanced packaging and test” with processes that can be executed outside TSMC’s CoWoS lines (requires the contract/pricing schedule).
Substrate demand signal: advanced packaging shifts demand toward high-end interposers/substrates (materials, photolithography/lamination, patterning, fine-pitch build). Look for supplier capex or customer-awarded capacity (needs primary sources).
Test throughput: if advanced packaging volume grows, wafer sort/final test capacity becomes a gating constraint; “second-order winner” likely tilts to automated test handling and high-throughput ATE ecosystem (requires deal-linked disclosures).
Lead time arbitrage: if routing changes, delivery schedules improve first—watch procurement lead times and backlog commentary in SEC filings (SEC search failed in-session).
Customer concentration: verify whether the agreement is for Nvidia-specific platforms or broader “strategic capacity” that various customers can draw from—without that, you can’t assign attribution.

Because the Nvidia→Amkor $1.5B figure wasn’t verified here, the supply-chain “routing around CoWoS” implication remains a framework, not a confirmed outcome.

Near-term vs long-term investor implications (conditional)

Market timing: advanced-packaging deals matter most when they change throughput before they change margins

If an OSAT deal truly adds advanced packaging/test capacity, it tends to show up first in fulfillment timing (less “weeks/months late” inventory flow) and later in financials (margin and utilization). The practical investor move is to monitor contract-linked operating metrics and capex/expansion completion milestones—rather than relying on the headline contract value.

Short-term (days–quarters): watch for backlog commentary, guided capacity starts, and any mention of “advanced packaging and test” yield/throughput ramp in company disclosures (not retrievable here due to SEC tool errors).
Long-term (1–3 years): the key checkpoint is whether OSAT capacity converts scarcity into normalized utilization, which compresses “capacity premium” pricing across competing OSATs and substrates.

Synthesis

What this article can conclude today: OSAT capacity is the real lever; the specific $1.5B deal needs primary confirmation

The credible investment thesis behind the brief is directionally sound: Nvidia’s external manufacturing model makes it sensitive to packaging bottlenecks, and OSATs like Amkor are structurally positioned to expand advanced packaging/test capacity. However, this session cannot responsibly claim that a specific “$1.5B multi-year Nvidia→Amkor deal” was signed or that it routes around TSMC CoWoS, because the necessary primary sources were not accessible/verified.

Without the signed contract language (or SEC/IR disclosure that states deal value, duration, and process scope), any “CoWoS routing” conclusion is speculation rather than research.
Deep Research Agent (this session)