Meta Signed 2 AI-Chip Partnerships With AMD and Broadcom in 50 Days
Advanced Micro Devices disclosed a 6-gigawatt GPU partnership with Meta in February 2026. Seven weeks later, Broadcom announced a multi-gigawatt extension covering Meta's custom MTIA silicon. Both companies' IR press releases — sourced directly from SEC-registered feeds — land in the same 90-day window as four additional gigawatt-scale AI-infrastructure partnerships from NVIDIA and Caterpillar. For retail investors tracking who's winning hyperscaler AI spending, Meta's two chip picks tell a specific story.
Meta appears as the named counterparty on two public-company AI-chip partnership announcements filed in the past 50 days: AMD's 6-gigawatt GPU deal (February 24, 2026) and Broadcom's multi-gigawatt custom-silicon extension (April 14, 2026). Both surface as ir_partnership signals in MetricDuck's index — and both land in the same 90-day window as two additional gigawatt-scale AI-infrastructure partnerships from NVIDIA (Thinking Machines Lab) and Caterpillar (American Intelligence & Power).
For retail investors trying to figure out which public companies are winning Meta's AI-infrastructure spending, the answer for early 2026 is not "one vendor." It's two — and they were announced seven weeks apart.
AMD: 6 gigawatts of GPUs, announced February 24
Advanced Micro Devices (AMD) disclosed the first of the two deals. The company's IR release put a specific number on it: six gigawatts. AMD trades at $284.49 as of April 21, 2026, up 42.63% in the past month and 231.73% over the trailing year; revenue over the trailing twelve months is $34.64 billion, up 34.34% year-over-year per the most recent 10-Q filing (period ending 2025-12-27).
AMD and Meta Announce Expanded Strategic Partnership to Deploy 6 Gigawatts of AMD GPUs
The deal is labeled a strategic partnership in MetricDuck's IR signal index. AMD also announced a separate strategic partnership with Nutanix the next day (February 25, 2026) focused on enterprise AI platforms — so the Meta announcement was not an isolated move but the larger of two back-to-back strategic deals for AMD's data-center business.
Broadcom: multi-gigawatt MTIA extension, announced April 14
Broadcom (AVGO) followed on April 14, 2026, with an extension of an existing relationship. The scale language in the title — "multi-gigawatts" — is less specific than AMD's 6-gigawatt figure, but the reference to Meta's "Custom Silicon, MTIA" is concrete: MTIA is Meta's in-house ASIC program, and Broadcom's custom-silicon business has been the named partner for that effort.
Broadcom Announces Extended Partnership with Meta to Deploy Technology to Support Multi-Gigawatts of Meta's Custom Silicon, MTIA
Broadcom trades at $402.17, up 23.78% in the past month; revenue is $68.28 billion over the trailing twelve months (+25.22% YoY) at a 67.82% gross margin per the most recent 10-K (period ending 2026-02-01). Notably, Broadcom's latest SEC filing is an 8-K dated the same day as this article's publication — an 8-K often carries material-agreement disclosures, though MetricDuck's current signal index does not cross-reference 8-K filings to IR partnership announcements.
The broader gigawatt-scale partnership wave
The Meta deals do not sit in isolation. A 90-day screen of ir_partnership signals surfaces four public companies with gigawatt-scale AI-infrastructure partnerships filed since late January 2026:
NVIDIA (NVDA) remains the incumbent: the same 90-day window carries four distinct NVDA partnerships in MetricDuck's index — Hyundai Motor + Kia (automotive AI, March 16), Thinking Machines Lab ("gigawatt-scale," March 10), Coherent (data-center optics, March 2) and Lumentum (data-center optics, March 2). Only the Thinking Machines Lab deal is framed in gigawatt language; the optics partnerships support AI data-center build-outs without specifying scale. NVIDIA's last 10-K guides Q1 FY2027 revenue to $78 billion ± 2% with 74.9–75% gross margins — the benchmark backdrop against which AMD and Broadcom's Meta wins will be evaluated. Caterpillar's January 28 deal with American Intelligence & Power added a different layer: two gigawatts of dedicated power for hyperscale AI data centers — a reminder that the chip deals sit on top of a separate physical-infrastructure build-out.
Get Quarterly Updates
We update this analysis every quarter after earnings. Subscribe to get notified when Q4 2025 data is available (February 2026).
4 emails/year. Unsubscribe anytime. No spam.
What retail investors should watch
Three concrete next steps:
- AMD and Broadcom earnings transcripts. AMD's last reported earnings were Q4 2025 (filed 2026-02-03); Broadcom last reported Q1 FY2026 (filed 2026-03-04). Management commentary on the Meta-deal ramp will be a first-read for whether "6 gigawatts" and "multi-gigawatts" translate to near-term revenue or multi-year commitments.
- SEC 8-K Item 1.01 filings. Material-agreement disclosures on the AMD-Meta and AVGO-Meta deals would confirm contract scope. Broadcom filed an 8-K on 2026-04-21 — worth reading for Meta-deal specifics.
- The next Meta chip announcement. Two deals in 50 days is a tempo. If Meta discloses a third public-company partnership in the next 90 days, it signals a diversification strategy rather than dual-sourcing.
FAQ
Who is winning Meta's AI chip spending in 2026?
Based on IR press releases filed in early 2026, Meta has disclosed multi-gigawatt partnerships with both AMD (February 24, 6 gigawatts of GPUs) and Broadcom (April 14, multi-gigawatts of MTIA custom silicon). Both deals are classified as strategic partnerships in their respective IR filings. No corresponding Meta-NVIDIA partnership surfaces in the same 90-day window per MetricDuck's IR signal index, though NVIDIA remains the broader GPU incumbent.
What does "multi-gigawatt" mean in an AI chip partnership?
Gigawatts refer to the electrical load the deployed compute will draw at scale — a proxy for physical data-center footprint and chip volume. AMD's 6-gigawatt figure implies hyperscaler-scale deployment. Broadcom's "multi-gigawatt" language and NVIDIA's "gigawatt-scale" deal with Thinking Machines Lab use similar framing. MetricDuck does not yet extract these figures as structured atoms — they appear here sourced from IR press-release titles verbatim.
Where does this partnership data come from?
MetricDuck's ir_partnership signal indexes strategic-partnership announcements from public companies' IR press-release feeds. Each row carries ticker, announcement date, partner names, and relationship type. This article used a 90-day signal screen executed April 21, 2026, returning 29 rows across 13 tickers (25 after dedup). Primary sources are each company's official IR press release.

MetricDuck Research
Structured filing analysis from MetricDuck's SEC intelligence pipeline.