✦ The OpenAI Deal & the 10-Gigawatt OrderNarrow moat

Broadcom (AVGO) — the future bets

Ten gigawatts promised by 2029 in the currency of 2025's enthusiasm — the order that reframed Broadcom, priced before a rack has landed.

The October 2025 announcement read like a treaty: OpenAI designs the accelerators and the systems, Broadcom develops and deploys them — ten gigawatts' worth, racks scaled entirely with Broadcom Ethernet, first deployments in the second half of 2026 and completion targeted by the end of 20291. It slotted into OpenAI's larger buying spree — roughly 26 gigawatts committed across Nvidia, AMD and Broadcom — and the market added about 9% to Broadcom's value the morning it learned who the mystery customer was2.

The 10-gigawatt calendarOct 2025 · dealannouncedH2 2026 ·first racks2027–28· the rampEnd 2029 ·10GW completeOpenAI designs; Broadcom builds and networks — inside OpenAI's ~26GW, 3-vendor spree.
A treaty with a timeline: the market repriced Broadcom ~9% on the announcement — delivery, not announcement, now has to earn it.

What makes the order a bet rather than a booking is the counterparty. OpenAI is the most ambitious spender in the industry and also its most creatively financed: the racks serve a company whose commitments dwarf its revenue, and ten gigawatts is a promise about 2029 made in the currency of 2025's enthusiasm. Broadcom's protection is structural — the customer owns the design and the capex while Broadcom collects on delivery — but a scaled-back OpenAI would still shrink the backlog that justifies the multiple. Watch the H2 2026 first racks land on time, and watch each subsequent gigawatt get funded, because the difference between 10GW announced and 10GW energized is measured in tens of billions of dollars.

Moat trajectory: Widening

Widening by announcement, pending by delivery: the 10GW order and the ~26GW OpenAI spree around it moved the whole custom-silicon market toward Broadcom. The arrow holds only as long as OpenAI's financing does — first racks landing on schedule in H2 2026 is the checkpoint.

The number that tests this moat
Moat Explorer calc
Broadcom's share of OpenAI's announced compute
10GW of ~26GW, about 38%

OpenAI has spread its commitments across three chip suppliers, and the first Broadcom racks are due in the second half of 2026. A shrinking slice, or a total that OpenAI cuts, would show the order moving elsewhere before it becomes revenue.

How it's calculated: Broadcom's 10GW divided by the ~26GW OpenAI has committed across Broadcom, Nvidia and AMD.
Source: CNBC, OpenAI partners with Broadcom (13 Oct 2025) ↗
References
  1. Reported10GW of OpenAI-designed accelerators on Broadcom silicon and Ethernet; H2 2026 first racks, end-2029 completion.
    OpenAI–Broadcom strategic collaboration (Oct 13, 2025): co-develop and deploy 10GW of OpenAI-designed AI accelerators; racks scaled entirely with Broadcom Ethernet; first deployments H2 2026, completion by end-2029 — Announced Oct 2025; deployments H2 2026–2029 · publ. Oct 13, 2025 · source ↗
  2. ReportedBroadcom rose ~9% on the reveal; OpenAI's commitments total ~26GW across three vendors.
    CNBC — Broadcom +~9% on the OpenAI custom-chip deal; OpenAI's total hardware commitments ~26GW across Nvidia, AMD and Broadcom — Oct 2025 · publ. Oct 13, 2025 · source ↗
Sources
Generated September 22, 2026