ReportGem ReportGem EN

普通外文研报

Cerebras Systems: Unique capability in fast inference drives large economic opportunity; Initiate at OW

发布日期: 2026-06-08研究机构: Morgan Stanley公司 / 股票: CBRS.OQ报告页数: 36原文语言: 英语证据页码: 2

研报英文原文证据摘录

Cerebras Systems: Unique capability in fast inference drives large economic opportunity; Initiate at OW

FoundationMthat create very high value.

Contracted revenue opportunities get them to $6 bn revenue - approximately 1%

market share of overall processors - by CY28, and we are optimistic that there

can be upside. The two leading processing companies in the AI space, NVIDIA and

Broadcom, have over $7 trillion in market value between them. Even with lower

margin potential of a niche solution, we think that every 1% share can be worth $50-

70 bn, and that's the base case here. Sales from Open AI - which should make up

~80% of revenues in the next few years - are contracted over the next 3 years, and

talking to customers there is clearly demand for more hardware if they can get

ahead of supply constraints.

We also expect competition in the category to intensify over time. In particular,

NVIDIA's acquisition of the assets of Groq - an AI processing company specializing in

a low latency approach - is worth watching. While this deal does validate the

category for Cerebras, and NVIDIA does not have wafer scale capability, they can

bring the world class networking capabilities of its Mellanox business, as well as

integration with its dominant GPU franchise, to bring a competitive offering. The size

and profit of this opportunity will draw competition, but Cerebras first mover

advantage and customer partnership is a good starting place.

We derive our $250 price target using a 12x multiple on our 2028 revenue estimate

adjusted for warrant contra revenue. The 12x sales multiple is in-line with SMID cap

AI names with near-100% AI exposure that are outgrowing the overall market.

While Cerebras operates a cloud-delivered infrastructure platform, we believe it is

本摘录由系统从所标注的 PDF 证据页直接提取并保留英文原文,不做批量翻译;登录后在阅读器切换中文时才按需翻译。

打开研报阅读器