Wall Street Recovers, Amazon Explores Selling AI Chips

CNBC's Investing Club reports that US equities recovered most losses from Wednesday's Federal Reserve meeting, with the S&P 500 up about 1% for the week, CNBC reports. Crude oil prices fell on headlines about oil tankers exiting the Strait of Hormuz, CNBC adds. CNBC reports Apple CEO Tim Cook indicated iPhone price increases as memory and storage chip costs climb, a signal markets read as persistent memory tightness; that helped gains in Sandisk, Western Digital, Applied Materials, and Lam Research, per CNBC. CNBC also says Amazon shares jumped after Bloomberg reported the company is in talks to sell its custom chips to third-party data centers. CNBC notes Amazon's custom-silicon family includes Graviton, Trainium, and Nitro, and cites Google monetizing its silicon via a deal to supply Anthropic with TPUs.
What happened
CNBC's Investing Club reports that US markets recovered most of Wednesday's losses after the Federal Reserve meeting, leaving the S&P 500 about 1% higher for the week, CNBC reports. CNBC reports crude prices fell on headlines that oil tankers are exiting the Strait of Hormuz. CNBC reports comments from Apple CEO Tim Cook indicating Apple may raise iPhone prices as memory and storage chip costs climb, a development markets read as evidence memory prices are not falling. CNBC reports those dynamics helped stocks such as Sandisk, Western Digital, Applied Materials, and Lam Research rally. CNBC also reports that Amazon shares rose after Bloomberg reported the company is in talks to sell its custom chips to third-party data centers, and CNBC lists Amazon's custom-silicon family as Graviton, Trainium, and Nitro. CNBC additionally notes Google has begun monetizing custom silicon, citing a past deal to supply Anthropic with TPUs.
Editorial analysis - technical context
Cloud providers and hyperscalers increasingly treat custom silicon as both an operations tool and a commercial product. Companies that have already moved to monetize chips, such as Google with Tensor Processing Units, provide a precedent for third-party sales that reduce customers' need to buy generic accelerator instances. Industry-pattern observations: monetizing in-house silicon typically requires software ecosystems, deployment tooling, and billing integrations that match existing cloud offerings, which raises integration and support costs for the seller while offering customers potential performance or price advantages.
Industry context
For the semiconductor value chain, persistent memory-price inflation amplifies demand across NAND/DRAM makers and boosts capital equipment vendors, a pattern CNBC highlights with the rally in equipment and materials suppliers. Observed patterns in similar market cycles: when component shortages persist, investors rotate into upstream capital-equipment names and materials suppliers ahead of broad fab-capacity expansion announcements.
What to watch
Indicators to follow include a confirmation from Amazon or Bloomberg on any chip-sales agreement; public pricing or terms if third-party sales proceed; announcements of software stacks or APIs to support external customers for Trainium/Graviton workloads; and memory pricing trends that continue to influence hardware vendors' revenue. Industry observers will also watch partner deals that mirror Google's Anthropic TPU arrangement as a barometer for commercial demand.
Practical takeaway
The story ties macro market moves to supply-chain dynamics and to a broader industry shift where cloud providers consider commercializing in-house accelerators. Readers tracking infrastructure and procurement should watch for formal announcements and technical integration details that would determine commercial viability.
Scoring Rationale
The Bloomberg June 18 scoop - Amazon AI chief Peter DeSantis confirming the company is actively in talks to sell Trainium chips directly to third-party data centers - is a concrete step forward from Jassy's April hint and is meaningful for cloud/AI infrastructure practitioners. The event is a market-roundup wrapper, which tempers the pure AI signal, making it a notable-tier rather than major-tier story.
Practice with real Retail & eCommerce data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Retail & eCommerce problems


