Samsung says neoclouds are about 8% of the AI market and rejects AI bubble concerns

AI & Compute

Samsung puts neoclouds at 8% and backs structural AI demand

Samsung estimates neoclouds account for about 8% of the AI market, arguing that long-term chip contracts—not speculation—anchor demand while debt-funded data centers remain vulnerable.

By
Rebecca Stone
Filed
Channel
AI & Compute
Read
4 min read

By Rebecca Stone

Samsung Electronics chip-business president Kim Yong-kwan estimated on Wednesday that neoclouds account for approximately 8% of the overall AI market, rejecting concerns that debt-funded data-center expansion could tip AI spending into a bubble.

What does the 8% estimate cover?

At a Morgan Stanley conference in Singapore, Kim called the segment a limited part of the market and said its financing problems should not threaten broader AI growth. Samsung, the world's largest memory-chip maker, issued that estimate without a disclosed revenue or capacity figure; the report named no process node or memory product family.

“Neoclouds represent a relatively smaller and limited portion of the overall market, approximately 8%,” Kim said. “We do not expect these concerns to pose a significant threat to the broader AI growth trajectory.”

Kim serves as president and head of corporate management, strategy and operation at Samsung's chip business. His comments separate two risks: the financial health of specialized compute renters and the demand outlook for chip suppliers. The distinction matters. The 8% figure was Kim's assessment, not a disclosed industry revenue share.

What does contract visibility change?

Samsung's argument rests on the structure of current AI spending rather than a forecast of uninterrupted memory prices. Kim called the build-out structural, not cyclical, and said long-term agreements between chipmakers and customers give companies more visibility into demand and supply than they had during earlier memory cycles.

That visibility matters because past rounds of speculation ended in overcapacity, according to Kim. Instead of relying on the spot market, chip suppliers and their customers are committing to each other over longer periods, reducing reliance on near-term forecasting.

“In short, today's investment is not a result of speculation, but actual tangible demand,” Kim said, according to Reuters.

The source did not disclose contract lengths, prices, memory product families or committed volumes. Nor did Kim quantify how much contracted AI-memory demand could offset unsold capacity. His case therefore depends on the contracts remaining firm as customers deploy servers and data centers.

Where does neocloud risk remain?

Neoclouds such as CoreWeave rent computing capacity. Reuters reported that their shares have been pressured by concern that debt-funded data-center expansion could unwind if AI demand slows.

The financing channel adds a second source of volatility. Some analysts and investors told Reuters that the departure of Jensen Huang as Nvidia CEO could tighten funding for smaller, highly leveraged players. That remains a market concern attributed to those investors, not a quantified forecast of credit losses or canceled chip orders.

Nvidia's Tuesday results also lifted SK Hynix shares and the Kospi on Wednesday, according to the report's linked coverage. It gave no percentages for either move and did not quantify how much AI demand sits with neoclouds.

Could AI memory demand be counted twice?

Samsung's own customer discussions have raised a more immediate question. Last week, a Korean paper reported that OpenAI's Sam Altman told Samsung executives that some AI-memory demand may have been double counted. The report did not specify the memory products, customer volumes or period involved. The duplication remains unquantified.

That caution sits alongside Kim's assertion of tangible demand. One statement addresses the basis of current orders; the other questions whether some orders represent distinct consumption. Neither resolves the other, but together they sharpen the need to compare AI infrastructure announcements with actual component shipments.

What will Samsung disclose in January?

Samsung also repeated that it will return the remainder of its planned shareholder payout through cash dividends and a share repurchase and cancellation program that the company described as substantial. Details are due in January after Samsung weighs factors including capital spending.

No dividend amount, repurchase value or share cancellation count was disclosed. The timing nevertheless links Samsung's shareholder return to the same capital-allocation choices facing a broad AI-memory expansion. Investors will get clearer evidence on cash commitments and spending priorities when Samsung publishes the specifics in January.

If long-term contracts remain intact, they could restrain the overcapacity risk that followed earlier memory cycles; if leveraged neocloud expansion weakens, financing rather than chip orders may become the first pressure point.

January's payout details will show how Samsung balances capital spending against shareholder returns.

Source: mlq.ai

Share this article:

More from Rebecca Stone

Rebecca Stone

Show full bio

Correspondent covering technology and product news in Semiconductors, chip manufacturing, hardware, AI compute at Chip Dispatch.

1,097 articles

Related articles

« Previous article