SK hynix gets offers to fund AI memory expansion

SK hynix, a South Korean memory chipmaker, has received offers from large tech firms to invest in new production lines.

These companies are also offering to help pay for manufacturing tools as they race to secure scarce AI memory chips.

The proposals include funding dedicated lines and financing equipment such as ASML extreme ultraviolet lithography machines.

However, SK hynix is wary of deals that could tie it to specific buyers or force lower chip prices.

The company has no spare capacity to set aside for any customer, including at the first phase of its Yongin fabrication plant in South Korea.

SK hynix, Samsung Electronics, and Micron have said they are discussing multi-year supply contracts with customers as the memory shortage continues amid strong AI demand.

🔗 Source: Reuters

🧠 Food for thought

Implications, context, and why it matters.

New supply deals are shifting toward upfront payments and price floors

These multi-year supply contracts mark a change from earlier deals, when customers mostly used them to guard against price swings 1.
Negotiations involving Microsoft and Google include upfront payments of about 10% to 30% of total contract value, up from an earlier norm below 5% 2.
Some new three-to-five-year agreements also include “price floor” clauses that set a minimum price for the full term 3.
The shift comes from a new priority. Locking in chip supply now matters more to customers than price volatility, which gives suppliers more bargaining power 1.

The AI memory crunch may be increasing financial pressure for some cloud providers

The memory shortage can choke the wider AI hardware market, including companies whose AI server shipments rely on High Bandwidth Memory (HBM), a type of advanced memory used in AI chips 4.
That strain can move more of the financial burden to cloud service providers that keep adding AI infrastructure.
Oracle’s capital spending rose more than 200% year over year, while free cash flow fell to negative US$10 billion as the company rushed to secure AI infrastructure 5.
This can create a shaky financing chain. Cloud companies borrow heavily to buy chips and build data centers, then lease that capacity to venture-backed startups. The model becomes vulnerable if end-user monetization does not arrive 5.