The Zhitong Finance App learned that SB Energy, a data center developer under the SoftBank Group, where Japanese investment legend Sun Zheng is at the helm, is considering postponing its initial public offering until next year at the latest. The company is building a large-scale artificial intelligence data center for OpenAI in Ohio, USA.
Tech media The Information reported on the 22nd that SB Energy is considering postponing the IPO until December of this year or next year. The media reported in August that the company is aiming to go public as early as September, which means that the listing schedule has been significantly delayed.
This change reflects the rapid cooling of investor interest in data centers and IPOs of AI cloud computing companies. The recent turbulent performance of cloud computing stocks, and the yield on some CoreWeave corporate bonds has soared, prompting other data center developers to reconsider their listing plans. Australian cloud service provider Firmus withdrew its IPO plans earlier this month.
According to the report, SB Energy publicly disclosed its prospectus on the 2nd of last month and submitted a revised version on the 21st of the same month. Companies usually launch a roadshow for IPO investors about two weeks after publicly submitting a prospectus to the US Securities and Exchange Commission. However, when market conditions become volatile, the preparation period from submission of documents to actual release may lengthen.
SB Energy is seeking to finance an Ohio data center through an IPO. The company plans to raise 5 billion to 7 billion US dollars, about 6.61 trillion to 9.25 trillion won, to build this facility to be leased by OpenAI. The data center will use Nvidia chips, is planned to be 8.5 gigawatts, and is backed by a $105 billion credit guarantee from Nvidia.
The investment bank valued SB Energy at about 60 billion US dollars, or about 79.35 trillion won. However, according to The Information, IPO investors have indicated that it is difficult to accept this valuation because the valuation is based entirely on expected future revenue when the company does not have any data centers in operation.
Ohio data centers account for the majority of SB Energy's total contracted data center revenue of $430 billion. The company plans to start operating the facility in phases starting in 2028 and earn revenue through a lease agreement lasting 15 to 20 years after the start of operation. In comparison, SB Energy's revenue for the first half of this year was $139 million, or $278 million annualized. A valuation of 60 billion US dollars corresponds to this revenue, and its market-sales ratio will exceed 200 times. A higher market-sales ratio means a higher valuation of the company compared to current revenue.
The Information quoted a person familiar with the matter as saying that SB Energy executives are closely watching Anthropic's large-scale IPO, which is expected to take place next month. Anthropic is seen as the first AI startup to go public; if its stock performs well after listing, SB Energy is likely to advance the offering as early as December.
The $60 billion valuation is getting cold: the pressure on AI financing is shifting from the bond market to the IPO market
SB Energy's consideration of delaying listing revealed the AI infrastructure's “huge long-term orders and recent lack of cash” financing conflicts.
The SoftBank-owned data center developer originally sought to go public as early as September, but is now considering extending it to December or even 2027. The planned 5 billion to 7 billion US dollars raised will be used to support the Ohio project leased by OpenAI. Investors' resistance to the valuation of about 60 billion US dollars is focused on revenue delivery time: the company has a data center contract of about 430 billion US dollars, but the relevant facilities have not yet been put into operation, and the main projects are scheduled to be launched in stages starting in 2028. The capital market is re-evaluating how high these forward rents can support today's valuations, and how much capital is needed before construction is complete.
This change comes after global assets have just experienced long-term interest rate shocks. The 10-year US Treasury yield had previously risen to 5.34%, a record high since 2002, significantly raising the discount threshold for forward cash flow. However, there have been changes in the latest market: the sharp decline in technology stocks on October 22 triggered safe-haven purchases. US bond yields fell by more than 19 basis points in a single day, and were about 4.97% during the Asian session on the 23rd. Therefore, the current more accurate main line of the market is: although the benchmark yield trajectory has declined, credit risk premiums and equity financing pressure on enterprises associated with AI infrastructure are still prominent.
The stock market has fully reflected this concern. On October 22, the Philadelphia Semiconductor Index fell 6.7%, down more than 26% from its June closing high; the Nasdaq Composite Index fell 2.60%. The SoftBank Group fell about 6.9% in Tokyo on the same day, retreating about half from its June high. Its credit default swap (CDS) price had previously risen to a four-year high. Although SoftBank recovered most of the intraday decline on the 23rd and ended up falling only 0.02%, the short-term price correction has not changed the market's scrutiny of its financing burden.
The credit market provides a more direct indicator of pressure for SB Energy, a data center developer under the SoftBank Group. According to the October 22 report, the yield on CoreWeave's 2031 bond reached 15.76%, and the 5-year CDS spread rose to 1,115 basis points; some media previously reported that its CDS pricing once corresponded to a 5-year cumulative implied default probability of close to 63% — this probability includes credit risk premiums and model assumptions, and cannot be directly understood as the actual probability of the company going bankrupt. SoftBank itself also paid significantly higher financing costs: in September, it announced the issuance of foreign currency bonds equivalent to about US$11.1 billion. Of these, 7.5-year dollar bonds were issued at face value, with coupon interest reaching 9.75%. The funds raised were mainly used to pay for investment in OpenAI and general corporate use. CoreWeave's secondary market yield and SoftBank's new bond dividends together show that investors are demanding higher returns before they are willing to continue funding AI expansion.
Orders lock in the future, and financing decisions can support the delivery process
SB Energy's key constraint is the timing mismatch between construction expenses and rental income. Land, electricity access, substation facilities, computer rooms and cooling systems need to be invested first. After construction is completed and delivery conditions are met, long-term leases can gradually generate revenue. The total contract amount of $430 billion represents the scale of business spanning many years. It is not a cash reserve or the present value of these revenues.
According to a special analysis by Reuters Breakingviews, only about 10% of the relevant data center contract revenue is expected to be realized within six years; this also means that even if demand actually exists, the project must still raise sufficient capital before production is put into operation and bear the financing costs continuously accumulated during the period.
The $60 billion valuation divided by $278 million in annualized revenue corresponds to about 216 times the price-to-sales ratio (P/S); however, current revenue mainly comes from existing energy businesses, and future valuations mainly rely on data centers that are not yet in operation. There is a difference in business stages between the two. What really determines valuation sustainability is how much cash can be left in future rents after deducting construction, operation, and financing costs, and when this cash will arrive. When the benchmark interest rate for the corresponding period and the issuer's credit spread raise the funding threshold, investors will demand a lower entry valuation; the IPO extension will also delay share capital supplementation, further increasing developers' dependence on other financing channels.
The $105 billion officially disclosed by Weida is the upper limit of the conditional guarantee, involving designated lease and electricity payment obligations corresponding to the approximately 4.25 gigawatt IT load in the Ohio campus. It usually takes effect with the commencement of each lease term and is triggered when a specific tenant defaults; it is not cash already paid to SB Energy, nor does it cover the full cost of the project. The 8.5 gigawatt planned scale in the report is not the same caliber as the officially disclosed guaranteed scope; gigawatts measure power capacity and do not directly represent the model's computational performance.
The investment meaning of the adjustment of SB Energy's listing plan is that the AI industry chain is making a stricter distinction between “owning an order” and “having the capital to complete the order.” The more secure the financing arrangement and the closer the project is to electrified delivery, the more certainty that procurement requirements will translate into chip, power equipment, and data center revenue. If Anthropic is successfully listed, it can test the market's acceptance of the growth and valuation of AI companies and improve the subsequent issuance environment; however, in the end, SB Energy still needs to use its own project economy and financing structure to persuade investors to pay the current price for future rents.