The Zhitong Finance App learned that SoftBank Group's mobile business had previously begun to set up “neocloud” (neocloud) operations in the US, and is now evaluating the impact of two events: the Democratic Party swept the US midterm elections last week, and the growing anti-data center sentiment.
“Major developments in the US will not be particularly beneficial to us.” SoftBank Wireless Business CEO Junichi Miyakawa told investors at a performance briefing in Tokyo on Monday. He added that overall global demand for AI data centers is still strong, but if the US situation worsens too much, “we may launch in other countries first.”
Performance is improving, but plans are being re-evaluated
As one of the largest cash cows within Sun Zhengyi Investment Group and the third largest wireless operator in Japan, SoftBank's financial results for the day were not weak: in the first half of the fiscal year ending September 30, revenue increased 11.7% to 3.8 trillion yen, and operating profit increased 2.1% to 642.1 billion yen, all record highs in the same period in history; cloud computing and AI-related revenue increased 37% year over year, driving the company to raise the annual revenue guide from 7.5 trillion yen to 7.9 trillion yen, and operating profit from 1.1 trillion yen to 1.1 trillion yen 1.16 trillion yen, net profit raised from 560 billion yen to 610 billion yen — the 600 billion yen net profit milestone originally achieved in fiscal year 2028 is expected to be reached two years ahead of schedule. By sector, financial business revenue increased 27.5% year on year, and distribution business grew 19.1%, which together with cloud computing formed the next growth engine.
During Miyakawa's tenure, the operator continued to expand into data centers and AI services to meet the surging demand for computing power from Japanese companies.
Where did SoftBank America play this game of chess
SoftBank has no small appetite in the US. The company announced on Thursday that it will establish a joint venture with the group at 51:49, and plans to rent AI computing power to US companies — including hyperscale cloud vendors — starting next fiscal year, with the goal of increasing the data center supply capacity to 10 gigawatts around 2030; at the same time, the planned Ohio data center project will reach 10 gigawatts and an investment of about 500 billion US dollars. OpenAI may become an early customer — SoftBank Group's cumulative investment in OpenAI is about 65 billion US dollars.
According to people familiar with the matter, if the US business expands smoothly, SoftBank's annual operating profit is expected to rise to 3 trillion to 4 trillion yen, which is three to four times the current level; Miyagawa characterized this as the company's “second venture.” Miyakawa's statement on Monday is tantamount to adding political variables footnotes to this blueprint for the first time.
The electricity puzzle is also in place: SoftBank is building a data center at the former Sharp LCD factory site in Sakai City, Osaka, and is also transforming a large-scale battery mass production line at the same site, with a target annual production capacity of about 1 gigawatt-hour — lithium iron phosphate first, zinc-halide batteries to be put into operation in 2027, and the annual revenue target for the battery business in FY2030 exceeds 100 billion yen — preparing for this resource-scarce country to keep up with the electricity demand for AI computing power.
Industry Headwinds: $3.5 Trillion Impairment Alert
Investors are increasingly questioning the economics of cloud services tied to AI and the sustainability of massive data center spending. An analysis released by research firm Gartner on October 14 shows that if AI service providers maintain the current pace of capital expenditure, AI infrastructure investments of more than $3.5 trillion may face the risk of impairment by the end of 2030: AI data center owners will need to generate a cumulative revenue of 8.5 trillion US dollars by then to cover capital expenses, financing costs, and achieve a 12% return on invested capital, while the highest AI revenue attributable to AI data centers is only 3.6 trillion US dollars ; By 2030, cost pressure and insufficient differentiation will force more than 80% of new cloud service providers to stop operating independently. Will Sommer, senior research director at Gartner, pointed out that hyperscale cloud vendors have multiple revenues and can support investment for many years, and new cloud service providers must earn returns on AI computing power leases as soon as possible, and the situation is most exposed.
This shadow just swept through the market last week: from November 4 to 5, the US stock storage and computing power chain plummeted for two consecutive days. Seagate's cumulative decline over the two days was about 23%. The trigger for the sell-off was the market's concerns about the tightening of data center approvals after the midterm elections. SoftBank's plan is the first heavyweight player to give a “backup route” in this shift in sentiment — and Miyakawa also left a balanced statement: Global demand for AI data centers is still strong, and the problem is only in the US.