Alibaba Priced HK$80B of AI Shares; SoftBank Readied ¥1T Bonds

Asia’s AI build-out reached two different pools of capital during the 24 hours ending 24 August 2026 at 09:05 in Tehran. Alibaba priced HK$80 billion of new shares and said every dollar of net proceeds was intended for full-stack AI. SoftBank Group filed to issue ¥1 trillion of seven-year bonds, aimed mainly at individual investors in Japan.

The amounts are large, but the financing structures matter more than a converted headline total. Alibaba is asking shareholders to accept dilution without a fixed repayment date. SoftBank is preserving equity ownership while proposing a contractual coupon and principal claim. One filing explicitly earmarks the proceeds for AI; the other specifies the bond terms but does not state how the money will be used.

The 24-hour brief

Samsung’s 21 August shareholder-return plan fell outside the locked window and is not treated as a development for this brief. No rumour or unnamed-source financing story is used as primary evidence.

Alibaba sold 3.57% of the enlarged company

Alibaba’s pricing filing dated 24 August says it agreed to place 710 million newly issued ordinary shares with at least six non-US professional, institutional or other investors at HK$112.70 each. The expected HK$80.0 billion gross proceeds become about HK$79.7 billion after commission and estimated expenses. Closing is expected on 26 August, subject to listing approval and other customary conditions.

The new shares represent 3.70% of the 19.175 billion shares outstanding before the deal and 3.57% of the 19.885 billion enlarged total. They are being issued under a general mandate approved at the September 2025 annual meeting, so the placement does not require another shareholder vote. The filing says the price was a 3.6% discount to its defined referenced share price and a 9.0% discount to the five-day average New York closing price after adjusting for the eight-shares-per-ADS ratio.

Most importantly, Alibaba states that 100% of net proceeds will go to full-stack AI capabilities, including expanding and improving AI infrastructure. It does not divide that amount among chips, data centres, models, inference, software or customer acquisition. The use is explicit; the allocation and delivery schedule are not.

Recent results explain the financing urgency

The placement followed a quarter in which AI demand and cash consumption both accelerated. Alibaba’s 20 August June-quarter filing reported RMB48.437 billion of AI Cloud and Compute Services revenue, up 45% year over year. Segment adjusted EBITA more than doubled to RMB5.628 billion. AI-related product revenue reached RMB12.376 billion and recorded a twelfth consecutive quarter of triple-digit growth.

That is meaningful operating evidence, not merely a capacity target. It is also incomplete. AI Labs and Applications produced only RMB3.338 billion of revenue and an adjusted EBITA loss of RMB13.861 billion, compared with a RMB3.224 billion loss a year earlier. The cloud infrastructure layer is showing operating leverage while the model-and-application layer remains expensive.

Capital expenditure was RMB67.678 billion for the quarter. Free cash flow, Alibaba’s non-GAAP liquidity measure, was an outflow of RMB44.670 billion, mainly because of higher cloud infrastructure spending. Cash and other liquid investments still totalled RMB474.505 billion at 30 June, so the placement should not be read as proof of a near-term liquidity shortage. It is a capital-allocation choice to raise fresh equity while a much larger internal liquidity pool is being consumed by several businesses and investments.

Net income fell 75% to RMB10.444 billion, but it would be misleading to assign that entire decline to AI capex. Alibaba attributed the change to lower income from operations, lower gains on investment disposals and lower mark-to-market gains. Non-GAAP net income fell 38%, with technology investment partly offset by improved cloud results and operating efficiency.

Our 21 August review of Alibaba’s cloud quarter separated earned revenue, incurred capex and free cash flow. The new event adds a fourth category: financing raised. None should be substituted for another.

The share price supplied an immediate limit

Alibaba’s Hong Kong shares fell as much as 10% to HK$110.10 in early trading, below the HK$112.70 placement price, according to Reuters reporting published on 24 August. That move is consistent with investors pricing the additional share count and questioning the timing of returns from heavy spending.

It does not establish the eventual return on the new capital. The operational test is whether the HK$79.7 billion of expected net proceeds creates capacity that is energised, filled by paying workloads and converted into durable gross profit before technology, chip availability or customer demand changes. A placement can fund that work; it cannot validate it.

SoftBank offered households a different claim

SoftBank’s official release timestamped 24 August and its two-page bond filing propose the 70th unsecured straight corporate bond. The total amount is ¥1 trillion, each bond is ¥1 million, the term is seven years and maturity is 16 September 2033. The preliminary annual coupon range is 4.30–4.90%, with final terms expected on 4 September. The Japanese offering is scheduled for 7–16 September and issuance for 17 September.

The bonds have no collateral or guarantee. They include a negative pledge, a mechanism to move from unsecured to secured status, and a net-worth maintenance clause. Japan Credit Rating Agency is expected to assign an A rating. Eleven securities firms are listed as underwriters, and the offerees are mainly individual investors.

If the full amount is issued at the preliminary range, the simple annual coupon cost would be ¥43–49 billion before issuance expenses. That is an arithmetic implication, not a final cost: the coupon has not been fixed and the sale has not completed.

The filing itself does not identify a use of proceeds. Reuters placed the deal in the context of SoftBank’s expanding AI and OpenAI commitments, but that context should not be rewritten as a contractual earmark.

SoftBank already has a large AI financing stack

The wider financing need is documented. SoftBank’s 27 February OpenAI investment notice committed another $30 billion in three $10 billion tranches, with the final tranche scheduled for 1 October. The company said completion would take its cumulative OpenAI investment to $64.6 billion and an ownership interest of about 13%, subject to closing conditions.

SoftBank arranged a $40 billion bridge facility maturing in March 2027. In its CFO message published with the 2026 group report, it said $20 billion had been drawn for the follow-on investment, another $10 billion was planned with the October payment, and takeout financing would combine asset-backed borrowing and possible asset sales.

That makes the new retail bond relevant to the AI capital story, even though the bond filing does not connect the two. It may add durable funding capacity or contribute to a broader refinancing plan; asserting a direct bridge-loan repayment would go beyond the disclosed evidence. This distinction echoes our analysis of NVIDIA’s conditional $105 billion Ohio backstop: a cap, commitment, guarantee, borrowing facility and cash payment are different financial objects.

Equity and debt move the risk differently

Equity can absorb an uncertain payoff without a refinancing cliff, but it permanently divides future earnings among more shares. Debt avoids that immediate dilution, but raises the amount that must be serviced before equity receives value. Neither structure proves that AI workloads will earn an adequate return.

What operators should take from the day

• Track financing proceeds separately from contracted equipment, construction progress, energised capacity and billable usage.
• Reconcile AI-cloud revenue with segment profit, model-layer losses, capex and free cash flow rather than selecting the fastest-growing number.
• Record whether proceeds are legally earmarked, described only as an intention or not allocated in the filing at all.
• Translate debt ranges into cash-interest scenarios, but wait for final pricing before treating them as obligations.
• Measure dilution against the post-deal share count and distinguish it from a one-day share-price move.
• Tie infrastructure spending to power, chip delivery, commissioning, utilisation and customer concentration milestones.
• Test whether a short bridge is genuinely replaced by longer-term funding instead of simply layered beneath it.
• Keep market reaction as a financing signal, not a substitute for operating evidence.

What remains uncertain

Alibaba has not published a project-level budget for the HK$79.7 billion of expected net proceeds. Its 100% AI designation spans a wide stack whose infrastructure, cloud and application layers have different margins and delivery risks. Completion also depends on the placement conditions being satisfied.

SoftBank has not fixed the coupon, completed the retail sale or disclosed the bond’s use of proceeds. Its OpenAI investment, bridge funding, data-centre plans and bond issue are documented separately. Connecting them is reasonable strategic context, but the cash path must be verified through later financing and balance-sheet disclosures.

The companies also have different business models, currencies, accounting structures and asset portfolios. The day is useful because the financing mechanisms contrast—not because their headline amounts form a clean league table.

What to watch next

• Alibaba’s 26 August placement closing and final net proceeds.
• A project-level split of Alibaba’s AI infrastructure, chip, model and application spending.
• AI Cloud and Compute Services external revenue, adjusted EBITA and capacity utilisation.
• AI Labs and Applications inference cost and adjusted EBITA loss.
• Alibaba free cash flow and liquid-investment balance after the new capital arrives.
• SoftBank’s final coupon, order demand and confirmed rating on 4 September.
• Completion of the ¥1 trillion issue on 17 September and any later use-of-proceeds disclosure.
• SoftBank’s planned 1 October OpenAI tranche and progress replacing the March 2027 bridge facility.
• Changes in SoftBank’s loan-to-value ratio, cash runway and asset-backed financing.

Alibaba has put an explicit AI label on new permanent capital. SoftBank has put a preliminary price range on a large retail credit claim. The next evidence is not another ambition figure; it is whether each financing route becomes deployed, revenue-producing capacity without eroding the financial flexibility it was meant to create.