Alibaba Group's board member and CEO Eddie Wu image generated by AI. Michael Burry, the investor who inspired "The Big Short," is far from satisfied with Alibaba's new share placement of 80 billion Hong Kong dollars. He stated on Substack that he cannot endorse Alibaba's continued issuance of shares and revealed he has already converted all his Alibaba holdings into JD.com. Burry's concern is that if AI investments keep rising while incremental profits fail to keep pace, the return on invested capital will continue to decline. Issuing new shares at this point means existing shareholders must bear the additional cost of equity dilution.

On August 24, Alibaba's Hong Kong-listed shares fell as much as 10.5%. The market's central question is why the company, which still holds 474.505 billion yuan (approximately 69.9 billion US dollars) in cash and other liquid investments as of June 30, 2026, needs to raise another 10.2 billion US dollars (80 billion Hong Kong dollars) through external financing.

Why Alibaba Chose Equity, and Nomura's Surprise

Nomura Securities said in its latest commentary on August 24 that the truly unexpected aspect for Alibaba was the financing method. For large tech companies with stable cash flows and strong credit profiles, debt financing typically better protects existing shareholders' interests: debt requires interest payments but can be repaid at maturity, while equity financing has no fixed interest cost but permanently dilutes old shareholders' claims on future profits and company value. This time, Alibaba chose the latter, placing 710 million new shares at 112.70 Hong Kong dollars per share to raise approximately 80 billion Hong Kong dollars (about 10.2 billion US dollars), all earmarked for AI infrastructure and full-stack AI capability building. The new shares represent approximately 3.6% of the enlarged share capital, with the placement price at an 8.4% discount to Friday's closing price.

However, in subsequent analysis, Nomura also attempted to explain why Alibaba opted for equity financing. Nomura estimates that the dilution to existing shareholders from this financing is approximately 3.7%, which it believes remains within a manageable range. At the same time, the formal confirmation of the financing scale and method has reduced market uncertainty about how much more capital Alibaba needs and what instruments it will use. Nomura places greater weight on changes in the external financing environment. This year, global tech companies have continued to expand capital demand around AI infrastructure, and the supply-demand dynamics in the bond market have shifted. Nomura points out that credit spreads on tech bonds have widened, new issuances require higher yield premiums, and subscription multiples have declined.

In other words, Alibaba does not lack the ability to issue debt, but the relative attractiveness of debt capital has diminished. This makes Alibaba's choice more realistic: if it continued issuing bonds, the company could avoid immediate shareholder dilution but would bear higher long-term interest costs and consume future borrowing capacity. If it issues equity, the cost is reflected once in dilution, in exchange for permanent capital with no interest payments and no fixed maturity date. The "Big Short" investor Burry opposes the former cost being transferred to old shareholders. Nomura believes that after the deterioration of the debt financing environment, this choice has become understandable. From this perspective, Alibaba's financing is not a choice between "cheap equity" and "cheap debt," but rather a trade-off between two types of capital that are both expensive. Which cost is ultimately lower depends on how high a return AI investments can generate in the coming years. If AI project returns are high enough, selling 3.6% of equity today will look expensive; if returns take longer to materialize, avoiding prematurely pushing the balance sheet toward high leverage also has practical value.

In terms of investor structure, according to foreign media reports, the placement received approximately 28 billion US dollars in orders, of which about 6 billion US dollars came from long-only funds and sovereign investors. Ultimately, about 40% of the issuance was allocated to such long-term capital, including sovereign wealth funds from Europe, Asia, and the Middle East. This at least shows that despite market concerns about AI investment returns and equity dilution, long-term institutions remain willing to take on the risk of Alibaba's future AI investments at current prices. In a large-scale equity raise of 80 billion Hong Kong dollars, Alibaba achieved nearly 3 times subscription, attracting substantial long-only and sovereign capital. For a company that has committed to years of sustained AI infrastructure investment, this provides a large pool of equity capital with no fixed repayment schedule.

For Alibaba's management, the priority is to secure the capital needed for the future before the strategic window closes. Just after the 80 billion Hong Kong dollar placement was completed, management announced share purchases. On August 24, Alibaba Group Chairman Joe Tsai bought 720,000 shares at an average price of about 112 Hong Kong dollars, while CEO Eddie Wu purchased 350,000 shares at around 111.6 Hong Kong dollars. Together, they added 1.07 million shares, investing roughly 120 million Hong Kong dollars. This price is almost identical to the 112.70 Hong Kong dollar placement price. Compared to the 80 billion Hong Kong dollar financing scale, the 120 million Hong Kong dollar impact on the capital structure is minimal, but management's signal is clear.

But one question remains: if debt financing is getting increasingly expensive, why do American tech giants like Amazon and Meta continue to issue bonds on a large scale, even extending maturities to 40 years?

Debt Is Costly, and US Giants Are Already Borrowing for 40 Years

Amazon issued 25 billion US dollars in bonds in July this year, divided into eight tranches with maturities extending from 2029 to 2066. Fixed-rate bond coupons ranged from 4.6% up to 6.25%. Based on the size of each tranche, 24.25 billion US dollars of fixed-rate bonds correspond to approximately 1.31 billion US dollars in annual coupon payments, plus 750 million US dollars in floating-rate bonds. Meta similarly issued 25 billion US dollars in long-term bonds in May this year, with the 2031 tranche carrying a 4.55% coupon and the 2066 tranche at 6.45%, representing about 1.43 billion US dollars in annual fixed coupon payments. As of the end of June, Meta had 84 billion US dollars in outstanding bond principal. These companies choose debt partly because they believe their AI project returns can significantly exceed the cost of capital.

Morgan Stanley's chief fixed income strategist Vishy Tirupattur recently noted that a core assumption behind major hyperscalers continuing to expand AI capital spending is that these projects can ultimately generate returns on invested capital above 25%. As long as this assumption holds, financing at 5% to 6% cost remains an effective form of leverage. But the market is demanding increasingly higher prices. Neil Sutherland, Schroders' head of US fixed income, described the credit spreads on tech debt as beginning to show "indigestion." This year, AI-related corporate bond issuance has surged, with the average credit spread of tech company bonds relative to US Treasuries widening to about 89 basis points. Some new issuances also need to offer new issue concessions to attract investors. Goldman Sachs Research, using a broader methodology, estimates that AI-related credit financing this year, including corporate bonds and project financing, has already approached 500 billion US dollars. Credit strategy head Amanda Lynam believes the importance of this shift cannot be overstated, because capital markets are facing not just a brief wave of bond issuance, but a multi-year financing cycle.

Creative AI Financing: Moving Data Centers Off Balance Sheet

As relying solely on parent company debt becomes increasingly insufficient to meet AI funding needs, tech giants are making their financing structures more complex. Alphabet is scouting different capital pools globally. On August 19, the company issued its first Australian dollar bonds, raising 5.5 billion Australian dollars at once, with maturities spanning 3 to 20 years. The 20-year tranche carried a 6.9% coupon, and orders exceeded 18 billion Australian dollars. Alphabet has previously entered multiple bond markets including sterling, Swiss franc, and yen. Which country has more abundant capital, which currency offers lower costs, and which investors are still willing to take on tech debt have become part of AI capital allocation decisions.

Meta has moved some data centers off its parent company balance sheet. In July, Meta and BlackRock established a joint venture worth approximately 14 billion US dollars for a data center in El Paso, Texas. BlackRock-managed funds hold 80%, while Meta holds 20%. BlackRock plans to invest about 4.9 billion US dollars in cash while raising 12.5 billion US dollars in project-level debt. Meta contributes land and related assets, securing data center capacity through long-term leases. This project finance structure, common in energy, utilities, and real estate, is now entering AI infrastructure. Its significance lies in allowing tech companies' computing needs to grow rapidly without letting parent company balance sheets expand without limit. Once data centers are carved out as standalone assets, infrastructure funds, insurance companies, pension funds, and bond investors can all participate in financing, while tech companies secure usage rights through long-term leases and capacity agreements.

Oracle may be going even further. To chase AI cloud demand, Oracle raised approximately 48 billion US dollars through a combination of debt and equity in fiscal 2026, while facing a free cash flow gap of 23.7 billion US dollars. Under financing pressure, Oracle has begun asking some large customers to prepay for GPU purchases, or letting customers buy GPUs themselves and hand them to Oracle to operate. Oracle has even stated that this structure can significantly reduce the capital it needs to raise. As a result, funding sources for AI infrastructure have rapidly expanded beyond tech companies' own cash flows to include corporate bonds, global local-currency bonds, project financing, infrastructure funds, customer prepayments, and equity markets. Goldman Sachs Asset Management warned in its 2026 investment outlook that during periods of sustained high capital spending, assessing a tech company's risk increasingly depends on whether its core business can continue generating sufficiently strong cash flows. This is also why the market's tolerance for new debt differs between Amazon, Meta, and Oracle.

AI Is Reshaping Tech Companies' Capital Allocation Priorities

Behind Alibaba's placement lies another shift. In recent years, large internet companies including Alibaba, flush with cash, have made buybacks an important capital allocation priority. The logic is that mature businesses generate substantial free cash flow, which companies return to shareholders while reducing share count and boosting earnings per share. Now, by issuing new shares, Alibaba is raising 80 billion Hong Kong dollars to supplement capital for AI infrastructure. This does not mean Alibaba has abandoned buybacks, but it does indicate that growth investment is gaining higher priority during the phase of rapid AI spending expansion.

This shift is not unique to Alibaba. Apollo estimates that hyperscaler cumulative capital spending could exceed 2.7 trillion US dollars from 2025 to 2029. As investment scales grow, external capital will play an increasingly important role in AI infrastructure construction. As AI moves from model competition into a phase of long-term infrastructure investment, tech companies' competitiveness increasingly depends on how much capital they can secure and at what cost. Divergence across capital markets will also become more pronounced.