The most pressing bottleneck for AI currently: cheap funding is drying up

Author: Gao Zhimou, Wall Street Watch

In the past few years, AI companies have easily raised hundreds of billions of dollars, with Wall Street competing to invest with little scrutiny, focusing on "plenty of money and little work." Now, lending institutions and investors are becoming more selective, and the terms are becoming increasingly stringent.

This week, the two most sought-after players in the AI industry hit a wall in the capital markets: Blackstone informed investors that Anthropic is likely to receive only a junk rating after its IPO; OpenAI has cut its new funding target from $30 billion to about $20 billion.

These two events point to the same change: cheap funding is becoming increasingly difficult to obtain for the U.S. AI industry.

Dan Ivascyn, Chief Investment Officer of Pacific Investment Management Company (Pimco), stated:

"The threshold for proving to lenders that these investments can translate into sufficient income is getting higher… the weakest links will be the first to be impacted."

Money is still there, but it's a buyer's market

Since the launch of ChatGPT, AI companies have nearly exhausted all funding sources, including venture capital, the stock market, banks, bonds, and private credit.

According to Bank of America statistics, the issuance scale of AI-related corporate bonds in the U.S. is projected to reach $394 billion by 2025, having already reached $655 billion by mid-October this year, with analysts expecting it to approach $1 trillion next year.

The demand for funding is increasing, but the investors are becoming more hesitant.

Investors' concerns mainly revolve around three points: resistance to data centers is rising in various regions, which may delay large projects; the AI profits of large-scale cloud providers and startups like Anthropic and OpenAI may not materialize as quickly as promised; and the clustering of stock and bond issuances has led to an overexposure to AI for investors.

New trading conditions are reflected in every corner of the market. Bonds from large tech companies like SpaceX and Meta are trading at significant discounts; AI cloud companies like CoreWeave, Nebius, and Nscale, which rely on debt to build computing power, have effectively been shut out of the bond market in recent weeks; even strong companies like Broadcom have had to offer better terms to banks to secure loans.

Anthropic is the latest case. It plans to raise up to $10 billion in its IPO in mid-November and stated in its prospectus that it intends to invest over $800 billion in computing infrastructure by 2045. According to the Financial Times, Blackstone recently informed investors that, based on guidance from major rating agencies, Anthropic is more likely to receive a BB rating or higher after its IPO, whereas Blackstone had previously expected it to obtain an investment-grade rating by the end of the year.

According to S&P and Fitch's systems, BBB- and above are considered investment grade, while BB+ and below are considered speculative, commonly referred to as "junk grade." Many insurance companies and pension funds are restricted from holding large amounts of junk-rated debt, meaning that a rating in this range would narrow Anthropic's borrowing channels and increase costs.

The $63 billion chip leasing deal reached between Broadcom and Anthropic this week was also negotiated under tense circumstances. According to the Wall Street Journal, Broadcom CEO Hock Tan personally intervened in the negotiations at the last minute to stabilize the banks' emotions; the deal's over $42 billion in senior loans was secured by Broadcom's high A rating. Anthropic also listed financing constraints from power suppliers as a risk factor in its prospectus: even if it can raise funds, it must worry about whether suppliers can.

OpenAI's situation is similar. According to the Financial Times, it has reduced its funding target to about $20 billion, with its valuation adjusted from $1.4 trillion to $1.25 trillion; the Wall Street Journal cited insiders stating that OpenAI has raised less than half of its initial $30 billion target. Some investors have lowered their subscription amounts, and others want to wait until Anthropic goes public before making decisions.

Although top-rated companies with good credit can still obtain funds, market patience is wearing thin.

This week, as tech companies reported disappointing earnings, AI-related stocks and bonds were sold off: Alphabet saw its market value evaporate by about $500 billion on Thursday, the largest drop in history; Intel plummeted 29% on Friday; the Nasdaq Composite Index fell 2.9% on Friday, marking one of the largest sell-offs since the AI boom began.

If volatility continues, financing will only become more difficult.

New cloud providers: the first to be shut out

The pressure is first falling on the weakest credit link.

One of the most active issuers in the AI debt market is companies referred to as "new clouds." They own or lease data centers, deploy AI servers, and sell computing power to large tech companies, trading firms, and AI startups.

Many new cloud providers have transitioned from cryptocurrency mining companies, have a short history of operating AI infrastructure, and had already faced high borrowing costs before the market turmoil, making them most sensitive to rising interest rates.

The turning point occurred in September. The Federal Reserve raised interest rates for the first time in three years; Oracle notified developers of a large data center in New Mexico that it would delay the start of full rent payments due to power supply and approval issues, prompting lenders to reassess the construction risks of their projects.

The bonds issued by CoreWeave earlier this year had a coupon rate of 9.75%, but now the yield in the secondary market has exceeded 15%. The drop in bond prices has led to rising yields, indicating that its borrowing costs have surged in the short term. Bankers have recently warned that the high-yield debt market for new cloud providers is essentially closed to new issuances, as yields above 15% are unbearable for borrowers.

Since 2024, Nebius has sold about $2.7 billion in stock to investors like Nvidia and issued billions in convertible bonds, but its funding needs remain unmet. It initially considered issuing its first high-yield bonds but ultimately abandoned that plan and opted to charge customers upfront.

CEO Arkady Volozh stated last week in Berlin that some funding sources, such as asset-backed securities and bond markets, "may be temporarily closed," but tools like customer prepayments "should be enough for us to at least build this 1 gigawatt of capacity." COO Ophir Nave noted that the asset-backed financing market is still open.

Equity financing is also not going smoothly. Insiders reported that Lambda sought to raise $4 billion in its final round of financing before its IPO, but after investors like Coatue decided not to participate, it lowered its target valuation from $14.5 billion to $13 billion.

The British new cloud provider Nscale originally considered going public in November but is now more inclined to postpone until December. It is also negotiating with investors like Nvidia for funding support for Anthropic's $45 billion data center in West Virginia, but the stricter terms requested by the other party after the market shift have slowed its progress in updating the prospectus. Some bankers have informed Nscale that its IPO is unlikely to reach the previously set valuation of $30 billion to $35 billion.

Giants: No matter how high the rating, investors will still "fill up"

Meta, Amazon, Alphabet, and Microsoft are on the other end of the credit risk spectrum. They have top ratings and significant borrowing capacity on their balance sheets, but the issue lies in the scale of financing itself.

This year, they have been actively issuing bonds in the high-grade debt market and have indicated that they will continue to do so in the future. The investment firms, insurance companies, and pension funds providing loans have felt the limits tightening and are also concerned about over-concentration in the same industry. "The real question is, how much can investors fill up on one name, one theme?" Ivascyn said.

On Friday, the bond prices of several large-scale cloud providers fell to near historical lows, with yields significantly higher than those of typical investment-grade corporate bonds. Despite their top ratings, their borrowing costs have exceeded those of ordinary investment-grade companies.

Everyone wants to get ahead of others. Meta plans to raise up to $97.5 billion through a series of new stock issuances, one reason being to act before competitors and to frame the financing of other giants in regulatory documents as a challenge to the capital market's capacity.

"Everyone wants to raise money for their construction before liquidity dries up," said Ben Connard, Chief Investment Officer of Carnegie Investment Counsel, "Ultimately, there may not be enough money to support this construction. Financing could become the final bottleneck that slows everything down."

Why the tightening of funds has become so critical

AI construction is increasingly reliant on external funding. According to an analysis of consensus forecast data from LSEG by Reuters, the capital expenditures of Microsoft, Alphabet, Amazon, Meta, and Oracle are expected to exceed their free cash flow by 2027, with every additional $1 of operating cash flow corresponding to about $1.57 of new capital expenditures. The money earned is not enough, and the gap can only be filled by issuing stocks and bonds.

The giants cannot stop either. A research report from Cornell Capital Group describes this as a "prisoner's dilemma": if one does not invest while competitors do, they may be eliminated.

Investor Citrini pointed out in an interview with the Monetary Matters podcast that this change alters the nature of risk. "In the first three years of this AI deal, the emotionally driven downturn had a bottom because most of the spending came from the operating cash flow of large-scale cloud providers. Now, incremental spending is almost entirely from equity and debt financing; if market sentiment turns, that incremental funding will dry up because it is no longer just one company spending money, but determined by the market."

He referred to this mechanism as "reflexivity": confidence supports financing, and financing supports investment; once confidence weakens, the contraction of funds will in turn impact investment and growth expectations. He also noted that current AI revenues are growing at about 25% to 30% annually, while capital expenditure growth rates are close to 60% to 70%, with the gap widening with each round of earnings guidance.