“Is there an AI bubble?” is such a tired thought. Here’s something altogether more wired: The AI boom is paying off, but not in a way that the current equities market has accounted for. The success of the technology in one area of the economy could make the bubble real in another, more precisely.
In a blog post published on Friday, Apollo Chief Economist Torsten Slok highlighted that the parts of the AI value chain with the highest profit margins—companies making AI models and applications—actually have the lowest levels of profitability, a departure from the standard business model of, well business, in which profit margins are higher for companies selling an end product to consumers.
Slok broke down AI companies into four categories: models and applications, cloud and compute, energy and grid, and silicon and equipment. Using data from Pitchbook and Bloomberg for companies including OpenAI, Anthropic, Microsoft, Amazon, Constellation Energy, Nvidia, AMD, and Micron, Slok calculated that silicon and equipment—such as chipmakers—has the highest profit margin, 41%, in the AI value chain. Meanwhile, models and applications—like Anthropic—have a -59% operating margin.
Slok warns that this sharp disparity is because money from the AI boom is not coming from natural demand for AI applications, but rather shareholders hoping to cash in on what they hope is the next technological revolution.
“AI boom’s profits are currently being funded by investors rather than earned from customers,” Slok said. “The upstream margins are real, but they are paid for out of capital raised by the layer losing money, not out of cash generated by end demand.”
Goldman Sachs now projects AI investments to swell beyond $1 trillion in 2026, but so far, the technology has little to show for itself, with no significant changes in economic productivity or profit margin growth outside of the Magnificent Seven. Should AI financing slow down, the lopsided profit margin structure threatens to topple the stability of the entire rapidly expanding industry, Slok warns.
“The bottom line is that the most profitable part of the AI value chain depends on the least profitable part continuing to grow revenue or raise capital,” he concluded. “Capital can bridge the gap for a while, but not indefinitely. And therein lies the risk: will the ROI show up for AI’s end customers fast enough to sustain the spending that is generating those upstream margins?”
Wider spread fears of an unsustainable AI expansion
Slok isn’t the first economist to sound the alarm on AI’s outsized reliance on investments. In its annual report published in June, the Bank of International Settlements noted the onslaught of AI investing, primarily from the five major hyperscalers, is outpacing earnings and free cash flow, which has led to these companies issuing debt to raise additional financing. A Bank of America analysis from last November found that in 2025, those five hyperscalers issued $121 billion in debt, four times the average debt levels issued by these firms annually over the previous five years.
“Disappointment in returns could trigger a sudden pullback in financing and turn the capex boom into a protracted investment bust, with potential knock-on effects on financial conditions,” Bank of International Settlements said in the report. “Should hyperscalers slow or halt the aggressive pace of capex deployment, many borrowers across the supply chain could struggle to replace lost revenue and service their debt.”
Tech writer Ed Zitron took this concern a step further, arguing AI spending is more precarious than it even appears on the surface. He used the example of Oracle, which has a negative cash flow of $23.7 billion, as of the end of fiscal 2026, and nearly $130 billion in outstanding debt and $260 billion in lease commitments for AI infrastructure projects that have yet to begin. Its massive gamble on AI buildout is in service of OpenAI, with whom it signed a $300 billion deal last September.
“Oracle’s existence — and Larry Ellison’s personal wealth — hinges on whether OpenAI can make good on its promise to spend $300bn in compute,” Zitron wrote in a Substack post in June.
He called this dynamic “most-obvious and under-discussed part of the AI bubble”: While 13-figure hyperscaler capital expenditures are fueling a semiconductor boom, there’s little evidence so far that the current AI boom will translate to widespread applications of technology that will justify all this spending.
More dangerous than just Oracle failing to deliver on its $300 billion promise to OpenAI is other tech companies giving up on their own exorbitant AI spending, Zitron explained.
“If Microsoft, Google, Amazon and Meta decide that it’s time to stop spending $30 billion or more a quarter on GPUs, RAM, storage, and data center construction,” he said, “that’ll tear a hole in the side of what people assume is a permanent supercycle.”
The question « Is there an AI bubble? » has become a somewhat cliché topic in economic discussions, yet the underlying dynamics of the AI market reveal a more complex reality. Recent insights from Apollo Chief Economist Torsten Slok indicate that while the AI sector is experiencing substantial growth, the profits generated are not aligning with traditional market expectations. Specifically, Slok highlights a significant disparity in profitability among different segments of the AI value chain, suggesting that the current boom may be artificially sustained by investor speculation rather than genuine market demand.
In his analysis, Slok categorizes AI companies into four distinct segments: models and applications, cloud and compute, energy and grid, and silicon and equipment. He utilizes data from Pitchbook and Bloomberg to illustrate that while silicon and equipment companies—such as chipmakers—enjoy profit margins as high as 41%, the models and applications segment, which includes companies like Anthropic, reports a staggering -59% operating margin. This indicates that the firms developing AI technology are currently losing money, which is atypical for a healthy business model where higher profit margins are generally found in consumer-facing products.
Slok warns that the current investment in AI is largely fueled by capital raised from investors rather than generated through actual customer demand. He states, « AI boom’s profits are currently being funded by investors rather than earned from customers, » emphasizing that the sustainable profitability of the AI sector hinges on the ability of its less profitable segments to continue raising capital. The consequence of this reliance on investor funding is a precarious situation where the entire industry’s stability could be threatened if investment slows down.
Goldman Sachs projects that AI investments could exceed $1 trillion by 2026, yet, as Slok points out, there has been little evidence of corresponding improvements in economic productivity or profitability outside of a select group of dominant tech firms—referred to as the « Magnificent Seven. » Should the pace of AI financing decelerate, this could lead to significant challenges for companies reliant on these investments, as the profitability of the more successful segments of the AI value chain is dependent on the growth and revenue of the less profitable segments.
The Bank of International Settlements has also raised concerns about the sustainability of AI investments. Their annual report highlights that the aggressive investment in AI, particularly from major tech players, is outpacing actual earnings and cash flow. This has resulted in increased debt issuance among these companies. For instance, a Bank of America analysis indicated that in 2025, major hyperscalers issued $121 billion in debt—four times the average annual debt levels over the previous five years. This trend raises the specter of a potential investment bust if returns on these AI investments do not meet expectations.
Tech writer Ed Zitron elaborates on these concerns, illustrating how precarious the AI spending situation is. He uses Oracle as a case study, revealing that the company had a negative cash flow of $23.7 billion and nearly $130 billion in outstanding debt, alongside substantial lease commitments for AI infrastructure projects. Oracle’s future is closely tied to its relationship with OpenAI, particularly a $300 billion deal that hinges on OpenAI’s ability to deliver on its promises. This situation exemplifies the risky dynamics at play: while major capital expenditures are driving a semiconductor boom, there is little evidence that the current AI investment surge will yield the widespread applications necessary to justify such spending.
The implications of these findings are profound. If major tech companies like Microsoft, Google, Amazon, and Meta decide to scale back their AI expenditures—which currently exceed $30 billion quarterly on hardware and infrastructure—the resulting contraction could disrupt the expectations of a sustained AI supercycle. Zitron argues that this potential outcome poses a significant risk to the perceived permanence of the AI market’s growth trajectory.
In summary, the current AI boom is characterized by a complex interplay of investment, profitability, and demand. While some segments of the AI value chain are thriving, the overall structure remains vulnerable, as the profitability of the more successful areas depends on the continued growth and capital influx to the less profitable segments. Without a clear pathway to sustainable revenue generation from AI applications, the industry’s future appears uncertain, raising questions about the long-term viability of the current investment strategies driving the AI boom. As stakeholders in the AI market navigate these challenges, the specter of an « AI bubble » looms large, calling for a careful reassessment of the underlying dynamics at play.
