�۶���Ƶ

Podcast: CIO analysts on "circular" AI deals, equity froth, and humanoid robots (13 min)
Video: Paul Donovan on what next for the Fed (8 min)
Podcast: (18 min)

While the number and size of deals in the artificial intelligence sector over the last few months has encouraged some investors that the value chain is evolving at pace, the details of some agreements have raised alarm bells in some quarters. In particular, some investors are questioning whether OpenAI’s strategic partnerships stand at the middle of an increasingly complex web of vendor financing arrangements, with echoes of market activity in the internet boom and bust of the early 2000s.

Recent OpenAI announcements include a USD 300bn cloud infrastructure agreement with Oracle, a USD 10bn custom-chip partnership with Broadcom, and strategic alliances with major semiconductor and memory providers. Notably, OpenAI’s partnership with NVIDIA involves up to USD 100bn to deploy 10GW (equivalent to around USD 500bn AI capex investment) of AI systems, while a deal with AMD could result in OpenAI acquiring a significant stake in the company. Additionally, OpenAI has secured agreements to purchase large amounts of memory capacity from leading suppliers—the size of which accounts for half of the world’s current capacity.

These interconnected deals have raised concerns about circular financing—where companies invest in each other and rely on mutual spending—drawing comparisons to the telecom bubble at the turn of the millennium. Investors are questioning whether such dependencies could pose risks if AI demand or monetization falls short of high investor expectations.

But while we are vigilant to the risks these types of commercial arrangements pose to both individual company results and to the wider market, we do not believe the current wave of AI vendor financing is a repeat of past tech bubbles.

The scale of circular deals, while significant, is not overwhelming. For example, the OpenAI-NVIDIA arrangement is estimated to represent up to 13% of NVIDIA’s projected 2026 revenue (based on market consensus revenue projections of USD 272bn). If a 1GW deployment occurs in the second half of 2026 (as suggested by a 22 September Reuters article), it would trigger a total capital investment of around USD 50-60 billion, with NVIDIA receiving USD 35 billion of that amount (based on NVIDIA accounting for approximately 60-70% of total AI capex). From this, USD 10 billion may be reinvested into OpenAI (according to NVIDIA), with further investments contingent on actual progress in AI monetization. This performance-based approach contrasts with the fixed, often speculative commitments seen during the telecom bubble.

The financial health of today’s leading AI firms is much stronger than telecoms at the turn of the millennium. Most AI capital expenditures, especially among the largest technology companies, are funded by robust operating cash flows. The “Big 4” tech firms are collectively expected to generate USD 203bn in free cash flow (after capex) in 2025, indicating considerable investment capacity, limited balance sheet risk, and a far more sustainable funding model than in previous cycles. This means that the value chain as a whole is not overly relying on debt or external financing to fund growth (although there are pockets of the AI value chain where financial vulnerabilities may be more pronounced).

Valuations are more reasonable and earnings quality is higher. In the late 1990s, internet leaders traded at forward price-to-earnings ratios around 60x, while today’s AI giants are closer to 35x, with stronger balance sheets and more predictable earnings streams. This suggests that the market is not pricing in unrealistic growth expectations, and that companies are better positioned to weather potential slowdowns. We continue to closely monitor key metrics such as AI monetization rates, funding trends, and the economics of GPU investments.

So although medium-term risks remain—especially if AI adoption or revenue growth disappoints—the near-term fundamentals should remain resilient, in our view. The current deal flow supports supply chain guidance and could even drive consensus earnings upgrades. Overall, we see the present environment as fundamentally different from prior bubbles, with AI companies largely adopting more prudent investment strategies and healthier corporate finances.

We retain conviction in our investment idea to seek beneficiaries of transformational innovation, with artificial intelligence at the forefront. But AI is not monolithic and nor should be investors’ positioning in AI stocks. We believe AI will be a multi-layered, multi-year opportunity that spans enabling technologies, intelligence platforms, and application solutions. Rather than concentrating risk in a single segment, we favor balanced exposure across the entire AI value chain. Semiconductors, for example, comprise about 23% of our recommended AI exposure, reflecting their critical role in powering next-generation systems. Our wider focus is on companies with strong fundamentals, sustainable cash flows, and the ability to innovate at scale. For investors, we believe a diversified, nimble approach that can pivot across layers of the AI value chain and invest across regions, sectors, and industries will likely be best placed to capture potential gains from transformational innovation while managing the risks inherent in this fast-evolving space.

For more details, please see the latest edition of CIO's Intelligence weekly publication.