Key Takeaways:
Semiconductor valuations have reached extremes unseen since the dot-com era, with Kioxia's market cap surging 46x in one year and hyperscaler capital spending now exceeding operating cash flow, Deutsche Bank's annual "WOW Charts" report shows.
The AI chip boom has pushed semiconductor valuations to extremes not seen since 1999, with Kioxia's market cap surging about 46x in one year to become Japan's largest company, Deutsche Bank's 2026 "WOW Charts" report shows.
"This is a structural shift in capital allocation that deserves close attention," Jim Reid, strategist at Deutsche Bank, said in the report published Monday.
In South Korea, Samsung Electronics and SK Hynix drove the KOSPI index to triple from its multi-year lows, pushing the nation's total stock market capitalization past Europe's largest exchanges. The rally reflects an AI capital cycle that has turned memory-chip makers from niche suppliers into trillion-dollar market cap players. Yet beneath the surface, Deutsche Bank flagged a structural imbalance: hyperscaler capital expenditures have surpassed operating cash flow, meaning companies are borrowing or tapping balance sheet reserves to fund AI infrastructure expansion.
The spending gap carries systemic implications. The Bank for International Settlements has classified AI infrastructure investment as a systemic risk, citing circular financing structures where Nvidia invests in OpenAI and OpenAI uses those funds to buy Nvidia chips — effectively making demand appear larger than it is. With US equity valuations near 1999 extremes and global fiscal deficits projected to exceed 2008 crisis levels for the next five years, the report raises the question of whether the AI buildout is a genuine industrial revolution or an over-leveraged bet.
The scale of the AI infrastructure buildout is unprecedented. Capital expenditures by Alphabet, Amazon, Meta, Microsoft and Oracle are projected to exceed $800 billion in 2026 and reach $1.1 trillion in 2027, according to the Kobeissi Letter cited in the report. That figure would represent about 3.2 percent of US gross domestic product — surpassing the nation's defense budget, which accounts for about 2.7 percent of GDP.
In the first quarter of 2026 alone, four of those five companies spent more than $130 billion on capital expenditures, putting them on track for $700 billion to $785 billion for the full year, according to JPMorgan and Moody's estimates. Epoch AI projects that cash capital expenditures will exceed operating cash flow around the third quarter of 2026, shifting the funding model from earnings-based investment to debt-financed expansion.
Morgan Stanley estimates about $2.9 trillion in global data center construction costs through 2028, with more than 80 percent of that spending still ahead. US data center electricity demand is projected to climb from 31 gigawatts in 2025 to 41 GW in 2026, then surge to 66 GW in 2027, according to Goldman Sachs research.
The BIS specifically flagged circular financing as a systemic risk. Nvidia has invested hundreds of billions into OpenAI, which in turn uses those funds to purchase Nvidia's chips. The same capital circulates through the system, appearing as Nvidia's investment, then Nvidia's revenue, and finally OpenAI's compute expansion — inflating demand metrics on paper.
On July 1, Nvidia introduced a new financing model combining revenue sharing with credit support, enabling AI cloud providers to pledge future computing power and pre-emptively access capacity. Harvard's Negotiation Project and multiple research institutions are monitoring such circular deals and issuing warnings.
The report also noted that global private AI investment remains heavily concentrated in the US, with distribution highly uneven. Token economics — the cost constraints of large language model inference — may become a major barrier to enterprise AI adoption at scale.
Deutsche Bank drew direct parallels between current US equity valuations and the 1999 dot-com bubble, noting that while market leadership has broadened beyond the Magnificent Seven, overall valuation pressure remains near historical extremes. The US still dominates global stock market capitalization, though non-US and emerging markets have begun showing signs of recovery after nearly two decades of underperformance.
Unlike the dot-com era, however, today's AI leaders — Microsoft, Google, Amazon, Meta — generate real profits and fund their spending through a combination of operating cash flow and debt. The data centers they build are physical assets capable of running real business operations and generating cloud revenue. Nvidia Chief Executive Officer Jensen Huang has estimated that demand for AI chips alone will reach at least $1 trillion before 2027.
Still, the structural imbalance between spending and cash flow means the bill will eventually come due. In the short term, shareholders absorb the cost through depreciation. In the medium term, creditors bear the risk if AI returns fall short of expectations. In a worst-case scenario, the entire financial system could bear the cost — as it did in 2008.
For investors, the key question is whether the $1.1 trillion in planned AI infrastructure spending will generate returns commensurate with the risk. Nvidia trades at about 35x forward earnings, while Microsoft trades at about 30x. If the AI buildout delivers on its promise, today's valuations may prove justified. If not, the most expensive pile of graphics cards in history will become a cautionary tale for the next cycle.
This article is for informational purposes only and does not constitute investment advice.
