# Nasdaq’s AI winner-takes-all bet faces pressure from falling prices and rising costs

By Riza Dagoc

Canonical URL: https://www.tokenpost.com/news/investing/42388
Published: 2026-10-28T06:01:36.000Z
Updated: 2026-10-28T06:01:36.000Z
Section: Investing

> David Wu argues that Nasdaq’s AI valuations assume concentrated profits, while falling token prices and rising computing costs could undermine growth and margins.

Expectations that a handful of AI companies will capture the industry’s profits face a challenge from price competition and rising computing costs. That combination could make the high valuations embedded in US technology stocks harder to justify.

Speaking on David Lin’s broadcast on October 27, David Wu, CEO and founder of David Wu Unbound, said Nasdaq valuations assume a winner-takes-all outcome for AI. He argued that commoditization was the more likely result.

Wu pointed to differences in how the US and China approach AI competition. He described the US strategy as seeking control over critical parts of the AI supply chain so a few companies can capture future profits. China, he said, encourages competition among numerous AI model developers and chipmakers to drive innovation. He argued that competition and falling prices in China could eventually shape the US AI trade.

He also questioned the growth assumptions needed to support Anthropic’s valuation. Comparing Anthropic and TSMC at an assumed valuation of $2 trillion each, Wu said his analysis suggested Anthropic would need annual revenue of about $1.2 trillion by 2030 to offer comparable investment appeal to the highly profitable chipmaker.

Even using a generous hypothetical annual revenue figure of $100 billion for Anthropic in 2026, Wu said reaching that level would require roughly twelvefold growth in four years. He compared the target with what he described as $1.22 trillion in worldwide software spending in 2025, questioning expectations that a single company could generate nearly that much revenue.

Wu said revenue growth was already showing signs of slowing. Citing figures that had emerged, he said Anthropic’s quarter-on-quarter growth had fallen from about 300% in the second quarter to around 45% in the third. He also cited spending management company Ramp’s data as showing zero growth in token spending across AI models in both August and September. Falling token prices could constrain revenue growth even as AI demand increases, he argued.

Computing costs, meanwhile, were facing upward pressure, Wu said. He argued that rising GPU rental rates, including for roughly three-year-old H100 chips, could not simply be read as evidence of strong demand. Limited power-grid capacity and permitting delays were slowing data-center construction, potentially leaving purchased chips unable to enter service.

Wu suggested some newer chips could remain in storage while depreciation begins before they are connected to power. He said Anthropic and OpenAI could face substantial pressure on margins as falling service prices slow revenue growth while computing costs rise.

In discussing Anthropic’s IPO, Wu questioned whether a $2 trillion valuation was achievable, suggesting investor expectations had shifted closer to $1.4 trillion or lower. Asked which financial measures mattered most in coming quarters, he identified revenue as the priority, with sustained growth and eventual profit margins central to assessing the company’s valuation.