U.S. technology stocks continued their decline this week, with the apparent trigger being a potential delay in OpenAI's IPO. However, analysts suggest that a deeper threat lies in the emergence of low-cost AI models, which not only impact enterprise clients of companies like OpenAI and Anthropic but also challenge optimistic forecasts for AI infrastructure spending.
Reports have surfaced that OpenAI is considering postponing its initial public offering, partly due to the muted post-IPO performance of SpaceX and heightened volatility in the tech sector. This news served as the immediate catalyst for the recent sell-off.
However, according to Jefferies strategist Christopher Wood, citing industry insiders in a client note, the GLM5.2 model from Zhipu "performs almost as well as Anthropic's, but costs only a quarter per token." He characterized the past week as "another DeepSeek moment."
During Friday's trading session, Micron Technology shares fell more than 7%, while AMD and Intel both dropped over 4%. Oracle continued to hit new lows after a 19% decline over the past five trading days.
Low-Cost Models Challenge AI Pricing Structures
Zhipu's GLM5.2 model is seen as a core factor pressuring the tech sector. A Jefferies analyst report on Thursday noted the model not only closely matches the performance of Anthropic's flagship product but also offers privacy protection comparable to leading models.
A Morgan Stanley trader stated on Thursday that Zhipu's new model demonstrates "extremely impressive programming capabilities" and poses direct competition in the enterprise market. The trader noted:
"The trend of enterprises and hyperscale cloud providers shifting to lower-cost models looks more like a recalibration of willingness to pay for AI, rather than a decline in AI demand."
Deutsche Bank analyst Jim Reid also highlighted similar pressures in a June 18 client report, noting that DeepSeek's V4-Pro model performs comparably to Anthropic's flagship Claude Fable 5 on "roughly 90% of daily tasks, but at a cost of only about 1.5%."
Enterprise AI Spending Undergoes Rapid Reshaping
AI pricing pressures are triggering chain reactions among corporate clients. In recent months, several companies have begun scaling back AI expenditures or exceeding their original budgets, with the token pricing of high-end AI models becoming a significant pain point.
The Jefferies analyst further pointed out that if low-cost models also offer reliable privacy protection, enterprises have an incentive to move AI workloads from cloud providers back to their own servers. This shift could fundamentally alter the investment thesis for AI infrastructure.
"Demand structure is clearly tilting toward low-cost models," wrote the Morgan Stanley trader in the report. This trend places valuation pressure on stocks across the NVIDIA supply chain, cloud computing platforms, and data center construction.
IPO Prospects for OpenAI and Anthropic Cloud Over
The combination of rumors about OpenAI's delayed listing and intensifying competition is making the market more cautious about the valuation expectations for these leading AI firms.
Analysts worry that if enterprise clients continue migrating to cheaper models, the revenue growth potential of OpenAI and Anthropic ahead of their IPOs could be eroded. A potential price war between them could further compress their respective valuation prospects before going public.
The continued emergence of cheaper, performance-competitive open-source AI models is seen as a more profound threat to the proposed IPOs of these companies and the broader tech sector. The market's previous assumption of "astronomical growth" in AI infrastructure spending is now facing significant reassessment.
Comments