Jefferies Warns Investor Returns May Derail AI Rally

The Real Threat to AI Investment

The greatest risk to the current AI-fueled technology rally is not semiconductor supply constraints but a potential investor awakening that hyperscalers and companies like OpenAI and Anthropic may fail to deliver returns on their massive capital deployments, according to Jefferies.

The investment bank argues that while AI capital expenditure shows no signs of slowing, funding anxiety could trigger a sudden withdrawal of investment. This risk is amplified by circular financing arrangements among major players, such as Nvidia providing capital to OpenAI for chip purchases.

Jefferies points to zero indication of AI spending deceleration for now, yet cautions that interconnected funding structures create systemic vulnerability. The brokerage's analysis suggests these circular arrangements could accelerate a market downturn if investor confidence cracks.

Chinese Models Capture Enterprise Share

Hong Kong-listed Z.ai, formerly Zhipu AI, launched its GLM-5.2 model on June 13, marking a significant shift in competitive dynamics. The model approaches Anthropic's performance for enterprise applications while charging one-quarter the cost per token, according to Jefferies.

This pricing pressure coincides with what the brokerage describes as a market reaction against "tokenmaxxing," likely to slow Anthropic's revenue trajectory ahead of its planned initial public offering. Anthropic's annualized revenue surged from $9 billion at the end of 2025 to $47 billion in May, according to Jefferies data.

Usage patterns on global aggregator platform OpenRouter demonstrate the momentum behind Chinese AI providers. The top Chinese models processed 21.37 trillion tokens in the week ended June 21, up from 4.37 trillion in late April. By comparison, top US models processed 5.76 trillion tokens during the same week in June, based on the nine most-used models tracked by Jefferies.

The GLM-5.2 launch demonstrates that enterprises no longer face trade-offs between intelligence and privacy, Jefferies noted. Companies are increasingly moving AI workloads from public cloud infrastructure back to local corporate servers.

Memory Sector Remains Bright Spot

Despite downward pressure on token pricing, Jefferies maintains a constructive outlook on memory suppliers, citing the Jevons Paradox. The economic principle suggests that falling token prices should drive higher DRAM prices as usage expands.

The brokerage argues the DRAM industry has undergone structural transformation, making the sector increasingly attractive. SK Hynix, Samsung Electronics, and Micron currently trade at 7.8 times, 6.8 times, and 9.2 times consensus 12-month forward earnings, respectively.

Reflecting this conviction, Jefferies is expanding tech hardware exposure across its portfolios. The firm will add SK Hynix and Kioxia to its global long-only portfolio with initial 4 percent weightings each, while increasing its existing Samsung Electronics position by one percentage point.

Taiwan Rides AI Infrastructure Boom

Taiwan is experiencing exceptional economic expansion driven by AI infrastructure demand. Real GDP grew 14.55 percent year-on-year in the first quarter of 2026, the fastest quarterly growth rate in nearly 48 years, according to Jefferies.

TSMC capital expenditure is forecast to reach $56 billion in 2026 and $65-70 billion in 2027. AI is expected to account for 31 percent of TSMC's revenues this year, underscoring the foundry's central role in the current technology cycle.

Cybersecurity Risks Compound Uncertainty

The Five Eyes Alliance has issued warnings that advances in artificial intelligence could dramatically accelerate cyberattack capabilities in the near future, according to Jefferies. Organizations may have only months to prepare for this threat landscape, the intelligence-sharing partnership cautioned.

This cybersecurity dimension adds another layer of risk to the AI investment thesis, particularly for enterprises moving workloads between cloud and on-premises infrastructure. The convergence of geopolitical competition in AI development and heightened cyber threats creates additional complexity for technology investors navigating the sector.