DeepSeek’s AI Model Raises Concerns Over AI Sector Profitability

Global and U.S. tech markets faced a sharp downturn on Monday as the emergence of DeepSeek, a Chinese AI startup offering a discount model, triggered investor concerns about the long-term profitability of AI and the sector’s insatiable demand for high-end chips.

At the time of writing, tech giants’ stocks, like $AMZN or $META, are down almost 4%, while leaders (possibly, former leaders?) like $NVDA register a 10% slump. The green AI giant sharp drop also influenced connected companies like its supplier Asvantest or its major AI-focused investor SoftBank, also down by 8%.

DeepSeek's assistant, which climbed to the top spot on Apple’s ($AAPL) U.S. App Store, reportedly utilizes lower-cost chips and less data, raising concerns about the viability of high-cost AI infrastructure investment. The Hangzhou startup’s AI model, DeepSeek-V3, used Nvidia’s H800 chips, initially designed with reduced capabilities to comply with U.S. export restrictions on China.

With training costs of under $6 million (vs $500 million for the training of Meta’s Llama 3), DeepSeek’s success challenges the prevailing narrative that AI's growth is dependent on ever-increasing capital expenditures (CAPEX) in high-end chipsets and data centers.

All of the above brought $NVDA to a very interesting trade level, around $128 a share. This also happens to be the lower edge of a 3-month long consolidation. If the price breaks out of it to the downside, it could potentially reach September’s or even August’s lows at $100 – $90 a share. 

Is the AI bubble about to burst?

Follow us on Telegram, Instagram, and Facebook to get the Headway updates instantly.